Ross Kelly
Meta’s recent “chaotic” restructuring efforts have taken another turn amidst reports the company has frozen hiring in its AI division.
Reports from the Wall Street Journal noted that the tech giant has paused hiring in its AI unit following the arrival of more than 50 staff.
The move follows reports of a huge restructuring effort at the company earlier this week, with The Information detailing plans to split its AI unit into four dedicated subdivisions, marking the latest overhaul in its broader AI strategy.
Meta formed its Superintelligence Labs in June, to be headed up by former GitHub CEO Nat Friedman and ex-Scale AI chief executive Alexandr Wang, and backed with billions of dollars in funding.
Under the changes reported by The Information, the new departments are expected to be an infrastructure team, a product team focused on offerings such as Meta’s AI assistant for consumers, Meta’s Fundamental AI Research (FAIR) lab, and a ‘TBD’ lab for unspecified projects.
It is likely that the final of the four will turn its attention to new frontier AI models aimed at exceeding the performance of its current Llama range.
Meta targets stronger focus on AI development
Speaking to ITPro earlier this week, Steve Grant, managing director at Figment, said Superintelligence Labs, with its newly-formed subdivisions, could mark a major reset for Meta, moving away from its iterative free model releases and closer to a focus on AGI alongside other frontier labs.
Grant added that the overhaul essentially suggests a "deliberate shift” toward a more sharpened focus on AI development.
“Each team handles a specific area, from developing new language models to integrating AI into products, scaling infrastructure, and pursuing long-term research,” he said.
“This structure is designed to make innovation more manageable and help the company respond faster to internal challenges, including staff turnover and previous model releases that did not meet expectations.”
By allocating specialist teams, Grant noted that Meta could benefit from a direct route to competing in the AI race, but that this would depend on targeted investment and harnessing the right talent.
Meta’s recruitment strategy has grabbed headlines in recent months, with the company embarking on a major hiring spree across the year so far.
Reports have detailed the company’s aggressive approach on this front, attempting to poach top industry AI experts with eye-watering pay packages reaching over a billion dollars.
The Wall Street Journal recently reported that Meta was targeting OpenAI researchers with attractive pay offers to fill positions at TBD Lab, for example.
Elsewhere, a whopping billion-dollar plus pay package for an exec at former OpenAI CTO Mira Murati’s startup shocked the industry.
Meta appears more than content to splash the cash on this front and across other key areas, such as infrastructure. In its Q2 earnings document, Meta predicted its capital expenditure for 2025 to hit between $64 and $72 billion, with higher figures expected in 2026.
The firm said it is “aggressively pursuing opportunities to bring additional capacity online to meet the needs of our artificial intelligence efforts and business operations”.
This includes a 5GW data center intended to significantly boost its AI training and inference capacity.
With reports now emerging of a hiring freeze, this adds further confusion to an already convoluted situation for Meta. The tech giant has been keen to emphasize this is a routine move, however.
According to reports from Reuters, a spokesperson for the company said this is nothing more than “some basic organizational planning” aimed at “creating a solid structure for our new superintelligence efforts”.
Steve Wilson, Chief AI and product officer at Exabeam, told ITPro the company’s recent strategy “has been chaotic”, noting that its huge investment and recruitment campaign is yet to show any tangible benefits.
“Despite pouring billions into labs, talent raids, and data centers, they’ve failed to translate any of it into meaningful business wins,” he said.
“They had a clear early lead with Llama in open source GenAI, yet never connected it to their core platforms or revenue streams,” Wilson added. “Meanwhile, OpenAI owns consumer mindshare, Grok grabbed social media (which should have been Meta’s backyard), Google tied AI to search, and Anthropic is making headway in developer tools.”
Is Meta sweating about catching up?
Aggressive hiring up to now and sizable investment on the AI front may suggest the company is aware that it’s fallen behind key competitors in the space. In late June, Demis Hassabis, CEO at Google DeepMind, suggested Meta’s hiring spree shows it’s lagging behind in the AI race during a podcast appearance.
Hassabis added that researchers could think twice about accepting a position at Meta if they’re seeking to help shape AI safety.
If Meta is conscious of its failings on this front, bringing FAIR directly into the fold at Superintelligence Labs could also signal a shift designed to ease fears about loose guardrails or lack of ethical oversight when it comes to Meta’s AI development.
More broadly, Superintelligence Labs will work toward Meta CEO Mark Zuckerberg’s stated aim of producing “personal superintelligence that empowers everyone”.
Though Zuckerberg hasn’t explained what this means in practical terms, he has predicted widespread upheaval in tech jobs, stating in January that mid-level engineers at Meta could be matched by AI before the end of 2025.
Meta’s AI efforts have courted controversy
Meta’s AI offerings have been widely embraced by the industry over the last two years. The tech giant’s models, which boast performance comparable to lighter models by Google and OpenAI, have also been used by the likes of DeepSeek to train its own models.
Despite the enthusiasm, it’s faced an equal amount of pushback from critics around how it’s framed these models. Llama 4, currently available as its lightweight Scout and mid-tier Maverick variants, was the subject of some scorn in the developer community when it was released.
Specialist communities on Reddit criticized the latest Llama models for their poor performance relative to competitor models and users have accused Meta of burning its goodwill with the locally-run AI community.
The backlash was such that Meta’s VP of generative AI Ahmad Al-Dahle felt the need to publicly deny claims the company skewed performance metrics for Llama 4 Maverick on public leaderboards - such as LMArena - by training it on test sets.
Similarly, the firm has also faced accusations of ‘open washing’ – with the Open Source Initiative noting in February that despite being advertised as open source, Llama models fall short of that definition.
Wilson noted that the company’s open source credibility is “being eclipsed” by alternatives on the market, particularly DeepSeek. This, he suggested, is another area where the company is moving aggressively to reduce criticism and position itself as a viable enterprise provider.
“Zuckerberg is trying to brute-force his way back into relevance with money and restructuring, but it’s the mark of a company that squandered its edge and is scrambling to keep pace."
Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.
MORE FROM ITPRO
-
AI isn't taking anyone's jobs, finds Yale study – at least not yetReviews Researchers say it's too soon to know what generative AI's impact will be on the workforce
-
A new 'top-tier' Chinese espionage group is stealing sensitive datanews Phantom Taurus has been operating for two years and uses custom-built malware to maintain long-term access to critical targets
-
Huawei executive says 'we need to embrace AI hallucinations’News Tao Jingwen, director of Huawei’s quality, business process & IT management department, said firms should embrace hallucinations as part and parcel of generative AI.
-
The rise of the UK’s ‘invisible’ AI workforceNews Research from Multiverse shows AI skills are becoming common in non-tech roles
-
Why Nvidia’s $100 billion deal with OpenAI is a win-win for both companiesNews OpenAI will use Nvidia chips to build massive systems to train AI
-
Shadow AI can be a tool for AI innovation with the right controls, say Gartner analystsNews Data-driven messaging and a supportive approach to securing AI tools are necessary for security staff looking to balance AI risks and unlock better funding
-
Zoom CEO Eric Yuan thinks AI will pave the way for a three-day week – and he’s not the only big tech exec excited about reduced working hoursNews Yuan joins Nvidia CEO Jensen Huang and Bill Gates in touting the potential for AI to unlock reduced working hours
-
DeepSeek’s R1 model training costs pour cold water on big tech’s massive AI spendingNews Chinese AI developer DeepSeek says it created an industry-leading model on a pittance
-
Global AI spending is set to hit $2 trillion next year – here’s where all the money is goingNews Huge spending increases in AI services and associated infrastructure are expected
-
OpenAI just revealed what people really use ChatGPT for – and 70% of queries have nothing to do with workNews More than 70% of ChatGPT queries have nothing to do with work, but are personal questions or requests for help with writing.
