‘In the model race, it still trails’: Meta’s huge AI spending plans show it’s struggling to keep pace with OpenAI and Google – Mark Zuckerberg thinks the launch of agents that ‘really work’ will be the key

Meta plans to invest as much as $135 billion in AI infrastructure this year, nearly doubling its capital expenditure, as it seeks to catch up with key competitors in the generative AI race.

Meta posted results with $59.89bn in quarterly revenue, topping analyst expectations of $58.41bn. For the full year, Meta posted revenue just above $200bn, up 22% from the year before, with net income of $60bn versus $62bn a year before.

"We had strong business performance in 2025," said Mark Zuckerberg, Meta founder and CEO, in a statement. "I'm looking forward to advancing personal superintelligence for people around the world in 2026."

In a call with analysts after the results were announced, Zuckerberg predicted "a major AI acceleration" including the launch of agents that "really work", adding that Meta would release new models and products in the coming months.

"I expect our first models will be good, but more importantly, we’ll show the rapid trajectory that we’re on," he said. "Our vision is building personal super intelligence."

Zuckerberg suggested that might show up in the company's smart glasses or other wearables — an area OpenAI is hoping to enter this year too.

In that call, CFO Susan Li said Meta was putting up the cash to back that up with capital expenditures in the range of $115bn to $135bn.

Last year, Meta was reportedly set to spend around $60bn, but that was increased to up to $72bn later in the year. Microsoft and Alphabet spent between $80bn and $90bn, while Amazon's forecasted spend was expected to top $100bn.

Meta is scrambling to catch up

Meta's results came alongside quarterly reports from AI rivals including Microsoft, which saw shares slide amid concerns from investors that spending too much on AI and affiliated infrastructure was impacting growth.

The bigger concern for Meta may be that it's at risk of falling behind due to lack of investment, according to Mike Proulx, VP researcher director at Forrester.

"Meta, just like every other big tech company, is now first defined by AI. In the glasses race, it’s ahead. But in the model race, it still trails," Proulx said. "If Meta wants to deliver on its 'personal superintelligence' vision, it needs to catch up with Google and OpenAI.

“That means heavy investment and real focus."

Proulx added that Meta shifting investment out of the metaverse was a "positive move", but suggested that the messaging from Zuckerberg on upcoming model releases might not fill users with confidence.

"Meta also tried to temper expectations around its next set of models, calling them merely 'good' and hinting that bigger upgrades are coming later. But by then will it be too late?"

Meta's Llama model range proved highly popular with users when first released in 2023, and the company pivoted hard to focus on generative AI alongside industry rivals.

Subsequent releases have proved controversial, however. The launch of Llama 4, for example, saw company executives deny claims that model benchmarking results had been exaggerated.

The pace of change in the industry has also been rapid, with big tech providers such as Google, OpenAI, and Anthropic all engaged in a war of attrition to roll out more powerful and efficient models.

OpenAI's initial lead alongside Microsoft on this front has since been whittled down by competitors such as Anthropic and Google.

Over the last 18 months, the latter has made great strides in catching up with the duo, and some industry analysts suggested that the launch of its Gemini 3 model range in November placed it firmly in the lead.

FOLLOW US ON SOCIAL MEDIA

Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.