While data center developers are looking forward to a prolonged period of expansion, they're still being held back by a shortage of available power.

In a mid-year update to its annual Data Center Power Report, Bloom Energy has found that developers are citing higher electricity prices, increased water consumption, and strain on grid reliability as the concerns most likely to influence projects.

Community scrutiny has only ramped up over the past six months, and, as of May, in the US, at least 18 state bills and 86 local moratoriums have been proposed. More than six-in-ten developers say they are planning to source their own power if the grid is unavailable, and nearly a third of these onsite-powered sites are expected to incorporate carbon capture by 2030, reflecting growing pressure to address emissions concerns while expanding power capacity.

"Access to power remains the biggest constraint to data center growth, but it is not the only issue. Community concerns are increasingly shaping which projects move forward," said Natalie Sunderland, chief marketing officer at Bloom Energy.

Developers are looking forward to adding significant capacity between now and the end of the decade, as AI adoption accelerates. Inference is now driving sustained demand for new data center capacity, and now accounts for more than 50% of AI compute, reflecting the transition from model building to real-world applications.

However, there's a gap in readiness that risks slowing AI innovation. Chip developers expect high-density architectures and rack-level DC designs to be adopted in 2028, a full year ahead of data center developers' plans.

"Our findings suggest that solutions that reduce strain on local infrastructure while helping developers bring new capacity online faster – such as clean onsite power – will play an important role," said Sunderland.

"We believe the winners in the next phase of AI buildout will be those that can grow in a way that works for both operators and the communities that host them."

The research follows repeated warnings that the power requirements of AI data centers are growing too fast for national grids to keep up. In a recent report, Wood Mackenzie concluded that with grid transmission build-outs five to ten years away, projects, markets, and consumers are all at risk. Colocated generation and flexible interconnection models aren't actually fixing the problem, said the firm, as these projects face huge technical, regulatory, and economic hurdles.

The problem's only set to get worse, according to real-estate services firm Savills, with only 850MW of power capacity delivered across EMEA since the start of this year, down by 11% on the same period last year.

Gartner has also warned that 40% of data centers could face constraints in power availability by 2027: consumption is expected to grow 160%, with the power required to run AI-optimized servers reaching 500TWh, more than two-and-a-half times as much as in 2023.

The UK is running on fumes in terms of data center capacity, Savills has warned, with surging demand pushing infrastructure to its limit.

Analysis from the real estate services firm found the national vacancy rate has fallen from 27% in 2016 to just 8% in the first quarter of this year.

Things are even tighter in London, where the figure stands at just 7%, thanks to existing cloud ecosystems, network density, and established availability zones. The city now accounts for 1,637 MW of the UK’s total live capacity of 1,803 MW.

“While activity remains anchored to tier-one locations around London, with Manchester cited secondary, beyond these two cities activity is limited reflecting how difficult it remains to deliver network density, occupier ecosystems and timely access to power," said Rupert Duckworth, associate director, Savills Data Centre Advisory.

"Having said that, we are forecasting growth in the North East given its renewable infrastructure and grid connectivity, and by a planning context that is more receptive to large-scale investment. Even so, this is best understood as a medium-term diversification story rather than an immediate release valve for the broader market.”

The firm said that national capacity is likely to increase in the long-term. However, outside the London area, the cloud market is still fairly shallow, as developers favor building on existing infrastructure networks where connectivity is already established and development risks are lower.

In 2024, 191 MW of IT capacity was delivered in the UK as a whole - a record high - followed by an even greater 231 MW last year.

According to Savills, this increase doesn't appear to be continuing, with just 48 MW delivered in the first quarter of this year. Of the 242 MW currently under construction across the UK, around 66 MW is scheduled to complete between Q2 and Q4.

This, the firm noted, shows just how limited the immediate delivery window is, relative to the headline pipeline.

“National supply is expanding on paper, but much of the pipeline remains early-stage and uncertain," said Duckworth. "Pre-letting has become a core feature of the market, with future capacity increasingly committed well before completion."

Capacity constraints could rise

Looking ahead, Savills said the UK is unlikely to see any fall in data center demand, capital, or strategic relevance; the big question is which projects can deliver capacity in a more constrained, regulated, and execution-sensitive environment.

To a large degree, delivery is now shaped by energy strategy as much as by real estate fundamentals.

"Market conversations suggest that, in some of the most constrained West London locations, developers are now planning for multi-year waits for firm capacity, long enough for operators to start to consider alternative power solutions less as a contingency plan than a strategy; using them to bridge the gap to grid capacity and secure position in the right locations,” said Lydia Brissy, Savills' director of European research.

FOLLOW US ON SOCIAL MEDIA

In 2016, Dell Technologies completed its acquisition of EMC. While its name lived on for another four years under the Dell EMC brand, it finally disappeared entirely in 2020.

You may wonder what this wander down memory lane has to do with anything – after all, the title of this article clearly states I was at Dell Technologies 2026 but bear with me. You see, for those of us who have been in this arena long enough, there may have been the sense that the ghost of EMC was haunting proceedings at this year’s annual Dell get-together.

First, a bit of background.

While the EMC Federation comprised a variety of vendor types, EMC itself was firmly in the enterprise storage business; according to a 2014 Forbes article, about 70% of the Federation’s revenue came from the EMC part of the business. While there was undoubtedly the attraction of VMware and Pivotal, acquiring the technology behind VMAX, Isilon, Data Domain and ScaleIO would have been a prime mover for the acquisition.

According to a 2025 retrospective analysis by M&A Watch: “Dell's acquisition of EMC was primarily driven by the need to transition from a PC-centric business model to a comprehensive enterprise IT solutions provider. By 2016, Dell's traditional market--personal computers and low-margin hardware--was stagnating due to declining growth rates and intense competition from HP, Lenovo, and Apple. This acquisition represented a strategic shift into higher-margin enterprise IT markets, including storage, cloud computing, and virtualization.

“Before the acquisition, Dell had a strong foothold in servers and networking hardware but lacked a comprehensive enterprise IT infrastructure offering. EMC's dominance in high-end data storage complemented Dell's existing portfolio, enabling the company to offer a full-stack IT solution.”

In May 2016, what is now the Venetian Conference Center Las Vegas hosted the final standalone EMC World with the strapline Modernize. Across the three days, this was split into three ideas: Modernize the Industry, Modernize Your Data Center, and Modernize Your Business.

With one of the tentpoles of Dell Technologies World 2026 being “The Modern Data Center”, (finally, I get to my point) I can’t help but feel an echo of that final EMC conference.

What is the Modern Data Center, anyway?

2026’s modern data center is, of course, all about AI and having the compute and storage available to carry out large AI workloads without resorting to the public cloud. Dell Technologies launched five new products, within the confines of existing lines, at Dell Technologies World 2026:

• PowerStore Elite. A descendant of Unity, the final storage ranged launched by EMC, PowerStore Elite has been pitched as “the biggest leap forward in the platform’s history”, with built-in AI software to handle load balancing and improve efficiency.
• 18th gen PowerEdge servers, which the company says offer up to 70% better performance. Many of the new servers are advertised as being excellent for AI workloads and high performance computing.
• PowerProtect One, a cyber resilience platform that brings together Dell PowerProtect Data Domain and Dell PowerProtect Data Manager into a single offering.
• Cyber Detect, an AI-powered ransomware detection product that plugs directly into PowerSroew and PowerMax enterprise storage.
• Dell Automation Platform, which itself consists of
  • Dell Private Cloud
  • Dell Distributed PRivate Cloud (previously known as Dell NativeEdge)
  • Dell Automation Studio, and
  • Dell Automation Platform.

The modern data center range, if we can call it that, wasn’t the only chunky enterprise hardware on offer, either. Updates to Dell Technologies AI Factory with Nvidia included the introduction of PowerRack, a rack-scale storage system that comes in three flavours depending on a company’s needs.

There was also the introduction of Dell Deskside AI, which, while it runs of various specialist PCs like the Dell Pro Max with GB10 and and Pro Max with GB300, is a firmly enterprise play – not just a business one.

If Dell Technologies 2025 was the year of the AI PC, 2026 was a hard push to reassert itself the data center as use of AI agents accelerates – and costs rocket in response.

The direction of travel, for the next 12 months at least, seems fairly clear to me: Dell is pulling on its own data center heritage and in particular the heritage acquired from EMC to position itself as the answer to the burgeoning cost of AI usage. This is particularly true of anything that uses a lot of memory or compute.

It’s funny how much and how little change a decade of progress can bring.

Research and consultancy group Wood Mackenzie has warned that the power requirements of AI data centers are growing too fast for national grids to keep up.

With grid transmission build-outs five to ten years away, the firm said, projects, markets, and consumers are all at risk. While data center operators are pursuing colocated generation and flexible interconnection models, these projects face far greater technical, regulatory, and economic hurdles than the industry understands.

"The power sector is fixated on data center flexibility, but that is not the end-game for grid operators or data center operators," said Ben Hertz-Shargel, global head of grid transformation and large loads at Wood Mackenzie.

"Firm grid service is the goal, backed by new transmission superhighways. But there is a lack of awareness throughout the power sector about the technical and regulatory risk confronting colocation projects, and the business risk of conditional interconnections."

Wood Mackenzie's figures see 16.4GW of gas capacity additions per year through 2035 to meet projected demand, despite only 4GW per year having been added between 2023 and 2025.

"Load growth and affordability are in direct opposition in the deregulated markets," said Chris Seiple, vice chairman, energy transition and power and renewables, Wood Mackenzie. "If prices rise to the level necessary to incentivize new generation, it will raise prices for all customers, prompting a political outcry."

In one example, mid-Atlantic grid operator PJM has 78 gigawatts of committed data center load against only 36 gigawatts of accredited generation capacity in its pipeline. The firm's attempting to deal with this by creating one elevated price tier for new resources contracted by large loads, alongside a lower tier for existing resources.

However, Wood Mackenzie warned, this could mean that existing gas and coal plants receiving lower capacity prices may retire, threatening reliability.

In Texas, meanwhile, current market prices of between $30 and $40 per megawatt-hour fall far below the $78-$100 needed to attract new gas generation.

Many data center operators are attempting to avoid the problem through colocation – there's currently more than 90GW of colocated generation now in US interconnection pipelines, indeed. However, the report finds that this is only achievable for the most sophisticated and well-capitalized hyperscalers.

Technical challenges include near-instantaneous changes in AI power demand, which can damage reciprocating engines and gas turbines. And while lithium-ion batteries can be used as a buffer to prevent this, they have a short useful life.

Battery response times need to be extremely short, relying on technology that has not been widely commercialized, and the irregular way in which AI cooling and GPU loads consume power introduces power harmonics, which, if unfiltered, cause equipment to overheat and degrade. Sub-synchronous oscillations, meanwhile, pose a fundamental stability risk to local generators and to distant ones on the transmission system.

And on top of this, the need to protect against downtime exposes data center companies to significant regulatory risk.

There's also, said the firm, a fundamental problem with cost allocation. While grid operators are planning close to $100 billion in transmission investments, much of this could be spread across all existing ratepayers, rather than assigned specifically to a data center.

"Grid operators are positioning flexible interconnections as a stopgap, not a long-term solution," said Seiple.

"The expectation – and often the requirement – is that transmission will eventually provide complete, firm service to large loads. That could cause costs to rise for existing customers if cost allocation methodologies aren't changed and if the data center demand doesn't materialize as forecast."

Data center operators are all too aware of the problem. Research from Savills late last year found that only 850MW of power capacity was delivered across EMEA during the first three quarters of 2025, down by 11% compared with the same period last year. They've been attempting to deal with this by concentrating facilities in the most favorable regions, with some pivoting to emerging markets offering more accessible land and power.

Equinix has announced a major expansion of its Fabric Geo Zones service as the data center provider aims to shore up customer data sovereignty capabilities.

Fabric Geo Zones is designed to ensure that enterprise data remains within defined jurisdictions. According to the firm, unlike solutions built within a single cloud or delivered as software overlays, Fabric Geo Zones enforces sovereignty at the network layer.

"Sovereignty can't be a setting you configure inside a single cloud. Global enterprises must enforce sovereignty at the network layer, across every cloud, provider and path simultaneously," said Arun Dev, vice president of digital interconnection at Equinix.

"Equinix Fabric Geo Zones is the only solution that enforces geographic boundaries as a property of the network itself. Traffic either flows along compliant paths or it's blocked. That's why enterprises across industries trust Equinix to move data across clouds without compromising sovereignty."

Fabric Geo Zones is aimed primarily at workloads requiring a high level of compliance – financial institutions, healthcare organizations, government agencies, and the like.

Global companies, meanwhile, can automatically apply jurisdiction‑specific routing rules to meet GDPR, LGPD, APRA and other regional requirements across their operations.

Data sovereignty in the spotlight

Inadvertently moving sovereign data across borders that organizations are legally required to respect is a growing risk when working in hybrid multi-cloud environments, said Courtney Munroe, founder of Apex Research.

"Businesses are facing one of the most complex global regulatory environments in history while at the same time facing huge pressure to deploy new technologies," he said.

"A global enterprise operating under GDPR in Europe, LGPD in Brazil, and APRA in Australia simultaneously needs different data routing rules for each jurisdiction, with every outage, failover, or congestion event a potential compliance violation."

Recent research from Kiteworks found that one-in-three organizations has experienced a sovereignty-related incident in the past 12 months. One-in-eight of these incidents consisted of unauthorized cross-border transfers.

In Europe, 44% cited provider sovereignty guarantees as their top barrier to cloud adoption – the highest of any region – despite near-universal GDPR compliance.

“Organizations across every region we surveyed are spending millions on sovereignty compliance, scoring high on awareness, and still getting hit by breaches, unauthorised transfers, and government access requests,” said Dario Perfettibile, EMEA GM of GTM and customer operations at Kiteworks.

Fabric Geo Zones is available in preview in countries including Australia, Brazil, Canada, Japan, Switzerland, the UK, and the US, with European Union availability expected in June.

FOLLOW US ON SOCIAL MEDIA

Data center operator Stellanor has dramatically boosted its UK presence by taking over eight data centers from UK managed services provider Redcentric in a £123 million deal.

The new facilities, serving around 450 enterprise customers, mean the company now has eleven data centers across the UK, with a total of 39MVA of secured grid capacity.

The firm said the deal will enable it to deliver high-density colocation capabilities, along with advanced cooling systems, robust security, and sustainable operations powered through renewable energy procurement across its network.

“This acquisition represents a transformative step in building the UK’s leading urban data center platform,” said Michael Tobin CBE, Chairman of the Board, Stellanor.

“We’ve scaled from two to eleven facilities in nine months, backed by the infrastructure fund managed by DWS Group. This enables Stellanor to serve the accelerating demand for enterprise colocation and AI-ready infrastructure across the UK, with further expansion into Ireland and the Nordic regions planned."

With the deal complete, Stellanor is now carrying out a series of coordinated infrastructure upgrades across the network, including high-density power capabilities, advanced cooling systems, and enhanced fiber interconnection.

This, the company said, will enable the platform to support AI inference, machine learning and real-time analytics workloads, without any service disruption for existing clients.

Redcentric, meanwhile, said the divestment will allow the company to solely focus on its core managed services business, strengthening the group's balance sheet by significantly reducing debt.

"As a dedicated UK MSP with vast experience in supporting organisations in highly regulated markets, we’re perfectly positioned to use this expertise to help business navigate an increasingly complex technology landscape,” said CEO Michelle Senecal De Fonseca.

“We’re confident our data center clients will benefit from Stellanor’s platform approach and DWS backing. We look forward to continuing to develop our partnership together and building on the strong ties we have built through this sales process.”

Stellanor has grown rapidly since its formation a year ago, backed by German investment firm DWS with the acquisition of two data centers in London from Colt Technology Services.

It acquired its Hemel Hempstead site, supporting GPU and AI chip workloads, from Imagination Technologies just last month in a sale-and-service-back deal that sees Imagination continuing to operate from the site as a fully managed client of Stellanor.

FOLLOW US ON SOCIAL MEDIA

It's no secret the demand for data processing and storage is exploding. Even before AI, between 2010 and 2025, the amount we created and stored went from around 2 to 181 zettabytes – with much stored and processed in data centers.

Of course, generative AI has massively increased data center demand. Last year Deloitte predicted data centers would consume 4% of global electricity by 2030.. According to the US Department of Energy, that figure's already 4-5% in the US.

Water use, too, is a major concern for data centers. Large data centers can consume the same amount of water for cooling as a town of 10-50,000 people, per data from the Environmental and Energy Study Institute. It’s clear something must be done to prevent putting additional strain on beleaguered electricity grids and water-stressed regions.

What’s being done now to address data center environmental impact and what are our options for the future?

The data centers of tomorrow

Data centers face steep power constraints. In theory, if you could completely clean up the power supply for data centers, you could dramatically reduce their overall emissions.

Solar, wind and hydro power have all been used for data center power and cooling here on earth, with renewables offering an increasingly cost-positive route to creating green data centers.

Some have suggested that space-based data centers might be the next step – and several serious plans for orbiting data centers are afoot, including one between an Australian compute module manufacturer and an Indian space-based infrastructure company. Nvidia has even released computing platforms intended for orbital data centers.

But despite the rate of advancement in space industries, we're a long way off orbital computing replacing ground-based server farms. The extra load on power needs and the environmental impact need action now.

Despite a long history of political and public mistrust, some are seriously considering nuclear. In 2024, Microsoft signed a 20-year power purchase agreement (PPA) with energy provider Constellation, to power data centers with nuclear energy from a reopened Three Mile Island plant. The same year, AWS announced a $650 million plan to acquire a data center next to the 2.5GW Susquehanna nuclear power station in Pennsylvania, in a deal with Talen Energy. The original plan was to power a 480MW data center, though this was later scaled back to just the initial 300MW on site.

While nuclear doesn't reduce total power or cooling loads, it has the potential to ease the demand on the local grid. Small modular reactors could achieve the same, at a lower cost and in a more scalable form factor, beating some of the most prevalent criticisms of nuclear power.

Most of the efforts towards reducing energy and water consumption are centered around optimizing current data center components. A Google data center in a former paper mill in Hamina, Finland, includes tunnels to the sea – pumps bring ocean water in where it cools components, taking the heat back out to sea again. Another small data center is situated in the Loire River in Nantes, France, cooling provided by the natural flow of the water. Future plans include an offshore model that does the same on the ocean swell.

Adnan Masood, chief AI architect at technology business consultancy UST, calls such systems “early proof that chillers [fan, freshwater and pump systems] aren't destiny”.

A body called The Ocean Sewage Alliance, which forges partnerships with industry to solve environmental problems related to the seas, says one possible solution is to use reclaimed or treated wastewater instead of potable or drinking water sources. Spokesperson Larissa Balzer says the adoption of reclaimed water for cooling is a “growing trend among major tech companies”.

Small wins

In trying to save power, water and waste in data processing, it's a good time to remind ourselves that all data isn't created, stored or accessed equally. There's a whole class of information that can work with lower power inputs, low latency and intermittent connectivity.

A study published in the January-March 2025 issue of IEEE Pervasive Computing talked about the possibility of taking discarded and retired smartphones and wiring them together to form little data centers. With processors and storage already on board, the promise is to kill two birds with one stone – reduce the breakneck demand for first generation data center equipment and do something about the growing e-waste problem.

Amit Chadha, CEO and managing director of L&T Technology Services, agrees about what he calls 'micro server farms'. "They can take on lighter but essential workloads like IoT data aggregation, local caching or microservices,” he says.

“They won't replace GPU-heavy AI clusters, but they can free up capacity in data centers for more intensive applications."

On a much larger scale this process already takes place with retired supercomputer parts, which are repurposed for less intensive workloads after they're deemed outdated.

Ezra Hodge, who works in AI and Data Centers at EMA Partners, likewise imagines idle smartphone chipsets in a globally distributed, low-cost parallel processing grid. "For emerging markets like education, or early-stage AI startups, it provides a way to access massive horsepower without the overhead," he says.

Former electric vehicle (EV) batteries may also be used as power sources in the near future. Reduced range and power and safety issues means EV batteries aren't used in cars by the time they reach about 75% of their original capacity, but data processing has far lower power input needs, particularly for off-peak workloads.

Batteries can be clean replacements for diesel generators in backup power, and they can be taken offline and recharged according to workload schedules to take up the slack. This year, battery recycling provider Redwood Materials used second life storage for a Nevada data center, the second largest battery-powered grid in North America.

And with the amount of e-waste in the world, the sky's the limit. Retired laptops and gaming consoles have been suggested to build micro-clouds for small organisations like schools. Displays from old phones or tablets can become monitoring dashboards for server racks. Microphones, accelerometers/gyros and cameras can become monitoring stations warning operators about server room sounds or vibrations, watching for fan or motor failures.

When do we want it?

Space-based data centers, nuclear power and mass e-waste recycling programs are going to take time, and the race for alternative ways of handling data better is a problem for today. What can we do with what we have now?

First is the need to handle data with the equipment we're using today more efficiently. Just one is separating compute from storage.

"'Helper' components like SmartNICs can move, filer, encrypt or prepare data while main processors, the 'brains' doing queries, analytics or AI, can focus on the heavy lifting," says Tobie Morgan Hitchcock, CEO and co founder at AI-native database firm SurrealDB.

If you decouple your systems, Hitchcock says the storage layer can deliver smaller, cleaner, more focused information to operate on. "It lets you scale the storage and compute components separately based on how much each really needs."

In some ways, this is similar to edge computing. While edge doesn't separate computer and storage, it does a lot of data preparation at the source where it's collected and stored – sorting, indexing, cleaning and packaging it neatly before being sent off to the data center for action.

"Embedding compact, high-efficiency compute modules into existing environments like factories or offices doesn't only help reduce latency but effectively turns everyday spaces into mini data centers, alleviating pressure on central facilities," says Chadha.

Reducing the demand on data centers might also remind you of a longstanding principle to reduce the amount of data we keep and store. "What's being overlooked in the discussion is the fact that companies are storing data they don't understand or don't need," says Maggie Laird, president of data management software provider Pentaho.

She says enterprises are storing exabytes of stuff spread across clouds, apps, endpoints and shadow IT, much of it ungoverned and unused and leading to ROT (redundant, obsolete, or trivial) data. "Companies are spending millions storing data they can't use, undermining the very AI initiatives meant to unlock value, while bloated storage needs are driving up energy costs and straining water access where data centers are built.".

As Masood puts it; "Not all data deserves a 24/7 heartbeat. Archive deep, run hot where it matters.

Turning to gas to power data centers means Microsoft's carbon footprint from such infrastructure will more than double.

That's according to an analysis from Stand.earth Research Group (SRG), which argues that three methane gas projects announced in the last month to power AI data centers for the company will send its carbon footprint skyrocketing 160%.

“Microsoft makes great claims about its climate credentials, and has shown real leadership in the past, which makes it even more disappointing to see the company so wholeheartedly turn toward fossil fuels – proposing nearly 5GW of new fossil fuel power in just the first few months of 2026 – while still promising to be on track for its climate goals," said Rachel Kitchin, senior corporate climate campaigner at Stand.earth.

"The gap between what Microsoft says, and what Microsoft does, seems to grow wider by the day."

The analysis comes amid a boom in data center construction driven by AI, which is already driving up energy demand. The rush has caused some tech companies to turn to alternative fuel sources, including nuclear, but also led to an increase in gas-powered data centers.

In recent years, increased generation of data center power using fossil fuels has sharply increased tech emissions.

Indeed, back in 2024, Microsoft's emissions surged 29% off the back of data center building, while Google admitted its emissions were up 50% due to AI energy demand. Despite such concerns, Google-owner Alphabet last month called for more power generation to feed AI.

In response, campaigners have demanded the UK government enforce stricter emissions rules for data centers.

Gas-powered AI

The SRG analysis points to a trio of data center projects powered by methane gas. The first is an exclusive agreement with Microsoft and Chevron for 2.5GW of power for a data center, while the second involves Microsoft leasing a 900MW facility in Abilene, Texas, which applications reveal has plans for methane gas powered turbines.

SRG calculates that those three projects alone will emit 15.52 million metric tons of carbon dioxide emissions – despite Microsoft's own plans to become carbon negative, including with data centers, by 2030.

ITPro contacted Microsoft for comment, but had not yet received a response at the time of publishing.

On-site power

Last month, Big Tech players in the US signed an agreement with the government to "build, bring or buy" their own power to data centers, in the hopes that would keep electricity prices down. However the agreement didn’t include any provisions to use renewable sources.

That, alongside dated electrical grid infrastructure, has led some data center companies to generate energy on site. In some cases that's achieved by co-locating with existing power stations such as nuclear facilities, but in others that's involved turning to gas.

SRG notes that at the end of 2024, data centers made up 5% of demand for methane gas power, but just a year later, that has leapt to 39%.

"Proponents of these fossil fuel-powered hyperscaler projects may, for example, point to the opportunity to leverage methane that would otherwise be burned off as part of the conventional extraction process, which is a convenient perspective to hold if you’re an oil-and-gas giant seeking to entrench new dependents," said Kitchin.

"However, if you’re one of the world’s leading technology companies that has for years proudly touted its decarbonization goals, the visions don’t line up."

"Renewables are the cheapest form of energy to build, aren’t subject to the massive price volatility compared to fossil fuels, and don’t drive up people’s utility bills," she added. "Microsoft has shown that it knows how to build renewables quickly and at scale, so to get back on track, it needs to cancel these pollution-producing contracts and recommit to around-the-clock clean energy."

Frustrated by the rocketing number of requests to connect to the National Grid, the UK government is planning a new system that could see data centers prioritized.

Over the last six months, the queue for demand connections to the transmission network has grown by 460%, thanks largely to speculative applications.

These are swamping the pipeline, delaying connections for strategically important projects and leading to waits of up to 15 years.

The government’s answer is to prioritize certain projects, including AI data centers, AI Growth Zones, and industrial sites that can deliver growth and jobs.

"The UK is home to Europe’s leading AI ecosystem, with firms like Nscale and Wayve pulling in billions of pounds worth of investment," said AI minister Kanishka Narayan.

"Delivering data centers - which we’re turbo-charging through our AI Growth Zones - is fundamental to this work, and all of this relies on access to the grid. These timely reforms will help us move at pace, to seize AI’s potential to help build a wealthier and fairer Britain."

Along with prioritizing connections for key projects, the reforms aim to tackle speculative applications and create a fairer, more efficient system by making it harder to join and remain in the queue.

Ofgem is now looking at how to achieve this, with measures likely to include increasing the financial requirements for developers in the queue – for example through deposits or fees which would be payable if key milestones are not met.

AI Growth Zones prioritized

The government will also publish a list of strategically important projects, particularly AI Growth Zones, which will be at the front of the queue as capacity is freed up or created.

It will also move to a “strategically aligned” process that prioritizes connections for facilities that are close to parts of the grid with high capacity, reducing the need for unnecessary new infrastructure.

"The surge in demand applications shows the strength of investment interest across Great Britain, but the demand connections pipeline must reflect projects that are credible, ready and committed to progressing," said Kayte O’Neill, COO of the National Energy System Operator (NESO).

"We are committed to working with government, industry and Ofgem to prioritise strategically important projects, while removing speculative applications. This will ensure data centres, industrial sites and vital public services can access clean, reliable power, while also supporting growth, innovation and jobs across Britain."

The initiative has been welcomed by the Confederation of British Industry (CBI), which has complained that delayed grid connections have consistently blocked business investment for several years.

“Prioritizing strategically important projects that bring jobs, investment and economic growth to communities across the country is the right approach to address a defunct and oversubscribed process that’s left major investors waiting up to 15 years to connect," said Jennifer Beckwith, CBI senior manager for energy transition.

“Alongside planning reforms, speeding-up grid connection is essential to give both domestic and international investors greater confidence that large-scale, high-value and economically significant projects can move more rapidly from development to delivery."

FOLLOW US ON SOCIAL MEDIA

Google is building a data center where the only water used will be in the kitchens, the company claims.

The tech giant announced a new data center under construction in Wilbarger County, Texas will use air cooling instead of water, with "clean" power added to the grid even before the data center comes online.

Google’s announcement comes amid a wider pushback against the impact of data centers – in particular in Texas – which are being built in a rush to meet growing demand sparked by AI.

Common concerns include increasing energy prices through high consumption and the fact these sites use significant volumes of water.

President Trump this week told data center building companies they need to start sorting out their own power supplies to avoid energy prices continuing to increase for people.

Google said that, once complete, the new data center will use "advanced air-cooling technology, limiting water consumption to only critical campus operations like kitchens."

Indeed, Google said the aim is to replenish more local water than it consumes and is working with local water charities. Google's head of data center public affairs for Texas, Andrew Hart, said the company was "proud" of its close collaboration with local leaders on this front.

"Texas is at the center of the world’s AI leadership, and our expansion into Wilbarger County is a testament to the state's incredible talent and infrastructure," he said in a statement.

The industry doesn't need to use water for cooling, but often does in a bid to cut power use. Evaporative cooling, for example, consumes vast amounts of water.

Alternatives include air cooling, which uses liquid carbon dioxide, while other operators use liquid immersion and closed loop practices. In short, there are options.

Google’s new data center power plans

When it comes to power, Google is working with local energy generation companies Wilbarger County through power purchase agreements (PPAs) to boost clean energy on the Texas grid.

The tech giant claimed it will add more than 7,800 megawatts of net-new energy generation and capacity.

Indeed, the new site will be co-located next to "clean" power plants built and operated by AES, which will also run retail and other services at the campus. Google said the co-location is a "key strategy" for easing demands on the grid while speeding along buildout of data centers.

The aim is for the project to be "power first" with the plants coming online ahead of the data center. That, along with $30 million announced in November to help local residents hit by rising energy costs, should help keep power affordable, the company said.

"We are proud of our long-standing partnership with Google and thrilled to bring them this advanced stage ‘power first’ clean co-located project,” said Kleber Costa, Chief Commercial Officer, AES US Renewables.

"By capitalizing on AES’s extensive development expertise and years of hard work in collaboration with the exceptional local leaders and communities, we are bringing to Google a site that is fully ready for construction, with land and interconnection agreements, as well as co-located generation under 20-year PPAs."

Texas pushback

Google's efforts to ensure its data center won't use much water and will generate clean energy come amid pushback against these facilities, largely focused on those two issues.

That's particularly true in Texas. In central Texas, the city of San Marcos recently rejected a data center project in the area after resident complaints — and a local judge — raised concerns about water use. As one activist said: "We are out of water."

Two other similar infrastructure projects remain in the works in the nearby area, with one in nearby Round Rock approved despite "hours of public opposition" at a hearing on the issue.

It's not just about water and energy, though. In north Texas, a "city sized" data center project faces criticism for the risk it causes to local Dinosaur Valley State Park, home to well-preserved dinosaur tracks, though water and power use are concerns too.

In Hood County, residents are also concerned about construction-related noise. As one resident noted to local reporters: "I get to pay property taxes to be audibly assaulted."

City and state authorities are quick to note that these projects not only bring in jobs but also tax revenues, important for areas facing restricted and deficit budgets.

FOLLOW US ON SOCIAL MEDIA

Data center operator Pulsant has made a £10 million investment to build a new, high-density data hall in Milton Keynes.

The 1.2 MW site is purpose-built to handle high-density computing tasks including intensive AI and machine learning workloads.

According to Pulsant, the main target market for the new site includes organizations operating in the financial services, healthcare and biotech, IT, and gaming sectors.

“The £10m expansion of our Milton Keynes data center is another big investment in our digital platform to meet hunger for high density compute power,” said Rob Coupland, CEO at Pulsant.

“UK digital infrastructure is facing unprecedented demand. With AI-ready capacity in short supply, bringing high performance, flexibility and choice to regional locations is critical."

The site will join Pulsant’s national platform of 14 UK data centers connected by a 400Gb-capable network. It opens low-latency access to more than 1,600 cloud services, network providers, and business partners, the company said, providing connectivity for global traffic routing and access to international networks.

Pulsant eyes out-of-London gains

Pulsant is touting its out-of-London location as a major selling point for the facility. It’s sited within the UK’s Oxford-Cambridge tech supercluster, home to businesses with 570,000 employees generating £135 billion in annual turnover.

Meanwhile, the firm claimed it can offer just two milliseconds’ latency to London Docklands and Slough and is highly resilient, with 10 carriers on site, including access to Megaport.

"For organizations looking for ultra-low latency, international connectivity and UK sovereign compute power, Milton Keynes is a great option compared to constrained and costly London data centers which lack the opportunity for expansion,” said Coupland.

"Our unique platform gives local, national and international clients the flexibility to circumvent some of the risks associated with the London cluster, while maintaining high performance, resilience and connectivity.”

Pulsant completed a £187 million debt refinancing round last summer following a year in which it acquired SCC’s Birmingham and Fareham data centers. The Birmingham site added 2 MW of capacity and 25,000 sq ft of white space, with Fareham offering a 3 MW capacity.

The company said it now plans to roll out its high-density model to other key regions.

A data center boom time for the UK

Data centers are experiencing something of a boom in the UK. Late last year, the UK government committed £500 million to strengthening domestic AI capabilities through a series of AI initiatives, including a new Sovereign AI Unit due to launch later this year. 

It launched new AI Growth Zones, in which data centers will get significant discounts on their electricity bills, if they are located in areas that could help to reduce pressure on the energy network.

Developers, meanwhile, could be supported to connect their own high voltage lines and substations to power their data centers, rather than having to wait for network operators to do it.

FOLLOW US ON SOCIAL MEDIA

Efforts by Amazon Web Services (AWS) to expand data center infrastructure across Europe face severe delays due to sluggish grid connection practices, a senior company figure claims.

In a recent interview with Reuters, Pamela MacDougall, AWS’ head of energy markets for EMEA, said the length of time it takes to get a connection to energy grids had become a major factor when deciding where to invest in new infrastructure.

Indeed, she said delays in Europe had hampered plans by the company to build data centers in high-demand regions, warning that delays were “challenging our growth aspirations."

"There's a misalignment. We want to expand and grow within two years,” MacDougall said. However, the AWS exec noted that grid connections can take up to seven years in Europe.

Last year, power constrains forced AWS to reportedly scrap plans for an industrial plant in Ballycoolin, Ireland that was going to build server racks.

Data from the International Energy Agency (IEA) shows that securing a grid connection ranges from two to ten years, depending on the country. MacDougall added that the problems weren't limited to a few countries.

"We're finding more and more across Europe that certainty of the delivery date has continued to be delayed."

An AWS spokesperson told ITPro it’s charging with plans to "rapidly expand its infrastructure across Europe, with ongoing investments in edge locations and data center regions in more than 20 countries”.

“We remain committed to expanding at pace to meet growing customer demand,” the spokesperson added.

Figures published as part of Amazon's Q3 2025 earnings report revealed the firm had added more than 3.8 gigawatts of power in the 12 months prior. The spokesperson told ITPro it plans to continue adding capacity across 2026 and into 2027.

AWS isn’t alone in grid woes

The IEA noted last year that the data center project pipeline equals 130% of existing capacity – but based on its own modelling, capacity will likely grow by just 70% because of delays and power constraints.

One solution is building data centers in regions with existing grid capacity, rather than focusing on existing hubs as is normally the case. The IEA added that smart grid tech can help boost capacity while battery systems can help balance demand, but noted that further investment in transmission infrastructure remains necessary.

Indeed, it's not just Europe that is struggling to keep up with data center power demands. The US is also facing "significant concerns" over the impact of data center growth on regional energy grids.

In the UK, the situation is much the same. In 2024, former National Grid CEO John Pettigrew called for "bold action" to boost energy infrastructure to keep up with surging power demands.

"Future growth in foundational technologies like artificial intelligence and quantum computing will mean larger scale, energy-intensive computing infrastructure," Pettigrew told attendees at the Aurora Spring Forum in Oxford.

This predicament for big tech shows no signs of improving any time soon, either, and experts have warned aging energy grids could pose a serious challenge to ambitious infrastructure projects.

Two-fifths of data centers have faced power constraints, according to research from Gartner, while analysis by McKinsey projects data center energy demands to triple by the end of the decade.

Big tech courts alternative energy sources

Big tech companies such as AWS, Microsoft, and Google have all been exploring alternative energy sources to power infrastructure projects.

While renewables have been a long-term focus for cloud computing providers, the advent of generative AI – and the associated pressure on data center infrastructure – brought nuclear power back into vogue.

AWS, for example, shelled out $650 million for a data center co-located with a nuclear power plant – though the project faced significant pushback from regulators.

Elsewhere, Microsoft signed a 20-year power purchase agreement (PPA) with energy provider Constellation in a move that will power data centers with nuclear energy.

The partnership involves the reopening of ‘Unit 1’ of the ‘Three Mile Island’ site in Pennsylvania.

FOLLOW US ON SOCIAL MEDIA

Imperial College London is firing up the UK's first direct liquid-cooled (DLC) AI deployment, through a deal with Digital Realty.

The DLC, high-performance computing (HPC) AI platform uses bespoke direct liquid-cooling technology supplied by Lenovo.

This allows it to deliver significantly higher compute performance per rack while reducing energy consumption, according to Digital Realty.

“Direct liquid cooling is becoming essential for modern AI and HPC systems, where performance and sustainability must go hand in hand,” said Kate Steele, director of HPC at Lenovo EMEA.

"We’re pleased to support this collaboration by providing a solution designed specifically for these intensive workloads, helping Imperial’s researchers unlock new possibilities while reducing the environmental footprint of high-performance computing."

Hosted at Digital Realty’s AI-ready Woking data center, the deployment will allow Imperial researchers to train larger and more complex AI models and support the acceleration of data-intensive simulations.

It could also help shorten the time from discovery to real-world impact across disciplines ranging from drug discovery and robotics to climate modelling and advanced materials.

"Advanced digital infrastructure is now as critical to research as laboratories and instrumentation,” said Jenny Rae, chief information officer at Imperial College London.

"This deployment gives our researchers access to the scale and performance they need to tackle some of the most complex scientific and societal challenges."

Imperial College is ramping up AI research

Imperial College is already one of the UK’s largest Tier 3 HPC university environments, researching AI, climate science, healthcare, and engineering.

This new deal builds on the Imperial College Intel Corporation Lenovo (ICICLE) initiative, first announced several years ago. This set out Imperial’s ambition to develop advanced, energy-efficient HPC and AI capabilities in partnership with leading technology providers.

“This deployment demonstrates what can be achieved when leading academic institutions, a global technology powerhouse, and digital infrastructure providers work closely together,” said Séamus Dunne, managing director, UK and Ireland, Digital Realty.

"Imperial’s research ambitions require an environment designed for extreme performance, resilience and sustainability, and we are proud to provide that platform at our Woking data center. Together, we are helping to create the technical foundations needed for the next generation of AI discovery in the UK."

The project was completed quicker than expected, with the contract signed in July 2025 and the fully operational environment handed over at the end of last year. This included structural floor reinforcement, installation of the DLC infrastructure, system integration, and commissioning.

"Working with Digital Realty provides us with a secure, resilient and sustainable platform, while Lenovo’s direct liquid-cooled technology allows us to grow our AI capabilities responsibly," said Rae.

"Together with our technology partners, we are building an environment that will serve Imperial’s research community for many years to come."

FOLLOW US ON SOCIAL MEDIA

The Irish cabinet has approved a new plan aimed at attracting more investment from large energy users like data centers – and there's been strong pushback.

The Large Energy User Action Plan (LEAP) aims to encourage investment in energy intensive sectors such as semiconductors, pharmaceuticals, precision engineering, and hyperscale data centers.

Thanks to their scale and energy consumption, these industries would benefit from coordination with national infrastructure planning, the cabinet said.

"The approach set out in the Large Energy User Action Plan will enable regions across Ireland to attract investment in the next generation of strategic industry, promoting long-term economic development and providing further employment across the regions," said minister for enterprise, tourism and employment Peter Burke.

"This will enhance Ireland’s proposition as a world‑class place to do business in and as strategic knowledge-intensive regional hub for the ICT sector, where a secure, sustainable energy system supports innovation, investment and long‑term industrial growth."

The plan encourages the siting of data centers in green energy parks, co-located with renewable energy resources, including offshore wind.

However, it also allows them to be developed outside these locations, in some cases until beyond 2030, in line with existing planning arrangements. One such area is Dublin, which has raised concerns among opposition parties.

“I have major concerns with plans which went to cabinet this morning which seek to lift the ban on the development of data centers in Dublin, which speak of 'sustainable' data center development and co-locating data centers with renewable energy resources," said Social Democrat environment spokesperson Jennifer Whitmore.

"The production of renewable energy sources should be prioritized for residents and small businesses – approving new data centers only means that we continue to chase our tail in terms of climate goals and energy demand."

There are also concerns as to whether, even with the 'green' provisions, Ireland's already heavily-loaded grid will be able to cope.

Labour climate spokesperson Ciarán Ahern noted that data center power demands have grown significantly over the last decade.

In 2015, for example, they accounted for around 5% of national electricity demand – today that figure stands at 20% and is projected to reach 30% by the beginning of the next decade.

"Families and small businesses expect a reliable electricity system, affordable bills and a government that protects them from reckless decisions that load costs onto their shoulders,” Ahern commented. “Instead, they are being asked to subsidise an industry whose electricity demand is growing at a staggering pace."

Data center energy costs dip while households rise

A key talking point in the infrastructure plans rests on current household energy prices. Ireland’s recently-approved Price Review 6, for example, sets network charges on energy bills for the next five years.

Pa Daly, Sinn Féin spokesperson for climate, the environment, and energy noted this will see costs rising for households, while falling significantly for data centers.

Sinn Féin plans soon to bring forward legislation covering a prioritization framework for grid connections.

"Instead of the current ‘first come, first served’ model which data centres can often easily muscle in on, we believe who can connect to the grid and when should be based on the public good," Daly said.

“These decisions should be based on the best interests of the Irish people, not the corporate bottom line."

FOLLOW US ON SOCIAL MEDIA

Google-owner Alphabet is buying data center and energy firm Intersect in a deal valued at $4.75 billion.

The move comes amid a race to rollout data centers – and the energy systems required to power them – to fuel AI.

In the last year alone, data center investment topped $61bn, though concerns continue over whether energy systems will be able to keep up with demand.

As part of the deal, Alphabet will acquire the $15bn in data center and energy assets Intersect already has in operation or is working to build, Reuters noted, adding that by 2028 Intersect had planned to have 10.8 gigawatts of power either up and running or in development.

"Intersect will help us expand capacity, operate more nimbly in building new power generation in lockstep with new data center load, and reimagine energy solutions to drive US innovation and leadership," said Sundar Pichai, CEO of Google and Alphabet.

Google targets energy innovation

Google already owns a minority stake in Intersect. However, moving forward Alphabet said Intersect would operate separately from itself and Google, keep the company brand, and continue to be led by CEO Sheldon Kimber.

The aim is for Intersect's team to work closely with Google's technical infrastructure division on joint projects, including a data center and power site already under construction in Haskell County, Texas, the duo said.

Intersect noted it will work to develop emerging technologies to diversify energy supply while also boosting Google's data centre projects.

"Intersect has always been focused on bringing innovation to the industry and we look forward to accelerating at scale as part of Google," said Sheldon Kimber, founder and CEO, Intersect.

"Modern infrastructure is the linchpin of American competitiveness in AI. We share Google’s conviction that energy innovation and community investment are the pillars of what must come next."

The deal doesn't include currently operating assets in Texas or California, which run as separate companies and will be bought out by existing investors, Kimber said in a blog post.

Powering AI

With infrastructure expansion gathering pace worldwide, experts have questioned whether supply will be able to keep up with demands placed on data centers as a result of AI.

Moreover, similar questions have been raised over whether current energy systems are up to the task of meeting future power demands. Data center power demand is predicted to top 106GW in the next ten years, according to a BloombergNEF report.

Google has already said it will pause AI data centers to help balance overloaded grids, and signed deals for renewable energy sources – including with modular nuclear reactor developers.

Kimber said deals like Alphabet's acquisition of Intersect are "pillars of what must come next".

"AI today is stuck behind one of the slowest, oldest industries in the country: electric power," he wrote.

"The country has racks full of GPUs that can’t be energized because there isn’t enough electricity for them. The grid is a patchwork operating system that’s been running for a century. An engineering miracle in its time, but not built for the AI era."

Bring your own power

Kimber said the US model is increasingly "bring your own generation".

"Companies like Google are thoughtfully building AI the responsible way," he added. "In the case of Intersect, that has meant renewables paired with flexible backup sources and energy storage, often on the same site as the data centers themselves. This approach can significantly accelerate speed to power and unlock scarce transmission capacity."

While Google has invested in renewables — including a 99% carbon-free data center in Norway — its greenhouse gas emissions have leapt 48% since 2019, driven by electricity use for data centers.

FOLLOW US ON SOCIAL MEDIA

Nearly 80% of data centers around the world are located in unsuitable climates, with 10% in regions that are too hot – putting pressure on cooling systems and raising safety risks.

In 21 countries, all data centers are located in climates that are hotter than the 18-27 Celsius range recommended by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE).

Non-profits RestOfWorld and Climate Central plotted temperature data from the Copernicus Climate Data Store – a project organized as part of the European Union’s efforts to open-source climate data – against locations from Data Center Map, a widely referenced resource and marketplace for data center-related services.

Crucially, researchers found that of the 8,808 operational data centers worldwide, nearly 7,000 are located in areas outside the optimal range. While the vast majority are in regions with average temperatures that are colder than the range, 600 are located in areas where average annual temperatures are above 27 C.

In 21 countries, including Singapore, Thailand, Nigeria, and the United Arab Emirates, all data centers are located in areas with average annual temperatures of above 27 C, along with nearly all data centers in Saudi Arabia and Malaysia.

Nearly half of Indonesia’s 170 data centers are in places that are too hot, while the figure for India – a key market for big tech and social media companies – is around 30%.

"Extreme heat not only places a heavier burden on cooling systems, it also reduces the efficiency of power transmission, raising the risk of outages," the researchers said.

Cooling technologies are getting better

Cooling technologies are improving, and big tech firms are touting their efforts to reduce energy use. Google, for example, has claimed that its data centers use about 84% less overhead energy for every unit of IT equipment energy, partly through the use of its DeepMind AI to optimize cooling.

Microsoft, meanwhile, is now using a new data center design it claims optimizes AI workloads and consumes zero water for cooling.

Microfluidics bring liquid coolant directly inside the silicon, removing heat more efficiently. The company’s lab-scale tests, it said, have shown that microfluidics performed up to three times better than cold plates at removing heat, depending on workloads and configurations involved.

Microsoft also reduced the maximum temperature rise of the silicon inside a GPU by 65%.

Elsewhere, this summer AWS said that it had shifted from air to liquid cooling. Its direct-to-chip approach puts a cold plate directly on top of the chip, with the liquid running in tubes through the sealed plate, absorbing the heat and carrying it out of the server rack. The liquid is then cooled and recirculated.

The RestOfWorld researchers point out that possible solutions in hotter countries could include building data centers underwater or underground, or using nuclear energy and hydrogen fuel cells.

FOLLOW US ON SOCIAL MEDIA

Data center M&A and investment hit record levels in 2025, according to new research, with more than 100 data center transactions in the first 11 months of the year.

The value of M&A, asset sales, and equity investments is already higher than the full-year 2024 record of just under $61 billion, according to S&P Global Market Intelligence data.

The boom is being fueled by a combination of factors, including hyperscaler expansion, private equity interest, and a surge in debt financing.

"We're seeing enormous growth in data centers and even acceleration," said Wim Steenbakkers, managing director at investment relations firm ING, during a recent press briefing.

"We do see a trend where the US has taken a lead, with Middle East growth eclipsing Europe at the moment."

The US has been leading the pack in terms of total deal value over the last two years, with the US and Canada totaling about $160 billion since 2019. In the same period, Asia-Pacific investment has hit nearly $40 billion, and Europe has totaled $24.2 billion.

Global data center footprints are projected to expand at a faster rate over the next five years than it has done in the previous five, spurred by demand for energy and compute-intensive AI workloads.

Bullish investment expected to continue

According to S&P Global, many investors are turning to new builds. Hyperscalers are making major investments, leveraging strong balance sheets and massive free cash flow to build new sites, while also looking for outside capital in the form of debt and collaboration with financial sponsors.

"The number of data centers being built by hyperscalers nearly doubled in 2024 from a post-pandemic low, while their total utility power more than doubled, according to 451 Research data," said S&P Global senior research associate Iuri Struta.

"This growth is expected to extend into 2026, with 114 hyperscaler data centers planned or under construction."

Struta noted that hyperscalers, frontier AI labs, and chipmakers are teaming up to buy assets or finance their construction, in a sign of the significant capital requirements needed to achieve the desired data center growth.

Meta, for example, partnered with Blue Owl Capital to build its 2.4 GW Hyperion data center, while late last year Microsoft and MGX Fund Management teamed up to create a $100 billion fund that would invest in data centers and power infrastructure.

Earlier this year, McKinsey researchers concluded that, by 2030, data centers will require $6.7 trillion worldwide to keep pace with the demand for compute power.

Those equipped to handle AI processing loads are projected to require $5.2 trillion in capital expenditures, while those powering traditional IT applications are projected to require $1.5 trillion in capital expenditures - nearly $7 trillion in total capital outlay.

"Investors will need to understand demand projections amid uncertainty. Companies should assess AI computing needs early, anticipate potential shifts in demand, and design scalable investment strategies that can adapt as AI models and use cases evolve. Second, investors should find ways to innovate on compute efficiency, McKinsey said.

"Third, they can build supply-side resilience to sustain AI infrastructure growth without overextending capital. This will require investors to secure critical inputs such as energy and chips, optimize site selection, and build flexibility into their supply chains."

FOLLOW US ON SOCIAL MEDIA

Pure Data Centres (Pure DC) has signed what it says is this year's largest standalone hyperscale data center lease in Europe.

One unnamed hyperscale customer is leasing an entire 78MW campus, situated in Westpoort, Amsterdam, and purchased by Pure DC on a long leasehold from the Port of Amsterdam.

Pure DC is investing over €1 billion to acquire and develop the 5.68-acre site, saying it has secured planning permissions and 100MVA of power via a private substation.

According to the firm, the development will create more than 1,000 jobs and support many more roles through the extended supply chain.

Once complete, the data center will provide around 80 permanent skilled jobs, including engineers, maintenance, security, and administrative staff.

The company said it plans to work with local schools and universities to provide training, career guidance, and outreach programs, supporting local charitable organizations, and working with community partners on environmental conservation projects.

Inside the Pure DC Westpoort campus

Named AMS01, the data center campus will consist of three 85m towers, powered by a private substation with a firm connection into the 50kV grid.

Each of these towers will house 26MW of data halls designed to support high density compute with high efficiency cooling. It will achieve the Netherlands energy efficiency target of a Power Usage Effectiveness (PUE) of 1.2.

The private substation is already constructed and live, with development of the data halls expected to begin in January next year.

“Amsterdam is one of Europe’s most constrained markets for digital infrastructure and Pure DC has again demonstrated its ability to unlock new low-latency, high-quality capacity," said Dawn Childs, CEO of Pure DC.

"This deal demonstrates how our specialist teams have the creativity and approach to deliver compelling proposals – delivering solutions for local authorities, potential customers, and our supply chain."

Who is the mystery customer?

Pure DC currently has two data centers in London, in Brent Cross and Park Royal, along with others in the UAE, Saudi Arabia, Spain, Indonesia, Ireland, and Finland.

It also recently received final planning approval for the first phase of a €400 million data center and private sub-station in Madrid, expected to have a final capacity of 70MW.

All told, Pure DC said it has more than 500MW of capacity live or under development.

Exact details on who the unnamed customer is are yet to be revealed. However, according to Research & Markets, the Netherlands currently has only two major hyperscale operators, Google and Microsoft.

The country's data center market was valued at $1.23 billion in 2024 and is projected to reach $3.39 billion by 2030, rising at a CAGR of 18.41%. It held a 2.8% share of the European market last year.

However, tight constraints on data center construction in Amsterdam caused by grid congestion have hampered the sector's growth, along with space constraints and concerns over energy consumption.

FOLLOW US ON SOCIAL MEDIA

Data center capacity will probably hold up in the face of increasing demand, according to Goldman Sachs – but that's by no means guaranteed.

The bank's analysts are forecasting that AI’s share of the overall data center market will double to 30% over the next two years, eating into the share from conventional cloud workloads.

By 2030, the firm said in a new report, overall power consumption from data centers looks set to jump 175% from 2023 levels – more than the 165% that the firm previously forecast.

The financial services giant has drawn up four possible scenarios to show how the data center boom may play out.

Goldman Sachs' base case is that the balance between supply and demand is set to narrow significantly over the next 18 months, with occupancy predicted to remain at peak levels through 2026.

After that, it expects supply constraints to ease off, with occupancy rates gradually falling back to around 90% by 2028 and then leveling out.

“What we’ve seen over the last nine months is that both supply and demand forecasts have gone up, but the demand side is increasing a bit more than the supply forecast," said Jim Schneider of Goldman Sachs Research.

In another scenario, demand for compute capacity to support data-hungry AI video content could jump significantly, while a new generation of GPUs could also consume more power than expected.

"The power efficiency is improving, but power demand is outstripping what a lot of people thought," Schneider says.

​In this scenario, data center occupancy rates would exceed 100% in peak regions by 2030 and come in 17 percentage points higher than the base case.

Another possibility, the firm said, could see AI demand falling, particularly from end users. If demand drops by 20%between 2025 and 2030, occupancy would run at eight percentage points less than the base case – leading to excess supply in data centers, and potentially forcing operators to lower their lease rates.

Deteriorating macroeconomic factors could also play a role in determining the trajectory of the industry, Schneider added.

“If the economy slows a bit and companies decide they need to be a little tighter with the usage of cloud services and they optimize, then growth rates could decelerate.”

Most of the capacity in data centers still supports run-of-the-mill cloud computing, with cloud and traditional workloads currently making up about 85% of data center demand.

However, Goldman Sachs warned that in a scenario where corporate spending on these basic services declines, occupancy in data centers could fall by four percentage points from the base case, even though AI demand remains steady. Efforts to optimize the use of cloud services could also reduce demand.

If heavy demand tips the occupancy rate into a significant lack of supply, researchers said hyperscalers will be leasing or utilizing their capacity as fast as they can build it.

"We believe occupancy will continue to tighten through the medium term before the market loosens," said Schneider. "And the market could remain tighter for longer, but that is yet to be seen."

MORE FROM ITPRO

• Is BESS the key to data center energy demand?
• Lack of power supplies hitting data center construction
• Where next for data center cooling?

As data center power density and uptime expectations rise, it's predicted that we'll see a rapid growth in the use of battery energy storage systems (BESS) in the next three to five years.

While there are utilities working on flexible load tariffs for which data center operators could use storage when called upon instead of curtailing, many are turning to off-grid solutions because interconnection for new loads is taking too long, says Allison Weis, Global Head of Energy Storage, Power, and Renewables at Wood Mackenzie.

For on-site power generation, liquid natural gas (LNG) is expected to be the power source of choice in the short-term. However, operators are also implementing geothermal, wind, and solar solutions. Then there's small modular reactors (SMRs), which may come out top in the long-term. The variability associated with renewable sources is a key driver for BESS, as fluctuations will need to be balanced. Therefore, solar and wind need to be paired with storage to increase their availability throughout the day.

"Batteries are well suited for shifting energy across short timeframes – hours rather than days. That makes them effective at absorbing excess renewable output," says Dr Denis Phares, professor at the University of Nevada and CEO of Dragonfly Energy, adding that they're also good for handling short-time oversupply from firm generation and reducing the need for fast ramping or curtailment.

Whatever sources an operator chooses for a data center's power generation, storage will be an essential part of the mix, says Weis. This is because data center operators recognize that battery storage is no longer just about backup power but about securing the 'license to operate' in a constrained market.

Stacey Yin, Senior Director Analyst at Gartner, says the most compelling reasons for adoption are now grounded in four drivers: enhanced resilience, cost optimization through peak shaving, improved sustainability and, most importantly, mitigating current grid limitations.

"In many regions the utility grid simply cannot supply the requested capacity without the buffer BESS provides."

Falling costs fuel growth

BESS have become more viable solutions due to battery costs coming down by more than 90% since 2010, notes Max Schoenfisch, Power Sector Modeler at the International Energy Agency (IEA). This has led to a growth in demand both for utility scale and behind the meter storage, especially in the US, China and Europe – and particularly the UK – but he expects this will broaden in coming years.

"In the Middle East we're already seeing Saudi Arabia emerging as a major player. We're also seeing a lot of development in Chile, which we expect will broaden to other South American markets in the future. India will also become an important country for battery storage over the next couple of years."

Data center use cases

When it comes to data centers, the highest impact use cases of BESS have graduated from passive backup to active grid participation and hybrid power architectures, says Yin. However, currently, Gartner's mainly seeing a move towards hybrid backup, where BESS reduces reliance on diesel generators.

"For example, at Google's facility in St. Ghislain, Belgium, it replaced traditional diesel generators with a large-scale battery system. This not only provides back-up power, but also enables the facility to operate as a carbon-free asset during outages," she explains.

Other use cases include AI power smoothing. With rack densities hitting 100kW+ for AI training, BESS can be deployed to handle massive, millisecond-level power spikes (transients) that AI clusters generate, acting as a 'shock absorber' to protect the grid connection, says Yin.

"What a BESS could do well is manage any peaks in compute – and therefore energy – demand, if such peaks exist. The presence of spikes depends on the type of data center and the operations being carried out. AI training is 'bulky', but when the system reaches certain checkpoints, the demand drops to near zero, and a BESS could help smooth this variable profile," adds Bogi Hojgaard, Energy System Expert at The Carbon Trust consultancy.

"Inference – when the model is already trained and responding to customers – is more likely to correlate with daytime working or daylight hours, which could lend itself to a BESS shaving benefit. The purpose of this shaving goes beyond optimizing electricity costs, which could also be addressed in part by onsite generation coupled with BESS."

Then there's also frequency regulation and revenue generation. "BESS assets are now being used to provide ancillary services, such as frequency response, to the grid. For example, Microsoft's Dublin facility successfully piloted a grid-interactive BESS that provides backup while interacting with Ireland's grid to balance wind intermittency," notes Yin.

BESS storage duration

In terms of the amount of storage BESS can currently offer, most solutions provide two, four, or eight hours. While these are lithium ion, sodium ion may take a piece of this market in the future, notes Schoenfisch, "particularly if lithium prices rise again, as sodium is so abundant. However ongoing technical development would be needed for it to take a bigger share of the market."

While there is interest in longer duration BESS, such as vanadium or zinc-air flow batteries, and thermal storage, Schoenfisch questions how competitive that would be on a cost basis.

"They would be competing with options such as pumped-storage hydro and other flexible, low-emissions generation technologies, so there are still some uncertainties."

According to Weis, BESS are still very expensive for solving 48-hour backup at 99.999% uptime guarantees, even mixed with renewables "as you could have a renewable lull for two days." Therefore, "the traditional datacenter application that demands 'five 9's' will not use storage yet as an alternative to backup diesel or gas. However, the big tech companies continue to invest in alternative long duration storage that may get to a cheaper price point eventually," she says.

A future-ready energy strategy

While there are still some questions to be answered, many analysts agree that BESS is evolving from a transitional solution to a core asset for data centers.

"While BESS initially served to bridge the gap between unreliable grids and decentralized renewables, it's now indispensable," says Yin. "Intelligent simulation technologies and digital twin integrations are accelerating this transition, proving that BESS is required not just for 'green' targets, but for fundamental operational resilience."

For product leaders and IT decision-makers, this convergence of operational resilience, cost benefits and grid limitations underscores the importance of investing and integrating BESS technologies immediately, she says, as doing so safeguards operations against unplanned outages and positions the facility at the forefront of a modern, future ready energy strategy.

The energy footprint of US data centers is expanding far more rapidly than previously anticipated, driven by a surge in massive-scale artificial intelligence (AI) projects that threaten to overwhelm existing power infrastructure.

According to a new report released this week by BloombergNEF (BNEF), titled AI and the Power Grid: Where the Rubber Meets the Road, demand for power from US data centers is now projected to reach 106 gigawatts (GW) by 2035.

The figures represent a significant revision of industry expectations. The 106GW forecast is a "36% jump from the previous outlook" published by BNEF just seven months ago, underscoring how quickly the generative AI boom is altering the physical requirements of the technology sector.

"Colliding with grid realities"

The report paints a concerning picture for IT leaders and utility providers, suggesting that the pace of digital infrastructure buildouts is outpacing the electrical grid's capacity.

"This boom in data center demand is colliding with grid realities," the report states.

BNEF analysts warn that specific regions are already facing precarious supply crunches. In the PJM Interconnection, a major regional transmission organization covering 13 eastern states and DC, data center capacity is forecast to reach 31GW by 2030. This figure "nearly matching the 28.7GW of new generation" expected to come online during the same period, leaving virtually no buffer for error.

The situation is similarly tense in Texas. The report notes that in the Electric Reliability Council of Texas (ERCOT) grid, reserve margins could fall into "risky territory after 2028," signaling that while short-term growth might be manageable, long-term reliability is in question.

The rise of the mega-campus

A key driver of this energy spike is the shift toward significantly larger facilities. As tech giants race to train more complex Large Language Models (LLMs), the physical footprint of the average data center is ballooning.

Of the nearly 150 new projects added to BNEF's tracker over the last year, "nearly a quarter exceed 500 megawatts." This represents "more than double last year’s share" of such massive facilities.

"The massive growth rate in data center power demand reflects more than a surge in the number of data centers in the pipeline; it also highlights the new centers' size," the report notes.

Geographic shift beyond "Data Center Alley"

The power constraints are forcing a geographical restructuring of the US internet infrastructure. Northern Virginia, long known as the world’s "Data Center Alley," is "nearing saturation," prompting developers to look elsewhere.

New projects are increasingly migrating south and west. BNEF highlights that developers in Texas are "transitioning former crypto-mining sites into AI data centers," capitalizing on existing connections to power and population centers. Meanwhile, hyperscale campuses are appearing in "suburban and exurban zones," typically within 30 miles of major cities in states like Georgia and Ohio.

This sprawl is reliant on robust connectivity. "Fiber optics form the digital spine enabling this sprawl," the report said, allowing these massive campuses to operate outside urban cores while maintaining the low latency required for high-performance computing.

For CIOs and IT procurement leaders, the report serves as a warning: the era of abundant, easily accessible power for digital expansion may be drawing to a close, marking an "inflection moment for US grids."

Difficulties accessing adequate power supplies are slowing the expansion of data centers, according to a new analysis from real-estate services firm Savills.

The company said it's only seen 850MW of power capacity delivered across EMEA since the start of this year, down by 11% compared with the same period last year.

Meanwhile, new take-up this year reached 845MW, around half the power capacity leased in 2024.

Savills puts this down to the limited availability of new facilities coming to the market, rather than a lack of demand. The underlying strength of demand, it said, is reflected in the total contracted power capacity, which has risen to nearly 14,500MW, up 12% year-on-year. Around a quarter of take-up is now pre-let compared with less than 20% three years ago.

As a result, occupancy rates across the region have climbed to 91% in the third quarter this year, up from 87% in the same quarter in 2022.

"The persistent imbalance between surging demand and restricted supply continues to underpin rental values. Following three years of sharp increases, average rents have stabilised across the region," said Cameron Bell, director, EMEA data centre advisory at Savills.

"Nonetheless, with accelerating AI-related requirements, rising energy costs, and sustained construction inflation, further upward pressure on pricing is widely anticipated for the rest of 2025 and beyond."

Partly because of these difficulties, data center supply tends to be concentrated in certain areas, said Savills. Established hubs grew strongly, with capacity in France up 15%, Germany 10% and the UK and Ireland up by 9%. But there was also growth in non-traditional locations such as Portugal (60%), Saudi Arabia (49%), Spain (25%), UAE (20%), and Sweden (17%). The Netherlands saw just 6% growth, thanks to an ongoing government moratorium on new developments.

"Geographically, we have seen demand remain highly concentrated. Despite discussion around location agnostic strategies when it comes to AI, very few such projects have translated into transactions," said Bell.

"Instead, operators are doubling down on existing availability zones, reinforcing consolidation in areas with established hyperscaler footprints, reliable energy supply, and space for scalable growth."

And there are other issues too. Across EMEA, data centre build costs – including land, shell, electrical and mechanical systems, cooling, fire safety and fit-out – now range between $7.3 million and $13.3 million per megawatt of commissioned IT load.

The sharpest annual cost increases are seen in Vienna, up by 28%, Warsaw, up by 25%, Stockholm, at 18% and Copenhagen at 17%.

"We are seeing developers responding with forward purchasing and closer ties with suppliers, whilst some are pivoting to emerging markets offering more accessible land and power," said Marc Edmondson, director and data centre expert in the building and project consultancy team at Savills.

Data center power consumption has become an increasing concern, with Gartner warning this time last year that 40% of data centers could face constraints in power availability by 2027. It forecast a growth in consumption of 160%, saying that by then the power required to run AI-optimized servers will reach 500TWh, more than two-and-a-half times as much as in 2023.

MORE FROM ITPRO

• More transparency needed on sprawling data center projects, activists claim

Data centers are cropping up across the world in a mad race to provide processing infrastructure deemed necessary to power the AI revolution.

Yet governments keep waving projects through without talking to locals or warning residents about potentially damaging impacts like water use or noise and air pollution.

That's according to a panel of speakers at this year's Mozilla Festival taking place in Barcelona, Spain – all of whom had investigated data center projects as part of their journalism or activism.

Paz Peña, an independent consultant and a Mozilla Senior Fellow, noted that big tech companies aren’t solely in the crosshairs for criticism – this is down to governments to regulate.

"It's not just about holding accountable companies, but government," she said.

Peña pointed to a case in Santiago, Chile where the community fought a Google data center via court, which found that the government should have informed locals about the impact on water use.

"The government was saying it's a corporate secret," she said, but the court disagreed — again holding the government to account rather than the company, though Google did have to rejig its plans with climate considerations in mind.

Consumer shaming

Tessa Pang, impact editor at Lighthouse Reports, predicted that messaging from the tech industry will start to include "consumer shaming" as projects globally continue at a rapid pace.

If you use tech products or AI, how can you complain about a data center in your own backyard? Peña agreed that it can feel "uncomfortable" to criticize tech giants and their infrastructure demands when you yourself use their services.

While Pang called for us all to use AI in ethical ways – and only when really needed – she stressed that consumers are not the "enemy", but the focus should be on government and companies instead.

This is particularly true when it comes to energy, now that data centers have become the biggest clients for energy companies, Peña added.

"Basically the energy transition is based on the needs of big tech," she said. "This is incredibly dangerous."

The panel agreed that there is a growing sense that AI and the data centers powering them may be a bad deal for many people – especially as energy prices start to rise – but added that it's difficult for local governments to push back.

Demanding better

Pablo Jiménez Arandia, a freelance journalist covering the area, called for attendees to imagine you're a small town mayor and Google or AWS shows up promising to spend billions on a data center that will provide local jobs – is it even possible to say no?

"We have to be aware of this power imbalance," he said.

That's made harder by big tech lobbying, and by well-oiled public relations machines. While those two activities add to the challenge, Peña noted that big tech's desire to maintain positive public relations suggested they were at least attempting to look good, offering a route to successful campaigns for activists.

Ultimately, they want to be seen as good actors, not bad actors, she noted.

The point isn't just outright banning data centers, but ensuring local communities have a well-informed voice that's defended by governments, and encouraging a better way of rolling out necessary infrastructure.

"This is not the only AI possible," said Arandia. "This is the AI of a group of elites in Silicon Valley. There are alternatives to build sustainable, fair systems."

MORE FROM ITPRO

• Colt DCS to expand West London data center campus in £2.5bn investment
• Lenovo is getting creative with future data center ideas
• Why tech giants are hitting the brakes on costly data center plans

Lenovo claims traditional data centers are not fit for purpose and is proposing some radical solutions for future designs.

Research from the company shows many are failing to meet sustainability targets, with 46% of IT leaders admitting their current infrastructure doesn't support energy or carbon-reduction goals.

That comes despite the fact 92% said they prioritize technology partners who reduce energy use and carbon footprint.

Meanwhile, an overwhelming 99% of IT and C-level decision makers across EMEA said data sovereignty will be critical in the coming years.

Unsurprisingly, AI is a key consideration in terms of future data center sustainability and design. Nine-in-ten believe that AI will significantly increase organizational data usage in the next decade, and 62% expect AI and automation to have the greatest impact on IT strategy.

This means that data centers of the future will have to be designed with AI workloads in mind, and many simply aren’t up to scratch right now.

“The data center of the future will be defined by how effectively it can scale for AI, deliver on sustainability targets, and operate with maximum energy efficiency,” said Simone Larsson, head of enterprise AI for EMEA at Lenovo.

“In EMEA, data sovereignty stands out as a particularly urgent priority, shaped by complex regional regulations and heightened scrutiny from CIOs and C-suites alike."

Lenovo is taking 'the cloud' to the next level

So what's the answer? Well, according to Lenovo, it's all about thinking outside the box. The company said it's been working with engineering firm AKT II and architect Mamou-Mani to dream up some ideas.

The most radical is the 'Floating Cloud' – perhaps taking the term 'cloud' rather literally, by suspending the data center in the air. These would operate at an altitude of 20-30km and make use of 24/7 energy from solar power. Pressurized closed liquid cooling loops would avoid any air pollution.

Closer to reality, other ideas touted by the company include the use of modular, stackable brick or pod systems of data centers linked to city needs. These would be sited close to water sources such as rivers or canals, the company said.

When sited next to geothermal power sources, designers suggested these could be referred to as 'Data Spas'.

Elsewhere, there's also the possibility of utilizing disused tunnels, bunkers, or transport systems, reducing land use while allowing for facilities to be set up in central locations with lower impact.

Such sites would benefit from enhanced security resilience, while the subterranean location would create a naturally efficient heat management system.

Some of these ideas are already starting to appear in the real world. Portus Data Centers, for example, operates an underground data center in the heart of Luxembourg, powered by 100% renewable energy sources.

Siting data centers near water sources is already pretty standard, and in areas with significant geothermic activity, many already take advantage of the free power.

Microsoft, G42, and KenGen, for example, are building a data center at the Olkaria geothermal field in Kenya.

MORE FROM ITPRO

• Why meeting data center demand is becoming a global challenge
• Gas-powered data centers: what's behind the boom?
• Majority of English data centers use less water than a 'typical leisure center'

Colt DCS has secured approval to expand its Hayes Digital Park data center campus in West London as part of a £2.5 billion investment.

The company says it plans to build three new hyperscale data centers and an innovation hub in partnership with Brunel University as part of the expansion scheme.

These new data centers, dubbed London 6, 7, and 8, will be powered using 100% renewable energy, according to Colt.

Power contracts for the development have been secured with the National Grid and are expected to provide an additional 97MW in power to the broader site, bringing total capacity to 160MW.

Xavier Matagne, Chief Real Estate Officer at Colt DCS, said the move marks “another important milestone” for the company’s expansion and broader UK digital economy.

“Data centers are a cornerstone of digital transformation,” Matagne said. “With this expansion, we can help power innovation, support the AI revolution, and contribute to the energy transition.”

Colt DCS eyes 2029 launch date

Colt said construction is expected to start mid-2026, with the first of three new data centers, London 6, scheduled to go live in early 2029.

Upon opening, Colt said the new facilities are projected to create over 500 permanent jobs, with the company planning to train more than 50 technical apprentices over a ten-year period.

Elsewhere, AECOM will lead development of the design proposals for the planned Innovation Hub. The hub will be used as a “base for innovation and community engagement”, according to Colt, hosting events for the local community.

“Our new campus in Hayes, including the Innovation Hub in partnership with Brunel University, will drive community value, from reusing heat for district heating to creating jobs, skills, and long-term investment,” Matagne commented.

“As one of the few operators capable of delivering new capacity in this area of London over the next decade, we’re proud to be helping power the UK’s future economy in a sustainable and inclusive way,”

Councilor Steve Tuckwell, Hillingdon Council’s cabinet member for planning, housing, and growth, welcomed innovation hub plans, describing the move as an “exciting new development that will help foster economic growth” in the local community.

“It will equip residents and smaller local businesses with the right skills, affordable workspaces, and opportunities to thrive,” he said.

MORE FROM ITPRO

• ]Equinix announces multi-billion pound data center investment in ‘huge win’ for UK
• Colt DCS starts work on third Frankfurt data center
• Why meeting data center demand is becoming a worldwide challenge

Equinix has bought an 85-acre plot in Hertfordshire, with plans to build a £3.9 billion data center with 250 MW of processing capacity.

The Hertfordshire Campus facility will double the company's processing power in the UK, and represents its biggest financial investment in Europe. All told, it will deliver more than two million square feet of data center space.

With construction expected to start in 2027 and set for completion in 2030, Equinix said it expects the project to directly generate 2,500 local jobs. Once it's fully operational, it will create more than 200 permanent roles, the majority of which will be highly skilled.

KPMG, meanwhile, has estimated that direct and indirect employment could contribute roughly £120 million in wages to the UK economy. It also projected the Hertfordshire Campus could add up to £3 billion in annual Gross Value Added (GVA) during the construction phase, and up to £260 million in annual GVA once operational.

"This £3.9 billion investment is a huge win for Britain. It will give businesses – from life sciences to high street banks – the ability to connect to thousands of other businesses across the world in an instant, powering our AI ambitions, boosting growth and creating hundreds of well-paid jobs,” said Liz Kendall, secretary of state for science, innovation and technology.

"This is about making sure the UK is at the forefront of the digital revolution and ensuring that every community benefits from the opportunities this new technology brings."

Inside the Equinix data center campus

The facility will be powered entirely by renewable energy, with dry cooling ensuring that water consumption will be comparable with conventional office buildings, according to Equinix.

More than half of the land will be kept as as open space, and the company has committed to creating new ecological habitats which will deliver a biodiversity net gain of at least 10%.

The site was bought by Equinix from DC01UK.

"The ambition behind the DC01UK project is extraordinary, and we are delighted that the vision for this project will be taken forward by leading experts committed to delivering it to the highest standards, maintaining our ambition, realizing the community benefits, and embedding themselves in the growth and future of Hertfordshire," said James Craig, director of DC01UK.

"For us, it’s a proud moment that also marks the beginning of an exciting new chapter as we continue to progress other UK data centre development opportunities to drive further innovation and investment in this space across the UK.”

Along with more than 250 data centers overseas, Equinix already operates 14 data centers across the UK, and has more than 1,200 UK-based employees. Of its 10,000-odd customers, more than 1,300 are UK-based, it said.

MORE FROM ITPRO

• Majority of English data centers use less water than a 'typical leisure center'
• Europe’s largest data center approved for development in UK
• UK data center industry growth faces serious challenges

Data center developer Argyll Data Development has signed a deal with AI infrastructure firm SambaNova to create the UK’s first sovereign AI cloud powered entirely by renewable energy.

Sited in the Killellan AI Growth Zone, a 184-acre green digital campus on Scotland’s Cowal Peninsula, the facility will use wind, wave, and solar energy generated on-site.

It will deploy SambaNova’s air-cooled SN40L systems, in which each rack draws roughly one-tenth the power of traditional GPU systems, eliminating the need for liquid cooling, the company said.

“Together with SambaNova and our strategic partners, we’re building a sovereign AI infrastructure powered by renewable energy, demonstrating that sustainability and scale can go hand in hand. Our goal isn’t just to make AI greener, but to make it competitive, compliant and cost-effective,” said Peter Griffiths, executive chairman at Argyll.

“This project gives UK enterprises the ability to innovate responsibly, securely and within our own borders, in full alignment with national AI ambitions.”

The first phase of the build will have 100 to 600 megawatts of capacity, scaling to over two gigawatts when the facility is complete. A private-wire renewable network and battery storage will enable 'island-mode' operation, with plans for future grid integration in progress.

Waste heat from the data center campus will be used to support vertical farming, aquaculture and local district heating, using a closed-loop, circular design.

“Argyll is a blueprint for scaling AI responsibly. By pairing renewable power with high-performance, energy-efficient computing, it shows what sustainable AI infrastructure can achieve,” said Rodrigo Liang, CEO and co-founder of SambaNova.

“With SambaNova’s chips-to-model platform, we’re enabling large-model inference with maximum performance per watt, while helping enterprises and governments maintain full control over their data and energy footprint.”

Argyll said the project is expected to enable £15 billion in total investment, create more than 2,000 construction jobs a year along with 1,200 permanent positions, and contribute £734 million in annual Gross Value Added (GVA) to the Scottish economy.

This week, SambaNova also announced a similar deal in Australia, launching SCX, the country’s first ASIC-based sovereign AI cloud powered entirely by renewable energy.

In Germany, meanwhile, Infercom is launching the country’s first Inference platform, providing fully EU-compliant, GDPR-safe AI infrastructure for startups, enterprises and government agencies, powered by SambaNova.

“Infercom’s new Inference-as-a-Service platform, built on SambaNova, directly addresses the growing demand from European startups, enterprises, and public sector organizations for sovereign, energy-efficient AI infrastructure,” said Liang.

“Europe needs sovereign, enterprise-grade AI infrastructure that delivers on regulatory, security, and performance expectations - and we’re proud to help make that a reality.”

MORE FROM ITPRO

• UK data center industry growth faces serious challenges
• Gas-powered data centers: what's behind the boom?
• Majority of English data centers use less water than a 'typical leisure center'

Basildon Council has granted permission for a new £1.3 billion datacentre in Wickford in Essex.

To be built by developer Caineal, the project is expected to create more than 600 construction-related jobs and 120 highly skilled operational roles. Meanwhile, Basildon Council said it could potentially support five indirect jobs for every direct position, through supply chains and service providers.

"This is a fantastic opportunity for our borough," said councillor Jessica Power, cabinet member for jobs and skills.

"It places Basildon at the heart of the UK's AI revolution and will act as a magnet for high-value, technology-led businesses, driving skills development and creating long-term employment opportunities."

Construction is expected to start on the 1.8-hectare site next to the A127 and Old Nevendon Road as soon as the final planning formalities are completed.

"This is a chance for the area to become a powerhouse in the AI data economy," said Caineal spokesperson John Bourne. "We're proud to be investing in Basildon and helping to position it as a centre of technological excellence."

There's little detail about the new facility, but planning documents show that it's set to include a substation and energy centre, along with the creation of public open space.

The original plans for the development have been scaled back from the initial application in December 2023, following local concerns about its impact on the green belt. However, the new plans include a commitment to on-site Biodiversity Net Gain and the creation of a new local nature reserve to the east of the site.

Caineal has also promised financial contributions toward education, employment, skills development and sustainable transport to sweeten the deal.

Essex data centers

Basildon Council said it was keen to see investment in digital infrastructure and AI across South Essex. And the county is already home to several data centres, including Fujitsu London East in Grays and ChunkyChips Data Centre in Basildon.

Google is also planning a facility on a 52-hectare site in Thurrock, including up to four datacentres. Nscale, meanwhile, is building a data centre in Loughton – claimed to be the UK's largest sovereign AI datacentre – that's expected to house the UK's largest supercomputer, recently announced by Microsoft.

The government is keen to encourage the building of more data centres, as part of its efforts to position the UK as an AI hotspot. At the beginning of this year, it announced the creation of dedicated AI Growth Zones to speed up planning for AI infrastructure.

These areas will be given faster planning approvals for the rapid build-out of data centres, along with better access to the energy grid.

MORE FROM ITPRO

• What is a green data center
• Google opens doors on UK data center ahead of Trump visit

Google has officially opened the doors of its $1 billion data center in the UK, part of a £5 billion investment into local infrastructure and the tech giant's first data center in the country.

The opening of the data center in Waltham Cross, in Broxbourne on the edge of London, comes as US President Donald Trump is set to arrive in the UK for a state visit that is expected to include a series of deals between the US and UK on technology and nuclear power.

Google said in a statement that the "state of the art" data center would be used to help meet demand for AI-powered services including Google Cloud, Workspace, Search, and Maps.

The £5 billion investment includes data centers like the one at Waltham Cross, as well as Google's capital expenditure, R&D, and AI-related engineering over the next two years in the UK, including DeepMind, the British-founded AI science company bought by Google in 2014.

"Google's investment in technical infrastructure, expanded energy capacity and job-ready AI skills will help ensure everyone in Broxbourne and across the whole of the UK stays at the cutting-edge of global tech opportunities," said Ruth Porat, President and Chief Investment Officer at Alphabet and Google, in a statement.

Inside Waltham Cross

Google bought the 33-acre site in Waltham Cross in 2020, saying last year that the project "represents an investment of $1 billion dollars in the country".

Technical details about the data center have been sparse, but Google has been keen to share information about energy plans – no wonder given the sensitivity around energy and water use by data centers.

Earlier this week, The Guardianreported that planning application documents for Google's next UK data center, in Thurrock in Essex, shows it will take up 128 acres and is expected to emit more than half a million tonnes of carbon dioxide a year – equivalent to 500 short-haul flights a week.

For the newly-opened Waltham Cross site, Google said it was designed to "minimize" environmental impact, including by using more advanced air cooling technology rather than traditional water-based cooling systems.

"The facility uses advanced air-cooling technology to limit water usage to domestic use and is also equipped to support off-site heat recovery, meaning heat from the data center can be re-routed and provided free of charge to help warm local homes, schools or businesses," the company said in a statement.

Google has also signed deals with local energy providers to encourage renewables, including a previously announced partnership with Scotland's ENGIE for wind generated energy as well as teaming up with Shell to use battery storage systems to help address intermittency issues with renewable sources.

Last year, Google said UK operations were "on track to operate at or near 90% carbon-free energy in 2025." This year, Google has adjusted its goal.

"Between the Shell alliance and Google's other clean energy initiatives, Google's UK operations are projected to run at or near 95% carbon-free-energy in 2026."

Keeping locals on board

The rise of AI has sparked a wave of data center construction, which isn't always welcome in locations where construction takes place.

In Waltham Cross, locals last year complained that noise from the construction site was keeping them awake at night, ruining local views and lowering house prices. The Daily Mail quoted one resident as saying: "I moved here for peace and quiet… It won't benefit me. It won't benefit locals."

Google said it had set up a community fund to be managed by Broxbourne Council to help support local economic development, and was "providing direct support" for local charities and social enterprises.

Councilor Corina Gander, Leader of Broxbourne Council, said in a statement provided by Google: "The Council has worked closely with Google to maximise the positive impact of the development. Google is heavily investing in community-based projects and is making an important contribution to the local economy."

MORE FROM ITPRO

• Majority of English data centers use less water than a 'typical leisure center' as operators embrace new cooling methods
• $200 billion to build a data center? Soaring AI demands mean costs could skyrocket within a few years
• Plans announced to resurrect former steelworks as a ‘green’ data center

Data center operators looking to capitalize on rising generative AI demand face a growing challenge – how to power these behemoths, which will consume more energy than existing regional grids are able to provide.

This has become the primary motivator for operators to adopt on-site generation, as Jon Healy, MD EMEA at IT consultancy and service provider Salute, explains.

“Traditional hubs are at or near saturation, which has created long connection queues with waits of typically seven-to-ten years and in some cases 13-15. The International Energy Agency (IEA) estimates that globally, nearly 20% of planned data center projects risk delay due to grid bottlenecks, and as a result developers are building ‘behind the meter’ primary generation. This offers speed to market in months rather than years.”

But gas generators also come with higher emissions and the potential for more pushback from those who live close to data centers.

Natural gas – a quick fix

Sustainability goals have driven many data center operators towards renewable power, such as solar and geothermal, but there are some opting for a more traditional choice – natural gas.

Crusoe, for example, has to date ordered 29 natural gas generators to power its data centers per Data Center Dynamics. Then there’s xAI, which uses natural gas generators at its Colossus data center in Memphis, USA. The company went so far as to double the number of turbines in use, a move opponents say was done without the required permits. Since then, the local county health department has approved 15 . Now looking to build Colossus 2 in the same city, xAI recently purchased a former natural gas power plant close to the site.

In discussions around data center power sources, Alvin Nguyen, senior analyst at Forrester has heard natural gas valued for its ease of transport to sites and lower emissions compared to other fossil fuels. But he tells ITPro that gas is mainly chosen because it’s the fastest and easiest way to generate your own power. Turbines and engines can be set up and begin providing reliable energy quickly, while renewable power generation can be intermittent without costly storage – and takes time and money to be set up.

Furthermore, in some cases natural gas may not be subject to the same level of regulatory scrutiny, he notes, which can be leveraged to operators’ advantage.

“If a generator’s mobile, perhaps mounted to a platform or on wheels, it may be allowed onsite for a certain length of time – even in parking lots or next to buildings. These enable you to get what you need now rather than waiting a few years to get the infrastructure in place. But I’m not sure how long that ‘loophole’ will be allowed to continue.”

Locations and geography may also make natural gas the most suitable power source. For example, it makes the most sense in areas with stranded natural gas resources, like flared gas from oil fields that would otherwise be wasted.

“If the gas is going to be burned anyway, better to turn it into AI compute than just smoke,” says Neil Sahota, CEO of simulation technology company ACSILabs. “This is the logic behind Crusoe’s flare-gas-to-compute model. Remote regions with poor grid access but pipeline availability also benefit by using gas to bootstrap datacenter operations until renewables or small modular reactors arrive. In some disaster-prone areas, gas offers a more stable fuel source than diesel, which is hard to deliver post-disaster. Ultimately, the key is aligning with regional realities,” he adds.

Trading sustainability for speed

While natural gas is the most environmentally-friendly fossil fuel, it still emits CO₂ when burned – and causes significant methane leaks when extracted and transported. Community opposition could therefore be an issue, as we’re seeing in Mississippi where local communities continue to voice pollution concerns regarding xAI’s data center sites.

It's worth noting, however, that operators do have some options to make natural gas work within their sustainability frameworks. Facilities equipped with carbon capture and sequestration technology can potentially meet ESG goals, says Jenny Gerson, senior director of sustainability at data center operator DataBank, who adds that this does “add significant cost and complexity”.

Choosing natural gas also risks locking the industry into a slower decarbonization path, adds Sahota. “Once companies invest in gas infrastructure, there’s financial and operational inertia to keep it running, delaying renewable adoption. Publicly, many frame gas as a bridge fuel, but historically, bridges tend to turn into highways. The optics also matter because critics paint AI as a climate laggard and accuse the industry of greenwashing, which will only increase if high-profile projects tie AI’s growth to rising fossil fuel use.

“The flip side is gas could displace dirtier diesel and buy time for renewable projects to catch up. The challenge is ensuring the ‘bridge’ has an actual off-ramp, like contractual sunset clauses, hybrid systems that integrate renewables, or modular gas units designed to be decommissioned quickly. Without those guardrails, the sector’s net-zero goals risk becoming aspirational rather than achievable.”

Another area of concern is that natural gas has similar limitations to that of more sustainable sources in terms of scalability. “In the case of hyperscalers it can only be supplemental,” explains Nguyen. “If you had tens or hundreds of thousands of racks, it wouldn’t be viable to have a gas generator per rack. This also means it wouldn’t be a suitable back-up power source as there wouldn’t be enough capacity. It becomes much more viable at enterprise-level however, because the energy levels required would be an order of magnitude or two lower.”

Gas today, renewables tomorrow?

Experts agree that the current focus on natural gas simply reflects the collision between explosive AI growth and grid limitations, and it’s not a sustainable long-term solution to the power needs of data centers.

What we’re really seeing is a temporary mismatch between demand and infrastructure – once the push for AI slows down and new energy infrastructure is built, the demand for natural gas could wind down.

“The most encouraging sign is that companies are already planning for flexibility by shifting loads, curtailing activities during peak demand and investing in cleaner alternatives,” notes Gerson. “Gas is filling a gap, but it’s not where the industry is heading long-term,” she concludes.

Google has unveiled that its first small modular (SMR) reactor from partner Kairos will be located in Tennessee, with operations beginning in 2030.

Last year, Google signed a deal to purchase nuclear energy from multiple small modular reactors (SMRs) developed by Kairos Power. That's in part to help power data centers, which are demanding ever more energy thanks to the rise in AI.

Google said last year that the deal was for seven SMRs, with the first online by 2030 and the rest following by 2035. Speaking at the time, Michael Terrell, Google's senior director, energy and climate, said this will provide 500MW of power to help "unlock the potential of AI for everyone".

Nearly a year on, the two companies have announced that the first SMR will be at Kairos Power's Hermes 2 Plant in Oak Ridge, Tennessee, offering up to 50MW of energy to Google's data centers in that state and neighboring Alabama, as well as bolstering supply to the wider Tennessee Valley Authority (TVA) grid.

Google’s nuclear plans gain traction

Kairos Power said in a statement that TVA was the first US utility to sign on to buy electricity from such a reactor. The company added that it will boost output at Hermes 2 from 28MW to 50MW with operations scheduled to start in 2030.

Google explained in a blog post that the TVA will purchase electricity from Kairos' Hermes 2 plant beginning in 2030.

"In this initial phase of the collaboration, we will procure clean energy attributes from the plant through TVA to help power our data centers in the region with locally sourced clean energy, every hour of every day," said Amanda Peterson Corio, Google's Global Head of Data Center Energy.

"By developing a technology, a supply chain, and a delivery model that can build an industry to unleash American energy, we can attract and support companies like Google and help America win the AI race," said Don Moul, TVA President and CEO, in a separate statement.

Moul added in the Google blog post that having the financial support of the tech company would help avoid customers having to pay for the high cost of development.

"Google is stepping in and helping shoulder the burden of the cost and risk for first-of-a-kind nuclear projects not only helps Google get to those solutions, but it keeps us from having to burden our customers with development of that technology," he said.

That point could be seen as deflecting criticism that such projects have a negative impact on local residents, particularly around AI demand for energy driving up costs for everyone.

The race for data center energy

Google isn't the only tech giant eyeing nuclear technologies to help power AI. The power-hungry technology is placing significant strain on data centers, prompting a rethink of energy sources in recent years.

Last year, cloud computing giant Oracle revealed it had secured building permits for three SMRs, with the aim of using them to power a 1GW data center.

Speaking during an earnings call at the time, CTO Larry Ellison said a key factor behind the nuclear push was the huge power demands as the company moved from AI training to inferencing, a more energy-intensive process.

Elsewhere, other industry giants are eyeing up the potential of nuclear energy to overcome power constrains.

Microsoft, for example, signed a power purchase agreement (PPE) with Constellation Energy in a move that will see 'unit 1' of the Three Mile Island nuclear plant used to power data center operations.

AWS also hopes to locate a data center next to a nuclear power station in Pennsylvania, though that has faced criticism from local residents and electricity providers over potential price rises.

The use of SMRs aims to help get plants up and running as soon as possible, versus the long lead times of traditional nuclear reactors, but this new technology could prove harder to roll out than existing, tested sustainable sources.

MORE FROM ITPRO

• Meta wants to join the big tech nuclear club
• Microsoft is using AI to get its nuclear projects approved in the US
• Hyperscalers go nuclear

A new report has laid bare the scale of water usage by data centers across England - and the results aren’t quite as devastating as one would imagine.

TechUK’s Understanding Data Centre Water Use in England, based on a survey in collaboration with the Environment Agency, suggests data centers are “not intensive water users”.

According to the study, 64% of commercial data centers across the country use less than 10,000 m³ of water each year, which techUK noted is less than a “typical leisure center” or equivalent to the water use of a Premier League football club.

Just 4% of sites reported use over 100,000m³ per year, the techUK report added, in line with industrial manufacturing requirements.

The figures from techUK come amidst rising concerns over the impact of water use by data centers, both in the UK and globally. A host of big tech companies, including Google, have come under fire for data center water consumption rates in recent years.

In 2022, for example, Google was called a “water vampire” for consuming around a third of an Oregon city’s supply, prompting an overhaul of practices at the site.

A similar report in early 2025 highlighted concerns over water scarcity, noting that Amazon, Microsoft, and Google were running data centers in some of the world’s driest regions and exacerbating existing supply challenges.

Data center operators have made significant progress in curtailing water use, largely thanks to advances in cooling techniques. Traditionally, large volumes of water are required to cool data centers, but the rise of ‘closed-loop’, direct-to-chip, and waterless cooling systems are becoming increasingly popular.

Indeed, more than half (51%) of data centers across England are now using waterless cooling systems, which the report noted use “no water beyond the regular functioning of any building”.

Meanwhile, 89% of operators said they no longer need to monitor water usage due to closed-loop systems.

Matthew Evans, Director of Markets and COO at techUK, said the report shows that data center operators are “actively innovating” to address water consumption despite negative public perceptions.

“Data centers are the backbone of the UK’s digital economy and will be central to delivering our AI and innovation ambitions,” he said.

“This report shows that, contrary to some public perceptions, most commercial data centers are actively innovating to use minimal water.”

Data center industry needs support

Data centers are estimated to contribute £4.7 billion in GVA to the UK economy each year, supporting more than 43,000 jobs. The sector also has huge potential, according to techUK, potentially bringing in an additional £44 billion in GVA between 2025 and 2035.

This is dependent on whether data center capacity can be increased across the country, however, and is an issue the trade body insists requires action.

“As demand for compute grows, we must plan ahead,” Evans said. “We need smart policies, resilient infrastructure, and stronger data to ensure digital and environmental resilience go hand in hand.”

The report from techUK noted that “major data and planning” gaps are hampering data center growth across the country, and called for closer collaboration between government, regulators, and industry to accelerate expansion.

Key recommendations from the trade body included the development of a UK-wide “water exploitation index” to help developers and planners assess potential water stress on a region-by-region basis.

Elsewhere, the report also called for greater industry focus on measuring and reporting water usage – as well as water usage effectiveness (WUE) – at sites across the country.

Investment in advanced cooling technologies and the development of cooling systems “based on local water availability” were also highlighted by techUK.

Richard Thompson, the Environment Agency’s Deputy Director for Water Resources, said future data center expansion across England “must go hand-in-hand with protecting public water supplies, food security, and the environment”.

“It is vital the sector puts sustainability at its heart, and minimizes water use in line with evolving standards,” he said.

“We are working with industry and other regulators to raise these to secure the best outcomes for our environment and our water supply for future generations.”

MORE FROM ITPRO

• Meta is working on a 5GW data center to supercharge AI infrastructure
• Google emissions have surged 51% in five years – but it’s making solid progress in data center efficiency
• Why meeting data center demand is becoming a worldwide challenge

Geothermal energy is becoming an increasingly attractive option for powering large-scale data centers with renewable energy. This is primarily because, unlike solar or wind energy which are reliant on weather conditions, geothermal power plants use the Earth's constant internal heat to generate energy.

Proponents push it as an alternative to using nuclear power for green data centers, which hyperscalers have been increasingly pursuing to generate the vast energy needed for AI workloads.

This is a power source well-suited to powering AI compute workloads, as it offers continuous, reliable baseload electricity. This is an essential requirement for data centers running high-density, round-the-clock tasks, notes Graham Grant, CEO at geoscience and engineering software provider at Seequent, The Bentley Subsurface Company.

Geothermal power plants can also be engineered to supply both electricity and direct cooling to data centers, he adds; a crucial advantage for AI workloads that generate substantial amounts of heat.

“By using the Earth as a heat exchanger, the temperatures nearer to the Earth’s surface can be enough to effectively cool data centers without the need for electricity or other fuels to assist the cooling process,” he explains.

Harnessing Earth’s heat

Conventional geothermal systems work by tapping into naturally-found underground hot water reservoirs, channeling steam from beneath the ground into turbines, which spin and generate electricity.

These typically have low lifecycle emissions, take up less land, and have fewer material inputs than other sources, notes Sanjay Purswani, senior knowledge analyst at Boston Consulting Group.

This traditional method is less scalable than wind or solar, however, as it’s only possible in certain geological settings such as near plate boundaries or volcanic hot spots. These areas include Iceland, Indonesia and western US states, such as California and Nevada.

This hasn’t deterred data center operators from embracing geothermal, however. Microsoft, G42 and KenGen are building a data center at the Olkaria geothermal field in Kenya, for example, while back in 2022, Microsoft signed a 10-year agreement with Contact Energy for the supply of electricity from the Te Huka geothermal power station in New Zealand for its local data center.

EGS expands what’s possible

What’s caused a step-change, however, is the development of enhanced geothermal systems (EGS). Meta’s interest in geothermal has grown alongside developments in this area, with the company signing a contract with Sage Geosystems for 150MW continuous geothermal power last year. The tech giant made a similar deal with XGS Energy earlier this year, while Google signed a deal in 2024 with NV Energy to power its Nevada data center using enhanced geothermal systems (EGSs).

EGSs enable the generation of geothermal energy to occur where there is heat in the rocks below the first two to four kilometers of the Earth’s surface but there’s no natural reservoir. In these cases, the approach is to either stimulate the hot rocks to create a reservoir or to use a secondary system to harness heat from the rocks such as a closed loop, which circulates fluid through a large loop that goes deep underground, absorbing heat from the surrounding rocks.

Innovative companies in this sector are opening up huge opportunities for geothermal growth through the development of these next-generation technologies, expanding the accessible geographical area for geothermal by using oil and gas technology like fracking to find underground heat, while advanced geothermal systems are pushing the boundaries of what’s possible.

“Advances in drilling technologies are helping lower costs and increase efficiency, and hybrid systems that combine geothermal with other renewables and improvements in thermal energy storage enable continuous, reliable power and cooling,” Grant says. “Together these innovations could significantly expand geothermal’s role in supporting data centers at scale in the coming years.

“The US president’s declaration of a national energy emergency earlier this year included geothermal as a priority,” Grant adds. “The recent introduction of fast-track permitting for proposed geothermal projects to a maximum of 28 days sets the scene for the expansion of the US geothermal energy market over the next few years.

“But this isn’t just a US trend – recent analysis from Wood Mackenzie showed that global commercial geothermal activity grew 80 percent in the first quarter of 2025, year-on-year,” he continues.

What’s holding geothermal back?

While interest in geothermal rides high, there are still limitations and challenges to overcome in order to scale this industry up.

One big challenge is that you still need hot rock deep underground, and drilling costs spike fast when the geology isn’t favorable, notes Purswani. “Drilling carries risk and if you don’t hit the right spot you’re out millions, so site selection and risk appetite matter a lot.

“Also, permitting and regulatory hurdles are no joke – especially outside of the US and Iceland,” he adds.

Operators also need to consider whether they want to base their data centers near sources of geothermal power, as the practicality of having a geothermal-powered data center is dependent on its location, points out Bob Johnson, VP analyst at Gartner.

“One of the big things with data centers is that they have to be able to communicate, so the questions are where are the main fiber-optic lines that the data center will need to hook up to and is it going to be difficult to run additional lines to the site?

“People often don’t realize that building these massive data centers goes hand-in-hand with major efforts to upgrade fiber-optic lines [to the site]. The goal is to boost capacity by adding more fiber through existing ducts, avoiding the need to lay miles of new pipeline wherever possible,” he says.

While geothermal has been a part of energy systems for over a hundred years, to date it’s played a limited role on a global scale. This may change in the coming years, especially if policies continue to support investment and uptake. But don’t expect it to become the default source for powering data centers notes Johnson – it’s more likely going to play a bigger role in the energy mix, alongside other renewable sources.

Microsoft has made no secret of the surging energy requirements required by its mushrooming AI infrastructure, and its latest attempt to improve sustainability stinks of greenwashing.

The company recently signed a deal with Vaulted Deep, a waste management company that, in its own words, turns “excess organic material into permanent carbon removal”.

Vaulted Deep will bury upwards of 4.9 million metric tonnes of waste deep underground. This largely comes from manure, sewage, and assorted “agricultural byproducts” that can’t be reused or contain harmful chemicals.

Simply put, the tech giant is burying poop to offset emissions – and it really is a perfect way to sum up the generative AI boom.

Energy consumption and emissions at Microsoft are skyrocketing amid the sharpened focus on the technology. In May, the company’s annual sustainability report showed its carbon emissions are soaring.

In a blog post dissecting the report, Microsoft President Brad Smith revealed the firm’s total emissions – including Scope 1, 2, and 3 – increased by 23.4% compared to its 2020 baseline due to AI and cloud expansion.

That baseline refers to greenhouse gas emissions which Microsoft uses as a marker in its 2020 commitment to become carbon negative by 2030, which at this stage has taken a few knocks.

Energy use also shot up by 168% last year, according to the report, but Smith was keen to point out that shows its efforts to improve sustainability are working given the disparity between emissions and overall usage.

Microsoft is burying waste - and its head in the sand

Generative AI is power hungry, to say the least, and the insatiable demand for services is placing huge strain on data center infrastructure globally.

The soaring power consumption rates have reached such a point that grid operators in Europe are struggling and scrambling to upgrade grid infrastructure. Data center water consumption, the often-unseen aspect of the problem, is causing serious problems for local communities.

Meta’s data center in Georgia, for example, is allegedly sucking the earth dry and limiting supply to local residents.

To its credit, at least Microsoft is doing something – it could always just shrug off concerns and criticism. In April, the firm signed a deal with carbon capture company AtmosClear to remove nearly 7 million tonnes of carbon emissions.

A similar deal with Chestnut Carbon, meanwhile, will see seven million metric tonnes of carbon offset through reforestation in the United States.

Elsewhere, infrastructure improvements have also been a key focus for the tech giant. These include novel data center cooling techniques and the use of cross-laminated timber (CLT) in data center construction to reduce its reliance on steel and concrete – both of which are large contributors to carbon emissions.

These are all positive moves, no doubt. But the volume and frequency of ‘partnerships’ and new-fangled attempts to offset emissions just reek of greenwashing and token gestures to prove they’re actually mindful of the effect they’re having on the environment.

Weigh the carbon removal numbers against Microsoft’s yearly emissions – around 15 million tonnes in FY24, per its 2025 sustainability report – and they don’t look so significant.

Moreover, perhaps I missed it, but Microsoft and counterparts also weren’t exactly bending over backwards before the generative AI boom to plant millions of trees or bury excrement thousands of feet underground.

It’s clear they’ve bit off more than they can chew with the beast they’ve unleashed on the industry. So now they’re scrambling.

Everyone’s looking for a quick fix

In an announcement confirming the deal with Vaulted Deep, Brian Marrs, senior director of energy and carbon removal at Microsoft, said this will provide a “differentiated, scalable approach to permanent carbon removal”.

"Its work delivers immediate climate benefits while stimulating local economies and addresses long-standing environmental challenges that communities face every day,” he said.

Cutting through the mind-numbing corporate jargon we’ve come to expect from big tech companies, the “immediate benefits” comment stands out here and speaks volumes about how Microsoft - and its counterparts in the industry - view the topic of sustainability.

They’re looking for a quick fix, because they’ve dived headlong into generative AI without considering the long-term impact. Sustainability and net zero were cool and trendy in 2020 when the company set its ‘baseline’ and pledged to reach carbon negativity by the end of the decade.

But AI has become such a moneyspinner for the industry that they can’t slow down at this stage or else be eclipsed by a competitor – and that would impact shareholder value, which of course is paramount.

Microsoft and friends have bombarded us with an array of ambitious plans to meet surging energy demands. From reopening nuclear power stations to geothermal energy projects and development of small modular reactors (SMRs).

At this stage it seems like they’re just throwing money at any conceivable idea to pad the numbers and catch headlines.

Google has pledged to pause AI workloads when electricity grids are hit by demand spikes.

In a blog post this week, the tech giant revealed it had signed agreements with two electrical utilities in the US to flexibly manage how its AI data centers consume power during times of peak demand.

The agreements are with Indiana Michigan Power and Tennessee Valley Authority, and will bring flexible demand into Google data centers to allow them to shift or reduce power consumption "during certain hours or times of the year."

Google's pledge comes amid the rise of AI and fears of what it might mean for energy supply, with data centers designed for the technology chewing through large amounts of electricity, as well as water for cooling.

"These capabilities, often referred to as demand response, have several advantages, especially as we continue to see electricity growth in the US and elsewhere," said Michael Terrell, head of advanced energy at Google.

"It allows large electricity loads like data centers to be interconnected more quickly, helps reduce the need to build new transmission and power plants, and helps grid operators more effectively and efficiently manage power grids,” Terrell added.

"These agreements represent the first time we’re delivering data center demand response by targeting machine learning (ML) workload.”

Google has admitted its emissions have surged 51% over the last five years, partially driven by the AI boom, and the company has invested in renewables and nuclear — as well as funding training to boost the number of electricians in the US.

Google has been testing similar setups

While this is the first time "demand response" has been used with AI data centers, Google has trialled the idea with Omaha Public Power District, reducing demand for power from data centers when storm Gerri hit the Midwest in January 2024, knocking power plants offline.

The company also uses a similar idea with YouTube, shifting workloads to regions with less intense power demands and to maximize use of renewables.

"The first data center demand response capabilities we developed involve shifting non-urgent compute tasks — like processing a YouTube video — during specific periods when the grid is strained," Terrell said.

"Through our ongoing partnerships with Centrica Energy and transmission system operator Elia in Belgium, and Taiwan Power Company in Taiwan, we've leveraged this capability to help grid operators maintain reliability during those periods of the year when demand is the highest."

Applying the idea to AI workloads makes sense given their high energy consumption and the limited power sources available.

"By including load flexibility in our overall energy plan, we can manage AI-driven growth even where power generation and transmission are constrained," Terrell said. "We believe this is a promising tool for managing large new energy loads and facilitating investment and growth."

AI risk to energy supply

The move was welcomed by Indiana Michigan Power. Steve Baker, president and chief operating officer at the energy supplier, said closer ties between industry and utility companies will be crucial to maximize efficiency.

"As we add new large loads to our system, it is critical that we partner with our customers to effectively manage the generation and transmission resources necessary to serve them," he said.

"Google’s ability to leverage load flexibility as part of the strategy to serve their load will be a highly valuable tool to meet their future energy needs."

Though the agreement is limited to two utilities, it's likely such deals will be struck in the future more frequently as tech giants building data centers seek access to grids.

Google admitted there are limitations, however, and not all services can be paused.

"Data center demand flexibility is still in the early stages and will only be available at certain locations," Terrell said.

"There are limits to how flexible a given data center can be, since high levels of reliability are critical for services like Search and Maps, as well as Cloud customers in essential industries like healthcare."

That means "demand response" isn't a solution to the looming challenges faced by electrical grids thanks to the AI boom, with both the US and Europe needing significant investment to keep pace with the technology's power demands.

The International Energy Agency has said AI will chew through the same amount of energy as Japan by 2030.

MORE FROM ITPRO

• Google Cloud is leaning on all its strengths to support enterprise AI
• A threat to Google’s dominance? The AI browser wars have begun – here are the top contenders vying for the crown
• Google is getting serious on cloud sovereignty

It’s well known that data centers are power-hungry beasts and one of the biggest challenges is fueling their appetite. Even before the advent of generative AI, the Semiconductor Research Corporation reported that the electricity demands of computation and data communication were growing significantly faster than global energy production. With AI now accelerating that consumption, the challenge has only intensified.

Analyst firm Gartner is currently predicting that in the near future, roughly half of new data centers will need some form of dedicated power generation – a microgrid – to either supplement or replace what’s available from the regional grid. With many looking to add weight to their green credentials, these are often from sustainable sources such as water, wind and solar.

Solar has attracted significant attention recently, with the announcement of the UK’s first solar-powered data center and hyperscalers such as Google inking a plethora of solar energy deals, including a recent contract with US developer energyRe to invest in and purchase renewable energy credits from a portfolio of over 600MW of solar projects.

This is particularly attractive for hyperscalers as they seek a quick fix for rising emissions linked to AI. Google's emissions have risen 51% in the past five years, while Microsoft's Environmental Sustainability Report 2025 revealed a 23.4% rise in emissions compared to its 2020 baseline.

With prices also continuing to fall, solar-powered data centers are likely to become more of a common occurrence. But might they ever become the default energy source for data centers, rather than just the most common kind of green data center?

Can solar power be a sole source?

Solar power is experiencing huge growth in regions such as the EU, where it’s just become the biggest power source for the first time, as well as in the US where wind and solar has now overtaken coal, per energy think tank Ember.

But in the near term, solar is unlikely to become the main power source for data centers, as solar power production isn’t always sufficient. Experts are seeing many deployments of solar as supplemental.

The energy provision for a data center is rarely a single source, notes Bogi Hojgaard, Energy System Expert at The Carbon Trust consultancy, pointing to the fact that while it could provide ample energy to power a data center on average, solar does come with its own challenges.

The main issues are of course generation and storage. Solar could only work as a single power source in an area with a large percentage of sunny days, and even then, you’d need a sizable area to build the solar farm.

“For solar, it takes between five and seven acres to generate a MW of energy, However, that refers to the peak generating capacity of a solar installation,” says Bob Johnson, VP Analyst at Gartner.

“In reality, since the sun isn’t always shining at its peak (or at all, like at night), the average power from a solar installation is roughly 25% of the maximum – per US Department of Energy statistics.

“In reality, you would need 20-30 acres per MW of capacity, or for a 100MW data center, 200-300 acres available for solar. Of course, this would depend on where the site is located, because some areas are better for solar than others,” he adds.

Since the sun doesn’t shine 24/7, storage is also a necessity and the capacity needed for a fully 24/7 solar solution would likely be very significant. While much has been mentioned about utility-scale batteries, adding sufficient battery storage for a 100MW data center to keep operating would add around US$1bn to the total cost, say Johnson, noting this is why he doesn’t think solar-alone is the best choice.

Choosing the right energy mix for the job

But then how does solar stack up against alternative power sources such as nuclear, wind, hydropower, and geothermal?

Aside from nuclear, costs are unlikely to materially change over the next five years notes Hojgaard, citing that comparisons of generation technologies on levelized cost of energy (LCOE) would yield a ranking of solar, wind, geothermal or hydro (depending on local conditions) and nuclear – with the latter three being substantially higher in cost.

Other factors when deciding on the most suitable power source includes location and gaining planning permission. Onshore wind can face issues such as public views on wind farms and the impact on the topography on power generation, while offshore wind is out of sight and higher capacity while also incurring a higher cost.

“There are some sites exploring offshore wind too though,” notes Bob Burgoyne associate director, Business Services at The Carbon Trust. “Blackpool is looking to develop a high-performance data center campus and reduce the curtailment of offshore wind by using this power for AI training loads.”

‘Traditional’ geothermal – such as in Iceland – requires ready access to superheated water/thin Earth crust, so necessarily located close to fault lines, which make them more prone to earthquakes. “Therefore, building data centers close to these would be a risk/cost trade off, although electricity can be transported, so that risk can be managed,” says Hojgaard.

‘Traditional’ geothermal – widely used in Iceland – requires ready access to superheated water/thin Earth crust, so necessarily located close to fault lines, which make them more prone to earthquakes. “Therefore, building data centers close to these would be a risk/cost trade off, although electricity can be transported, so that risk can be managed,” says Hojgaard.

Relatively new is advanced geothermal, which doesn’t require the same access to hot water, instead drilling deep boreholes to access the heat present across the world. “This has intriguing potential, with the first pilots underway in Nevada and Utah, but not until the next decade,” says Johnson.

Hydro is a great option, but basically all the available hydropower’s already taken Johnson adds, bringing us, finally, to nuclear. In terms of ‘traditional’ nuclear, many countries don’t have the skills or capability to build and for those that do the size of plants typically range from single reactors of 1,000MW up to multiples of up to 1,800MW – Hinkley Point C and Sizewell C for example, will total 3,320MW.

“A hyperscale data center demand typically ranges between 20 and 100MW, so building a dedicated plant for one wouldn’t make sense, though it could be part of an energy portfolio for a group of data centers,” Hojgaard notes.

A new class of nuclear reactor – the small modular reactor (SMR) – is under development and has “the longer-term potential for being ideal. However, due to the complex regulatory environment, the first ones won’t become operational until around 2030, so in reality they’re a next-decade solution,” Johnson says.

Each power source clearly has its own set of challenges and while many don’t believe it’s going to become the de-facto power source for all future data centers, solar is clearly one of the best options for those that don’t want to, or simply can’t, rely fully on a local power grid.

Data centers look set to continue being powered from a mix of sources, but experts agree that solar will definitely continue to take market share, as both photovoltaic and battery costs are declining.

Breakthroughs in cooling management could also improve the match of solar output to data center demand, while battery chemistry advances may drive down storage costs. While it’s too early to say how far advancements will go, solar is poised to claim an increasingly larger share of the data center energy mix — even if it never becomes the sole power source.

Meta CEO Mark Zuckerberg has revealed plans to build a 5GW data center, with one cluster alone set to be nearly as big as Manhattan.

Detailing the plans in a social media post this week, Zuckerberg said Meta plans to invest “hundreds of billions” of dollars as part of the move to build out AI infrastructure.

Meta, along with industry competitors, has been ramping up investment in AI infrastructure over the last 18 months, with capital expenditure at the tech giant surging to $68 billion.

The pledge comes as Meta has offered large sums — some in the hundreds of millions — to entice top AI talent away from rivals, successfully poaching staff from Google, OpenAI, Anthropic, and Safe Superintelligence, among others.

In the last two weeks, reports suggested Meta sought financing of as much as $29bn to fund data centers, and bought 1GW of renewable power as part of the push.

Meta is big on data centers

Writing on Threads, Zuckerberg outlined some of the company's plans when it comes to competing with AI, and in particular developing superintelligence.

OpenAI has long targeted the idea of artificial general intelligence (AGI), and Meta recently launched its own Superintelligence Lab.

"For our superintelligence effort, I'm focused on building the most elite and talent-dense team in the industry," he wrote. "We're also going to invest hundreds of billions of dollars into compute to build superintelligence."

Zuckerberg noted that analyst firm SemiAnalysis reported Meta would be the first lab to bring a 1GW "supercluster" online, but said that Meta is already working to build several multiple-gigawatt data centers.

The first of these will be named Prometheus, he revealed, and is set to come online in 2026.

"We're also building Hyperion, which will be able to scale up to 5GW over several years,” he added. “We're building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan."

"Meta Superintelligence Labs will have industry-leading levels of compute and by far the greatest compute per researcher. I'm looking forward to working with the top researchers to advance the frontier!"

Where will the data centers be built?

In New Albany, Ohio, the Prometheus system is a 1GW cluster of data centers using both Meta's own infrastructure and leased systems, according to SemiAnalysis, noting that this means "Meta ramps faster".

It will have two natural gas plants on site, though Prometheus will also use renewable energy.

Hyperion, meanwhile, is in Richland Parish, Louisiana; work started last year, and the first 1.5GW phase is expected to come online next year. It will be powered in part by renewables but also feature natural gas.

The same day that Zuckerberg made his enthusiastic Threads posts, a story in The New York Times described Meta's construction of a $750 million data center in Georgia, highlighting the impact on the local community including local wells going dry.

While a Meta spokesperson disputed the account, the local water authority said the data center uses 10% of the county's total water supply.

Bigger data centers means bigger challenges

The rise in AI has led to a huge increase in data center investment and funding for associated infrastructure.

Earlier this year, the US government unveiled plans for the Stargate Project, a $500 billion scheme aimed at building out AI infrastructure across the United States.

With backing from OpenAI and cloud computing giant Oracle, the project aims to meet projected demand for compute power as enterprises across the country continue ramping up AI adoption.

But with infrastructure investment comes new demands - and concerns. Particularly around the environmental impact and strain placed on energy grids.

A host of big tech companies, including AWS and Microsoft, are investing heavily in nuclear energy to meet expected power demands as a result of their own infrastructure build out.

Indeed, the International Energy Agency said AI will chew through the same amount of energy as Japan by 2030. Google reported earlier this month that its emissions have surged 51% over five years, largely down to AI infrastructure development.

Concerns over energy constraints are also growing on the other side of the Atlantic, with research earlier this year warning that European energy grids might struggle to contend with surging power demands.

MORE FROM ITPRO

• Meta wants to join the big tech nuclear club
• Oracle wants to power a new gigawatt data center with three small nuclear reactors
• Data center water consumption is out of control

You can measure it in gallons, you can measure it in swimming pools, you can measure it in oceans but, however you decide to calculate the amount of water being used to cool today’s data centers, the number is staggering. According to some reports, certain areas of the US are using seven billion liters of water per year to cool these lifelines for tech innovation.

One of the key culprits of this water usage is evaporative cooling, in which warm air is drawn through pads that have been soaked in water, evaporating it and lowering the air temperature behind the fan.

While this method is generally more energy efficient than which uses drastically more water than traditional air cooling solutions. The demands on these systems are only deepening as data center real estate gets bought up, rack densities increase, and environmental concerns come further to the forefront. In some newer areas of the market, such as the Middle East, air cooling is a non-starter.

Sanjeev Khanna, the COO at Caddis Cloud Solutions and a former Intel, Microsoft, and Oracle employee sees three main alternatives to traditional evaporative cooling: closed loop systems, direct to chip cooling, and immersion cooling. He says that, especially when it comes to direct-to-chip cooling, there is a perception that this is entirely new technology. It’s something that he disputes.

“I joke about it like I've been schlepping direct-to-chip for 20 years. The good thing is, there are a lot more people. Innovation is going into it because I think there is demand now. It never took off [before.]”

He says that focus on new technology is being pushed by the increasing power demands of data centers worldwide and that the industry is taking notice.

“I was at Nvidia GTC and you couldn't walk ten feet without looking at another cooling solution…There's a physics problem now, and to solve the physics problem, people are going to need liquid. It's not a choice.”

The changing ways we're cooling data centers

Jenny Gerson, DataBank’s senior director of sustainability, says that – like for many companies – the way they’ve approached cooling has shifted as their customers’ needs and wants have evolved. Earlier on, that meant evaporative cooling solutions, but now they are edging towards direct to chip cooling with a default design that uses a closed loop. Microsoft recently announced that it is prioritizing this style of design as well.

“As data centers are continuing to evolve, as there's more high density and higher cooling needs, we're bringing the cooling closer to the chips,” Gerson explains. “So, either at the rack level, at the server level, or at the chips themselves. And so we'll do whatever customers need for their density.”

While adoption and education can prove to be bottlenecks in the short term, Gerson says that the main sticking point in any newer approach to cooling in the sector is power availability and the capability to distribute that power effectively.

For every person who spoke to ITPro for this story, power efficiency was a core consideration. Joe Capes, CEO at Liquid Stack, says that as we reach the outer limits of single phase cooling, liquid cooling is going to have to be the future.

“With these rising chip temperatures and rack densities, the only way to really deal with this is through liquid cooling,” he says.

“Whether it's two phase immersion, two phase direct to chip or hybridization of multiple liquid cooling technology, technologies like combining single phase and two phase systems together. And I think that what we're seeing from companies like Nvidia and others is looking forward and farther down the path of these roadmaps to see how two phase technologies can be adapted.

Pace of cooling innovation

Another question is the pace of innovation. Aaron Lewis, the chief commercial officer at Johnson Controls, says that the staggering speed at which data centers are expanding means an expedited timeline for those bringing products to market.

“When you think about new product introductions in the HVAC space that used to be a three to five year program that you would put in place,” he tells ITPro. “Today, if you were to wait that long, you'll miss three cycles of different iterations of data centers based on new chip requirements that are coming out, or new cooling technologies within the cold chain itself.”

Lewis says that another key shaper of innovation is where data centers are tending to be placed. In a warmer climate, like Arizona or the Middle East, mechanical air cooling solutions are limited in their effectiveness because of the high ambient temperatures. In a cooler climate, he says other forms of cooling can be relied upon.

“We developed data centers in Montreal, where it's very cold most of the year, and you can utilize some type of free cooling system, so you're able to run water through either the dry cooler or a free cooling coil on the chiller itself, so that you can maintain the space without having a mechanical cooling.”

For Capes, the pace of innovation means that the infrastructure and technology just can’t quite meet in the middle yet.

“The adoption and scale up of liquid cooling has really kind of transpired over the last 12 months. And so, as an industry, we're not quite there yet. It's going to take some time before, before data center facility designs and working fluid temperatures catch up with the way that these liquid cooling systems are designed to operate.”

Immersion cooling

Closer to the frontier of new innovation in cooling is immersion. This is where the rack is set into a pool of oils, often petroleum or vegetable. Khanna isn’t sure that the method is ready for primetime. For him, the logistical challenges—including difficulties regarding maintenance—mean that widespread adoption is a ways off yet.

Immersion cooling could be a gamechanger for data centers in extremely hot environments, as racks submerged directly in cooling liquids can disperse heat far more efficiently. Submerged racks are also protected from other environmental factors such as humidity.

But Khanna isn’t sure that the method is ready for primetime. For him, the logistical challenges – including difficulties regarding maintenance – mean that widespread adoption is a ways off yet.

“I think people will figure out a way to get creative there and how we can solve the operational challenges,” he says, adding “personally, I don't think it's ready yet. People who build these, they think that they're ready to go, but I don't think this is ready yet.”

That perspective won’t stop people like Seamus Egan, the general manager of immersion cooling at Modine Manufacturing. He says that immersion cooling allows for an interplay between already existing evaporative cooling techniques and this newer approach, with evaporative cooling being used to reduce the ambient temperature of the data center.

“When we take that entire server and immerse it in a liquid, and these are dielectric fluids, then we've got the ability to capture 100% of the heat…We can now take the heat that's captured inside of that liquid, do a liquid to liquid transfer, and then evacuate that heat on a full liquid path, end to end, out of the actual data center.”

While Eagan is bullish on the potential for this style of cooling, he isn’t naive when it comes to the barriers that will need to be addressed before widespread adoption can be on the horizon.

“Number one is that all customers want to use the least disruptive techniques that align with what they have today. So if 95% of today's centers in the market today are air cooled data centers, then people will tend to gravitate towards those or try to have a solution that's least disruptive, right? The second thing is just experiences and working with them.”

The pace of innovation is creating an environment where the technology to cool data centers is struggling to keep up. Water availability, power availability and distribution, and reducing cost and ecological impact, are all things the industry is grappling with. While the ideal solution isn’t apparent, what is clear is that the modus operandi of the last fifty years or so, air cooling, is no longer fit for purpose. Liquid, it appears, is the future.

Google emissions continue to climb amid the AI boom, new figures show, up 51% from 2019 to 2024 and 11% in the last year alone.

The tech giant’s total carbon dioxide emissions reached 11.5 million tonnes last year, according to its annual environmental report, marking a steep increase on the 7.6 million tonnes recorded in 2019.

A key factor behind this increase lies in the huge strain placed on data center infrastructure by generative AI - and it’s an issue that Google isn’t alone in tackling.

The International Energy Agency warned that AI will chew through the same amount of energy as Japan by 2030, with emissions from leading AI companies up by 150% since 2020.

There are different types of emissions, and Google has made headway in curtailing impact across a number of areas. Scope 1, for example, refers to sources that a company directly owns, such as buildings.

Google cut this type by 8% year-on-year, the report noted. Meanwhile, Scope 2 refers to indirect emissions from energy used by the company; those fell by 11%.

Scope 3 refers to supply-chain emissions, which climbed 22% year-on-year — and they make up 73% of Google's carbon footprint.

Google said that the increase in Scope 3 emissions were "primarily due to increases in data center capacity delivery (i.e., emissions generated from the manufacturing and assembly of technical infrastructure hardware — including for AI — and their logistics, as well as from data center construction)."

The company said it would continue to build out infrastructure next year and aimed to "decouple this growth from a proportional increase in emissions", in part by encouraging clean electricity in the supply chain and shifting to low-carbon data center construction.

Google emissions are being addressed

Still, the company highlighted its progress. Notably, it slashed data center energy emissions by 12%, with Google data centers using 84% less overhead energy than the industry average.

In a blog post examining the report, Kate Brandt, chief sustainability officer at the tech giant, said that despite a 27% increase in electricity demand for data centers, the company has “successfully decoupled our operational energy growth from its associated carbon emissions.

"This was largely due to more than 25 clean energy projects we’d contracted over the past several years — some as far back as 2019 — coming online in 2024.

Google also claimed that its products more than made up for its own emissions, noting that the combined emissions savings for 2024 alone from Nest thermostats, Google Earth Pro, Solar API, Google Maps' fuel efficient routing, and Green Light was 26 million metric tons of CO2 equivalent — more than double Google's own tally.

Cleaner energy required

Brandt admitted that "it's important to be up front about the hurdles we're facing", highlighting that total admissions were up 11% year on year.

She pinned the blame on slower progress with decarbonizing energy in some key regions, notably Asia Pacific, and said challenges remain with newer forms of energy production.

"When it comes to the next-generation clean technologies we need — the solutions that go beyond wind and solar like enhanced geothermal, advanced nuclear, nuclear fusion, and more — they’re still in their early stages," she added. "They’re not yet deployed at the scale required and can remain expensive to get online."

Google has teamed up with various energy companies on smaller modular reactors and other advanced nuclear, as well as geothermal and biomass, the report noted.

In last year's version of this research, Google saw a 13% increase year-on-year for GHG emissions and a 48% increase since 2019, with the report noting: "This result was primarily due to increases in data center energy consumption and supply chain emissions."

MORE FROM ITPRO

• Just 10 tech companies are responsible for half of all data center energy consumption
• Data center water consumption is out of control
• Pressure rises to bolster data center energy efficiency as IT leaders worry demand is pushing the grid to breaking point

Cloud providers in Europe are calling for data centers to reuse waste water and help fund public infrastructure improvements, but they’ve also urged the EU not to single them out for water use restrictions.

That comes amid concerns that data centers are too thirsty. Most use large amounts of water for cooling — in 2022, for example, Google was called a "water vampire" for soaking up a third of one Oregon city's water supply.

The rise in demand for AI data centers has led to growing concerns about potential water scarcity, with a report from SourceMaterial earlier this year noting that Amazon, Microsoft and Google were all running data centers in some of the world's driest regions.

Last month, the European Commission (EC) said water shortages are being exacerbated by leaks and pollution across the bloc, but also specifically pointed to high usage from data center operators.

Cloud Infrastructure Providers in Europe (CISPE) admitted that "water scarcity is rapidly becoming a pressing challenge," saying the organizations' members support the Commission's inclusion of water resilience in its Green Deal.

But it also called for the Commission to not target any specific industry in a post titled "don't throw the baby out with the bath water!"

Data center water use is out of control

In that post, CISPE said its cloud members play a role as "responsible water users and as innovative enablers of better water management".

The trade group insists the sector is "already among one of the most responsible water users." As evidence of this, CISPE pointed to the Climate Neutral Data Centre Pact, which includes measuring Water Usage Effectiveness to highlight data center consumption rates.

"Water resilience is a pressing issue that should be addressed by the Commission," said Velimira Bakalova, sustainability policy manager at CISPE.

"But the Commission should also recognize not only the significant work that the sector has done, but its leadership in this area and its potential to deliver and support innovative solutions that will significantly improve water management for the benefit of all."

The EC proposed minimum performance standards alongside investment in technology to find new cooling methods. The trade group also proposes to run pilots to promote water efficiency across all areas of industry.

"In particular, to promote water savings across data centers, the Commission will rate their energy efficiency and overall sustainability and propose minimum performance standards, including for water consumption," the strategy document noted.

There are technologies that limit that consumption, such as liquid cooling, but they remain niche. Via reclamation projects, AWS has said it will return more water to communities around the world by 2030 than it uses in its operations.

Similarly, Microsoft last year unveiled a new data center design that promises to use no water for cooling at all.

Recommendations for more efficient water usage

CISPE's policy paper included recommendations for how the EC could improve water use without such restrictions. That includes creating frameworks to support water reuse, such as using municipal wastewater for data-center cooling.

The organization also recommended setting up public-private partnerships to modernize water infrastructure by using private innovation and money that's overseen by the public sector.

Alongside those practical ideas, CIPSE also called for the EC to promote the use of the cloud to help manage water systems, including digital twins and digital sensors.

Notably, the trade group urged restraint with regard to regulatory action.

"Adopt a cross-sectoral, holistic approach to water policy that avoids singling out specific industries,” CISPE said. "The sector is already navigating high energy costs, uneven electricity market access, and complex regulatory demands."

"Imposing new, standalone water regulations could increase costs, create regulatory fragmentation, and deter investment. This risks shifting infrastructure outside the EU-undermining both sustainability and sovereignty goals," CISPE added.

"Such regulatory uncertainty could also reduce Europe’s attractiveness for climate-neutral infrastructure investment at a time when other regions offer clear and stable frameworks for green data growth."

However, such restrictions may not only come via EU regulations — back in 2023, utility Thames Water said it may restrict data centre water use in London during droughts.

MORE FROM ITPRO

• Just 10 tech companies are responsible for half of all data center energy consumption
• Pressure rises to bolster data center energy efficiency as IT leaders worry demand is pushing the grid to breaking point
• Europe needs more energy and better grids to meet data center power demands in the age of AI

Ravenscraig was Scotland’s largest steelworks before closing in 1992, but now it could be set for a new lease of life as a green AI data center.

Proposals put forward as part of the UK government’s AI Growth Zone scheme will see the site create more than 2,000 jobs alongside an estimated £3.9 billion in capital investment.

The drive to transform the site is being led by Edinburgh and York-based renewable energy developer, Apatura. Under the plans, the Ravenscraig development will include battery energy storage spanning 160 acres and powered by 550 megawatts of grid connections, which are due to come online by 2030.

Ravenscraig is one of five ‘AI-ready’ sites Apatura is currently working on across Scotland’s Central Belt. The company said it aims to harness the country’s “growing supply of renewables”, drawing on wind and solar generation.

“This proposal plays strongly to Scotland’s strengths - in green energy, in AI, in education and in skills - and would complement the country’s significant capabilities in big data and research,” said Apatura CEO Giles Hanglin.

“We’ve secured the land and grid, and we’re already working closely with relevant organizations locally and nationally to make sure the benefits are widely felt.”

Apatura put forward Ravenscraig as a candidate for the government’s AI Growth Zone initiative last month, and it is now under consideration by the Department for Science, Innovation & Technology (DSIT).

Hanglin said its selection as an AI Growth Zone would be a “significant endorsement” of Scotland’s digital sector and help position the country as a “powerhouse of green, AI-enabled digital infrastructure”.

The rise of green data centers

Green data centers have become a key talking point over the last two years, particularly amidst growing concerns about infrastructure energy consumption in the wake of the generative AI boom.

As ITPro previously reported, being a truly green data center is about more than just renewable energy sources. Indeed, these sites aim to improve sustainability across the breadth of operations.

This includes a strong focus on e-waste management, for example, spanning the decommissioning of servers, networking equipment, and assorted hardware.

Energy efficiency and water management are equally important considerations for green data centers, with operators making great strides toward lower power usage and reducing the volume of water consumed in cooling processes.

The UK is now home to a range of green data centers. In October last year, for example, colocation and managed services firm DSM Group opened a sustainability-focused data center in Peterborough.

The site draws upon renewable energy, including a 200kW solar farm, alongside cutting-edge cooling technology.

There is growing pressure on government and industry to increase investment in renewable energy sources and green data centers.

A recent study from Oxford-based Aurora Energy Research called for greater strategic collaboration between government and industry to compensate for rising data center power demands across the country.

Analysis by the firm found that closer ties could unlock billions of pounds in low-carbon power investment, but also support around 5 GW of green data center growth over the next decade.

MORE FROM ITPRO

• UK data center industry growth faces serious challenges
• Data center waste heat to warm London homes as development partner found for UK-first project
• Europe’s largest data center approved for development in UK

Emissions from four leading AI companies increased by 150% since 2020, with electricity consumption by data centers up by 12% annually — four times faster than global growth.

That's according to a report from the International Telecommunication Union (ITU), part of the United Nations (UN). Those figures only go up to 2023, the most recent year of full data, the ITU noted, so the more recent acceleration of AI is likely to be using much more energy and causing even more emissions.

AI is predicted to use as much energy as Japan by 2030, according to a report earlier this year by the International Energy Agency, but only 13% of firms are tracking their AI energy use, according to another study.

The rise in energy demand is sparking concerns about whether aging grids can keep up, while tech companies themselves have begun investing in their own power sources, including nuclear.

According to the ITU report, electricity consumption was up 12% per year from 2017 to 2023, and the 164 digital companies that reported their electricity consumption accounted for 2.1% of the global total of electricity use — with just ten companies making up half of that.

The US had the largest share of global data center electricity consumption at 45%, followed by China at 25% and Europe at 15%.

Although tech companies promote their renewable energy efforts, greenhouse gas emissions made up 0.8% of all global energy-related emissions in 2023, a 1.4% increase from the year before.

The companies assessed in the report have a carbon footprint above the combined emissions of Argentina, Bolivia and Chile, and the data collected in the report shows the industry is falling short of wider calls to cut emissions by 45% by 2030.

The top ten emitters from the sector make up 53% of all direct emissions, known as Scope 1 emissions.

"Advances in digital innovation — especially AI — are driving up energy consumption and global emissions," said ITU Secretary-General Doreen Bogdan-Martin.

Some improvements to data center energy consumption

There were signs of improvement, despite the AI driven increases — indeed, the ITU noted a rise in efforts to make AI infrastructure more efficient, pointing to Google's sixth-gen TPU that uses 67% less energy.

Of the 200 digital companies included in the report, 23 were using 100% renewable energy, up from 16 the year before. And 49 released climate reports, while 110 publish Scope 3 targets on indirect emissions from their supply chains.

"Digital companies have the tools and influence to lead the global climate transition, but progress must be measured not only by ambition, but by credible action," said Lourdes O. Montenegro, Director of Research and Digitisation at the World Benchmarking Alliance.

"This report provides a clear signal to the international community: more companies are stepping up, but emissions and electricity use continues to rise."

The report called for the tech industry to strengthen its data verification on climate reporting, disclose the full environmental footprint of their AI operations, and continue increasing adoption of renewable energy.

"Despite the progress made, greenhouse gas emissions continue to rise, confirming that​​ the need for digital companies to adopt science-aligned, transparent, and accountable climate strategies has never been greater,” said Cosmas Luckyson Zavazava, Director of ITU's Telecommunication Development Bureau.

“ITU's work in monitoring the environmental impact of the sector is a crucial step towards achieving a sustainable digital transformation.”

MORE FROM ITPRO

• Tech industry takes vast lead in green energy spending, biggest companies vie for top spot
• What is a green data center and why are they attracting big investment?
• Pressure rises to bolster data center energy efficiency as IT leaders worry demand is pushing the grid to breaking point

Microsoft has announced plans to invest $400 million in Swiss data centers to expand its cloud and AI infrastructure in the region.

The move comes alongside data sovereignty moves — keeping European data in Europe — which has driven billions in Microsoft investment in data centers in the region.

Earlier this year, Microsoft said its EU Data Boundary for the Microsoft Cloud was finally complete and would include Switzerland.

Microsoft launched its first data centers in Switzerland six years ago, and in 2022 opened a local Microsoft Innovation Hub. Microsoft didn't say how many jobs the investment might create, but reports from Reuters suggest the company has about 1,000 staff in Switzerland.

The $400 million in investment will focus on several projects, but primarily aims to expand and upgrade its four existing data centers near Zurich and Geneva to meet demand for cloud and AI services, the company said.

The upgrades will serve 50,000 existing customers. So far, all of Microsoft's electricity consumption in Switzerland has been covered by renewable energy purchases, and it aims to continue that sustainability work.

Microsoft ramps up data sovereignty efforts

Expanding access to those data centers inside Switzerland is key for regulated sectors like healthcare, finance, and government, according to Microsoft president Brad Smith.

In a statement confirming the move, Smith said the investment will help “further strengthen Switzerland’s long-term economic resilience and competitiveness” while ensuring the company adheres to regulatory requirements.

Switzerland ranks second globally in GitHub AI contributor share, with the number of contributors doubling since 2022. Microsoft added that the number of its Swiss customers engaging with its AI tools had increased by three percentage points to 31% over the last six months.

"We are steadfast in our mission to empower our customers and partners, as AI's true potential is unlocked when innovation meets real-world implementation," said Catrin Hinkel, CEO of Microsoft Switzerland, in a statement.

"Together, we'll continue to build on Switzerland's tradition of innovation to ensure technology delivers meaningful value to the economy and society."

Beyond data centers

Alongside the data center investment, Microsoft will partner with Switzerland Innovation Parks, which supports SMEs and startups, while also funding AI training for one million Swiss people by 2027.

The latter includes training designed specifically for apprentices, non-profits, and SMEs, as well as "broad AI literacy" through websites focused at general users.

"To ensure that Switzerland has the right skills to stay competitive, Microsoft aims to help skill one million people in Switzerland by 2027 with the goal of developing AI capabilities for education, industry, nonprofit organizations, and citizens," Microsoft said.

The tech giant also said it will invest further in working with the international organisations headquartered in Switzerland, notably creating digital services for refugees to support UNHCR.

"Microsoft will expand its collaboration across the International Geneva ecosystem, working with the United Nations and international organizations to advance AI policy conversations, support global AI skilling programs, and address challenges in health, humanitarian action, and human rights," the company said.

MORE FROM ITPRO

• Analysts think Microsoft's data center rollback is bad news for the AI boom – but the company says not to worry
• Microsoft signs off on $100 billion investment plan for data center development
• First Microsoft, now AWS: Why tech giants are hitting the brakes on costly data center plans

There’s an old saying that money is what makes the world go ‘round. However, to hear some IT professionals tell it, data center availability may soon join the ranks of the societally indispensable.

We do live in a world where the British government is investing in data center infrastructure while Texas is openly courting firms to set up shop in the state, not to mention the almost daily announcement of a new private investment in the data center space.

The world over, data center demand is rising, making it hard to find a single site without firm commitments.

“Suffice to say that the declining availability of data center space is not understated. The data center vacancy is declining year over year. It is at historic lows. Tier one markets will sit at sub 1% vacancy. In some places [it is] perilously little.”

Those are the words of Chris Opat, senior vice president of cloud operations at BackBlaze. He says that areas of the US where data centers have been centralized are very quickly being bought up.

“All but 1% of the build that's scheduled for next year in Northern Virginia is under contract already,” he tells ITPro. “Of what they plan to build, 99% of it's under contract, right? So, these are just cosmic circumstances that I don't think anybody could have foreseen, but they are real.”

To that end, a CBRE report from August last year noted that, despite a 10% jump in supply in primary markets – places like Dallas-Fort Worth, Northern Virginia, and Chicago – the vacancy rate in those data centers over the first half of the year dropped to a never before seen low of 2.8%, half a percent lower than the same period the year before.

At the same time, amid reports that Microsoft was cancelling leases on the equivalent of two data centers, there has been concern in some circles that Microsoft’s decision to redistribute their investment was an indicator of things to come. Bina Khimani, a co-founder of Kinesis Cloud and veteran of AWS and IBM, disagrees with that assessment.

“This Microsoft news is, I think, like a drop in the bucket,” Khimani tells ITPro. “It's not a big thing…sometimes you are shifting your strategic alignment.”

What she does see is a continued need to focus on efficiency to increase the availability of capacity at the same time as the industry recognizes a need to expand and grow into new geographical areas.

“Now you see [a] $2 billion investment in Utah, 400 acre data center. You see data center investments in [the] Nordics because it has more access to renewable energy. You see data center industry growing very rapidly in India.

Khimani adds that innovative approaches to powering these data centers and reducing the load on already existing infrastructure are a growing part of the sectoral picture. The industry is working hard to figure out what it means to best use and build infrastructure, she tells ITPro.

“A lot of money is flowing into it, but it's not always flowing in a smart way. People are still experimenting.”

That forward focus on efficiency bears out in the data. In a November 2024 report Gartner estimated that the capacity available worldwide will more than double by 2028 compared to five years earlier. That equates to an increase of approximately one and a half zettabytes. At the same time, with advancements in optimization and efficiency – some of which is evolving thanks to AI technology – the authors expect the capacity growth rate to drop from 51% in 2024 to 9% in 2028.

Part of the challenge, especially for comparatively small companies, is that hyperscalers like Google and AWS dominate the market. That same Gartner analysis estimates that hyperscalers will be utilising 70% of available capacity within the next three years. A similarly timed report from McKinsey and Company found hyperscaler cloud service providers (CSPs) are driving most of the rising data center demand, with 60-65% of AI workloads expected to be hosted on CSP and hyperscaler infrastructure by 2030.

However, the whole story isn’t just about which data centers are needed or where they’ll be located. Herb Hogue, chief technology officer at Myriad360, says that the clamoring for infrastructure is a case of both hyperscalers dominating the market and companies wanting to make sure they have capacity when they need it down the line.

“There's a bit of a land grab around building these facilities, not because of the facility so much, but because of the power availability,” he explains. “And power availability is driving a lot of the necessity to consume, because if you don't get it, and it's finite, which it kind of is, you're not going to have it later…I think the demand, at least for the next 18 to 24 months, is sustained. Can we, any of us, know further than that? I don't think so.”

The race for efficiency

In the current moment, with private and governmental investment ballooning, it can feel as if the data center industry—particularly as AI drives demand—is moving at a breakneck speed.

Pascal Jaillon, SVP of product in the US for OVHcloud, who spent the early part of his career in France before moving to the US, says that part of the reason the American market moves quickly is because of a lower level of regulation. One area in which he sees that most vividly is in how American labor law makes business reorganization easier.

“The hiring and firing is actually much [more] loose here and allows companies to grow very quickly when there is demand, and to scale back when this demand disappears, where in Europe there is always a bit of latency… that is put in place by those regulations,” he explains.

Jaillon said cloud demand continues to grow, with the example of remote work becoming the norm during the pandemic, and that AI would feed into this trend.

Looking a little into the past, Opat sees similarities between the current lack of availability and the challenges created by the Bitcoin mining boom of just a few years ago.

The staffing shortage

To add more to the pile of problems for prognosticators and business owners across a myriad of supply chains, Hogue sees another issue at play: a low number of skilled workers for data centers to bring new availability to market.

“I think the one thing that gets missed in this hyper cycle of investment is the byproduct of it is it's requiring even more high level, highly skilled people in multiple domains. Whether that be next generation construction, next generation fab building, next generation operation skills or interactive skills.

“You may have a supercar sitting out there in the driveway, but if you don't know how to drive a five speed, it's useless to you.”

So, while there is heavy focus on increasing supply to meet the expanding demands of the moment, there’s also a lack of skilled labour, a set of markets of differing sizes, a power problem, and a construction build-time problem. This is all without mentioning the environmental concerns associated with data centers and the immense amount of water the sites consume to serve workloads such as AI models.

How these seventeen problems in a trenchcoat will end up being solved is yet to be determined. What is clear is that there is a whole industry keen to find out what those solutions might be.

There’s a reason why most coal power plants were built by the coast. It’s the ocean, nature's liquid chiller, where they can pipe cooling seawater through their systems and reduce the heat from production.

Liquid cooling is by no means a new technology, having been first used in the 60s to chill mainframe computers. Most data centers today already use some fashion of liquid cooling, with the average facility consuming roughly 300,000 gallons of water per day.

And, as liquid cooling becomes ever more popular, we must seek ways to implement it in a way that doesn’t cost the Earth. One way to do it is currently being developed in Portugal by the data center company Start Campus which has repurposed the infrastructure of a discontinued coal power plant.

In collaboration with Schneider Electric, Start Campus has seen the construction of a sprawling AI data center campus on the Portuguese coast. The site, which has been operational since October 2024, has been built to be Europe’s largest and most sustainable AI-ready data ecosystem.

The location is crucial; situated in the port town of Sines (pronounced Sin-edge), 100-odd miles south of Lisbon, where there is a deep basin of ocean water to easily pipe through the facility. But the basin is more specific than that – it’s a man-made basin, built to cool the coal mine which still looms in the background of the Sines DC.

Decommissioned in 2021, the power station will soon be pulled down entirely. But the Sines DC will reuse its basic idea for cooling, something quite common in coal and nuclear power stations. Start Campus and Schneider Electric hope it will be a new standard.

“The cooling was a bonus, to be honest,” says Robert Dunn, the CEO of Start Campus. “The two most important things that we looked at when we were considering 100-plus sites were renewable power availability and connectivity. So the power availability was there, we had 500 MW from day one and the transmission network was already there. So we didn't have to wait years to get that power. And then connectivity was here through the cable landing station which is just a few kilometres away.

“So they were the main reasons and then when we looked at this site, we saw a great opportunity to repurpose some of the existing infrastructure and we always wanted to do something that was innovative, sustainable, and efficient on the cooling side. So being able to repurpose a seawater facility was an obvious choice for us.”

The pipes are still being connected to the first installation in the data center, SIN01. Once fully live the system will cycle around 1,000 cubic meters of seawater per minute for cooling the server racks. The water will then make its way back to the ocean at a slightly cooler temperature, to meet EU and Portuguese regulations. No water will be lost or evaporated in the process, according to Start Campus, as it is expelled back into the ocean where the north-to-south current around Portugal will take it away from the basin.

However, when the seawater system is completed, the Sines DC will still not be 100% liquid-cooled, as there will be just too much residual heat. This, as Dunn explains, is because cooling in the data center is getting more complex and energy-intensive. Within the data Hall of the SIN01 there are the liquid cooling loops and heat exchanges, but there is also air cooling technology. This comes by way of the Schneider Electric acquisition of Motivair. This will form the third element of the facility's hybrid and high-efficiency cooling solutions – liquid cooling, air cooling, and heat rejection.

“It’s not just about the seawater that we've put in,” Dunn explains. “Cooling and PUE is probably a small part of what we can do, but we have to look at every decision that we're making. Including the fuel we're using in the generator and what we're doing with the landscaping and just trying to support the local community at the same time. It takes daily attention to make sure that you can be as sustainable as possible.”

Data centers consume some 75 billion liters of water per year, according to the Start Campus. Computation is a hot business and the temperature only goes up when you add AI. But with its innovative use of seawater and the system of a decommissioned coal power plant, the Sines DC in Portugal might just be the coolest AI-ready data center in the world.

Microsoft has published what it said is a first-of-its-kind look at the total environmental footprint of data center cooling technologies, from raw materials to retirement.

The 'cradle-to-grave' life cycle assessment (LCA), published in Nature, tots up the total carbon, energy, and water usage across four major cooling methods: air cooling, cold plates, one-phase immersion, and two-phase immersion.

"In a nutshell, we’re trying to understand the trade-offs," said Teresa Nick, director, natural systems and sustainability for cloud operations and innovation at Microsoft and co-author of the paper. "You’re trying to understand the context of what you’re doing and what the impacts are."

While the standard approach for data centers has been air cooling, the industry has more recently been exploring cooling technologies that rely on liquids, which can dissipate heat much more directly and efficiently than air.

In the case of cold plates, a coolant is pumped in a loop to a flat container that sits directly on top of the chips in a server rack.

One-phase immersion involves operating servers in a tank that has cooling fluid pushed through in a circuit; while in two-phase immersion, server racks are in a tank filled with a different fluid that boils at low temperatures, with the vapor rising to condense, thus cooling, and return to the tank.

Compared with air cooling, the study found cold plates and the two immersion cooling technologies can reduce greenhouse gas emissions by between 15% and 21% over their entire life cycles, along with cutting energy demand by 15% to 20% and water consumption by between 31% and 52%.

However, while two-phase immersion showed potential for improvements in all areas, it currently uses liquid polyfluoroalkyl substances, or PFAS - under regulatory scrutiny in the European Union and the US, meaning that they may be unavailable in the future.

"It was interesting to see that cold plates could be as good as the two immersion cooling methods," said Nick.

Microsoft eyes renewables gains

Meanwhile, researchers found transitioning to 100% renewable energy can cut emissions by as much as 90%, regardless of what cooling technologies were used.

Microsoft said it plans to use the findings to improve new data center designs and cloud operations, and to help meet its broader sustainability goals. It's also making the methodology available to others in the industry through an open research repository.

"Our intention is not to say, ‘this is the right technology.’ They all could be. There are different circumstances that make you use a technology," said Husam Alissa, director of systems technology in Cloud Operations and Innovation at Microsoft and leader of the lifecycle assessment study.

"What we’re trying to do here is tell the industry, ‘Here’s how you build an end-to-end life cycle assessment that takes cooling into account. And here is a tool for you that you can customize to your specific needs and then make a decision'."

MORE FROM ITPRO

• Why tech giants are hitting the brakes on costly data center plans
• $200 billion to build a data center? Soaring AI demands mean costs could skyrocket within a few years
• Data center water consumption is spiraling out of control

Surging AI-related infrastructure demands mean data centers could soon cost as much as $200 billion to build, according to a recent study, with most owned privately rather than by governments.

Research by Epoch AI tracing the computing powering AI by analyzing 500 systems shows that AI supercomputers — Epoch's phrase for AI data centers or GPU clusters — are doubling in performance every nine months thanks to improvements in chips.

"Two key factors drove this growth: a yearly 1.6x increase in chip quantity and a yearly 1.6x improvement in performance per chip," the researchers said in a blog post.

"While systems with more than 10,000 chips were rare in 2019, several companies deployed AI supercomputers more than ten times that size in 2024, such as xAI’s Colossus with 200,000 AI chips.

While computational performance is up 2.5 times per year, power requirements and hardware costs are also doubling annually. Epoch said hardware costs were up by 1.9 times each year, while power needs were doubling annually.

For example, xAI's Colossus cost $7bn to build and uses 300MW of power. On the upside, these systems are becoming more efficient, with computational performance per watt increasing by 1.34 times annually.

"If the observed trends continue, the leading AI supercomputer in June 2030 will need 2 million AI chips, cost $200bn, and require 9 GW of power," researchers said, noting projects like Stargate, the $500bn infrastructure project in the US, mean the funding side should be sufficient, though a lack of power may continue to be a problem.

"To overcome power constraints, companies may increasingly use decentralized training approaches, which would allow them to distribute a training run across AI supercomputers in several locations," the researchers added.

The research highlights two key concerns for AI developers. First, the cost of infrastructure, with critics — and some investors — suggesting too much is being spent; earlier this year, Microsoft said it may take 15 years to see a return on investment for its AI spending and Anthropic CEO has said it may one day cost $100bn to build an AI model.

Second, such systems come with high energy demands that existing power infrastructure may not be able to keep up with. On both sides of the Atlantic, concerns have been raised over the ability for grid networks to keep pace with demand as AI adoption continues to accelerate.

Last year, the CEO of the UK’s National Grid suggested an overhaul of the country’s grid network could be required to accommodate data center infrastructure growth in years to come.

Shift to private systems

Another change is ownership, the study noted. Now, 80% of such machines are owned by companies rather than governments, compared to industry owning 40% back in 2019, the research showed.

According to Epoch's research, the shift in ownership is down to a 2.7-times annual growth rate for AI supercomputers at private companies versus 1.9 times in the public sector.

"In addition to faster performance growth, companies also rapidly increased the total number of AI supercomputers they deployed to serve a rapidly expanding user base," the blog post noted.

The US owns 75% of the total computing power in Epoch's list, with China in second place at 15%.

"Meanwhile, traditional supercomputing powers like the UK, Germany, and Japan now play marginal roles in AI supercomputers," the researchers noted.

"This shift reflects the dominance of large, US-based companies in AI development and computing. However, AI supercomputer location does not necessarily determine who uses the computational resources, given that many systems in our database are available remotely, such as via cloud services."

MORE FROM ITPRO

• How will energy prices affect data center costs in 2024 and beyond?
• Why tech giants are hitting the brakes on costly data center plans
• Is this the future of data center sustainability? New architectures could slash energy consumption, lower carbon emissions, and drive cost savings

Google has said it will invest in training electricians in the US as part of efforts to shore up power systems needed to feed energy hungry AI models.

The move comes as the rise of AI has sparked concerns about its massive energy use and whether existing power systems can keep up with the skyrocketing demand.

Google announced the move in a whitepaper, Powering A New Era of American Innovation, which includes 15 of what the company calls "policy opportunities".

That includes ideas to boost American power infrastructure, including alternative sources of energy, boosting the grid, and investing in building a pipeline of necessary skilled labor.

The whitepaper also called for heightened investment in carbon capture as well as investigating geothermal, natural gas, and advanced nuclear as sources, with Google saying approvals for work and other permits needed to be expedited.

A key talking point centered around upgrades to electricity grids to meet sharp demands as a result of AI.

"By unlocking advanced electricity resources and grid infrastructure, the United States would enable all sources of electricity to contribute to a more reliable and affordable energy future," Google said in a blog post.

"They now require constructive public and private collaboration to advance."

In the paper, Google pointed to the Federal Energy Regulatory Commission last year tripling its five-year demand forecast, with energy demand from AI driving 15-90 gigawatts of new US data centre demand by 2030.

That said, Google also mentioned that energy intensity for AI workloads was decreasing by a fifth each year, down to improvements in hardware and models.

Google's big recruitment drive

Google said getting power networks and supply ready to meet the looming demand of AI needed effort now, pointing to a McKinsey estimate that suggested an additional 130,000 electrical workers would be needed by 2030 to build out data centers and manufacturing in the coming years.

But only 7,000 people enter the field annually and around 10,000 depart through retirement or career changes.

"In particular, a shortage of electrical workers may constrain America’s ability to build the infrastructure needed to support AI, advanced manufacturing and a shift to clean energy," Google said in the blog post.

To address that, Google said its charitable arm, Google.Org, would financially support training for 100,000 electrical workers and a further 30,000 new apprentices via the electrical training Alliance (etA), an educational group created by the International Brotherhood of Electrical Workers and the National Electrical Contractors Association.

Reports from Reuters suggest the funding could total around $10 million annually. That support will help the etA increase the number of electrical workers available by 70% within five years, Google said.

MORE FROM ITPRO

• Google signs fresh renewables deal amid surging energy consumption
• Google's greenhouse gas emissions have skyrocketed in the last five years
• Google is going nuclear

Amazon Web Services (AWS) has paused plans for some data center leases, according to analysts, sparking further concerns about the cost of AI infrastructure spending plans.

According to reports from Reuters, Wells Fargo analysts spotted that AWS had paused discussions over data center leases in overseas markets specifically. It wasn't clear, the analysts said, exactly how much capacity was being reconsidered.

The AWS report follows similar moves by Microsoft, which were seen by analysts as a sign of weaker than expected demand for AI or a reconsideration of the costs — perhaps no surprise given the recent turmoil in the tech industry sparked by US tariffs.

Amazon and Microsoft have both pledged to significantly increase their spending on data centers amid the AI boom, with AWS saying it would spend $83 billion and Microsoft $80bn this year alone.

The latter has stressed that investment isn't being cut back despite data center lease changes.

AWS is letting data center plans simmer

Wells Fargo analysts said the AWS rollback was similar to Microsoft's own aren’t outright cancellations at this stage. Instead, the company appears to be "digesting aggressive recent lease-up deals".

"Over the weekend, we heard from several industry sources that AWS has paused a portion of its leasing discussions on the colocation side (particularly international ones)," Wells Fargo reportedly said in its client note.

"It does appear like the hyperscalers are being more discerning with leasing large clusters of power, and tightening up pre-lease windows for capacity that [would] be delivered before the end of 2026.”

Analysts noted that Meta, Google, and Oracle all remain active in data center leasing.

Meta, for example, has pledged to spend $65 billion in data center investment this year alone while Google recently reiterated its plans to shell out $75 billion on such infrastructure.

Similarly, Oracle is helping with Project Stargate, a government-backed $500 billion project to build more data center infrastructure in the US over the next five years.

Amazon is still bullish on AI demand

When approached for comment on the reports, an AWS spokesperson pointed ITPro to a recent LinkedIn post by Kevin Miller, AWS vice president for global data centers.

Miller said the tech giant still intends to charge ahead with infrastructure expansion plans and that the company continues to see “strong demand” for generative AI and foundational workloads.

That said, Miller noted there were indeed changes in data center plans but stressed this move represents “routine capacity management”.

"There haven’t been any recent fundamental changes in our expansion plans," he said.

"We have almost two decades of experience delivering data center capacity to meet customer demands, when and where they need it.

"That experience has taught us to consider multiple solutions in parallel. Some options might end up costing too much, while others might not deliver when we need the capacity. Other times, we find that we need more capacity in one location and less in another."

MORE FROM ITPRO

• What is a green data center and why are they attracting big investment?
• AWS just rolled out Amazon Q Business in Europe – and it includes new data residency features
• Pressure rises to bolster data center energy efficiency as IT leaders worry demand is pushing the grid to breaking point

The energy demand of AI data centers will top that of Japan by the end of the decade, new research shows – and that’s providing that energy grids can even keep up.

A new report from the International Energy Agency (IEA), which sought to quantify the impact of AI on energy networks, noted that investment in data centers has nearly doubled since 2022, with data center electricity use up by 12% since 2017.

That's set to double by 2030, largely driven by the rise of AI as the tech industry steams ahead with adoption.

"Global electricity demand from data centers is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today," said IEA Executive Director Fatih Birol.

Varied demand

The IEA report noted that a typical AI data center consumes as much electricity as 100,000 households — but the largest under construction now may chew through 20 times as much.

However, the report noted that data centers are just one driver behind the growth of global electricity demand, making up about one-tenth of that demand — less than industrial motors, air conditioning, and electric vehicles.

Developing economies may see less impact from data centers, as their electricity demand growth was already high; advanced economies, which had largely stagnant electricity demand, will experience a more dramatic impact.

"The effects will be particularly strong in some countries," Birol added. "For example, in the United States, data centers are on course to account for almost half of the growth in electricity demand; in Japan, more than half; and in Malaysia, as much as one-fifth."

The report noted that in the US, by the end of the decade, data centers will use more electricity than aluminum, steel, cement, chemicals, and all other energy-intensive goods combined.

Energy grids are creaking

The IEA warned that strain on electricity grids could delay as many as a fifth of data center projects, noting that there are already long waits for connecting to grids and access to critical grid components such as transformers and cables. For gas generation, turbine deliveries already face lead times of several years.

"If the electricity sector does not step up, there is a risk that meeting data center load growth could entail trade-offs with other goals such as electrification, manufacturing growth or affordability," the IEA said.

This isn’t an issue isolated to the US, either. Last year, the head of the UK’s National Grid warned the country’s energy network was already struggling to keep up with rising demand - and that’s a trend expected to continue across the decade unless major investment is secured to facilitate upgrades.

One solution presented by the IEA included choosing better locations for new data centers, looking for opportunities in areas with power and grid availability.

"Grid operators could also examine incentives to locate data centers in areas where grids are less constrained," the IEA said. "We find that 50% of data centers under development in the United States are in pre-existing large clusters, potentially raising risks of local bottlenecks.

Beyond those direct concerns, the IEA also warned about security. Attacks on critical national infrastructure have increased rapidly in recent years, and with data centers now representing a crucial cog in the modern economy, the risks are growing.

"Cyber attacks on energy utilities have tripled in the past four years and become more sophisticated because of AI," the IEA said in a post. "At the same time, AI is becoming a critical tool for energy companies to defend against such attacks."

AI could be the curse and the cure

AI could help address some of these challenges by optimizing existing sources, reducing consumption, and boosting grids — such efforts are already in place, with the UK government's AI Energy Council tapping AI leaders including Microsoft, Google, and Amazon to help tackle local energy concerns.

The report said that AI-based fault detection could reduce outage durations by 30-50%, while remote sensors paired with AI management could increase the capacity of existing transmission lines.

"Up to 175 gigawatts (GW) of transmission capacity could be unlocked if these tools are applied, without any new lines being built," the report said.

Perhaps highlighting the impact of AI, the IEA also unveiled a new AI agent to help answer questions on the report's contents, though it warns that "AI can make mistakes, and you should verify the answers independently with the original source materials."

MORE FROM ITPRO

• Data centers are growing in size and number as AI prompts widespread global expansion
• What does a data center look like in the AI era?
• Europe needs more energy and better grids to meet data center power demands in the age of AI