Meta Raises 2026 AI Capex to $145B: More Than Its Last Two Years Combined

Meta's Q1 2026 earnings brought an announcement that rattled investors but told a clear story about where the AI arms race is heading. The company raised its 2026 capital expenditure guidance to $125-145 billion — up from an already-aggressive prior forecast of $115-135 billion — citing AI infrastructure as the primary driver. The new figure is larger than Meta's combined capital expenditure for all of 2024 and 2025.

What Meta Is Spending On

The bulk of the increase is going toward AI infrastructure: data centers, custom chips, GPUs, and the physical facilities to house them. Meta attributed the raised guidance to "higher component pricing this year" and "additional data center costs to support future year capacity."

That last phrase is the tell. Meta isn't just building for current workloads. It's pre-investing in infrastructure that will handle demand it expects in 2027 and beyond. That's a calculated bet that AI adoption will continue to accelerate and that securing infrastructure capacity early — even at premium cost — is worth the upfront investment.

The spending funds Meta's AI efforts across several fronts: its Llama open-source model program, the Meta AI assistant (now one of the most-used AI products globally), the Reality Labs division, and the growing Meta Superintelligence Labs initiative focused on developing frontier AI capabilities.

Why the Stock Fell

Markets reacted negatively, with Meta stock dropping roughly 6% in after-hours trading following the announcement. The reaction reflects a real tension in how investors are thinking about AI spending right now.

On the positive side, Meta's AI investment has produced visible returns. Meta AI has attracted hundreds of millions of users. Ad targeting improvements powered by AI have driven substantial revenue growth. The Llama models have created a developer ecosystem that builds on Meta's platforms, increasing strategic leverage.

On the other side, $145 billion is an enormous number, and investors want a cleaner path from infrastructure investment to earnings per share. At this scale, even modest execution missteps carry major financial consequences. The concern isn't whether AI is important. It's whether $145 billion, deployed at this pace, is the right amount at the right time.

The Bigger Picture: Big Tech's Infrastructure Race

Meta's spending doesn't exist in isolation. Microsoft, Alphabet, and Amazon have collectively signaled roughly $725 billion in capital expenditures for 2026, almost entirely earmarked for AI. Some analysts now project that Big Tech capital expenditure could top $1 trillion by 2027.

This is one of the largest coordinated infrastructure buildouts in history. The companies involved aren't just betting that AI is important — they're betting that whoever builds the most infrastructure earliest wins a durable competitive advantage, because training the next generation of frontier models will require compute that takes years to build. The cost of being late to that buildout is potentially worse than the cost of overbuilding.

That risk of overcapacity exists, but it looks increasingly manageable given the pace of AI product adoption over the past two years.

What This Means for the Broader AI Industry

Meta's spending decision ripples through the entire AI supply chain.

For NVIDIA, Broadcom, and custom chip manufacturers, Meta's commitments translate directly into chip orders and revenue visibility. It's a key reason why NVIDIA's revenue projections remain high even as broader stock market conditions fluctuate.

For data center operators and power infrastructure companies, Meta's continued investment validates the ongoing buildout of AI-specific facilities — the same category of spending represented by Nvidia's recent $2.1 billion IREN deal and $3.2 billion Corning partnership announced this week.

For AI developers and startups, Meta's infrastructure investment indirectly expands what's possible. More compute at Meta means more training runs for Llama models, which means better open-source foundations for applications built on top of them.

Meta's Strategic Position

Meta's spending increase also signals something about competitive positioning. Unlike OpenAI or Anthropic, Meta isn't primarily trying to sell access to a frontier model. It's using AI to improve its advertising business, build its consumer AI products, and establish strategic credibility in the AI ecosystem through Llama.

That dual goal requires staying competitive on frontier AI training (to produce capable Llama models) while also excelling at applied AI (to improve Meta AI as a product). Both require substantial infrastructure. The $145 billion figure reflects the cost of running both tracks simultaneously at global scale.

What's Next

Meta has indicated this infrastructure investment trajectory will continue through at least 2027. Alongside the capex increase, the company also disclosed plans to cut approximately 8,000 employees in May, part of a broader restructuring in which AI-driven efficiency is reducing headcount while AI-related infrastructure investment simultaneously expands.

That juxtaposition — fewer employees, dramatically more infrastructure — captures the current moment in enterprise AI precisely. Companies are using AI to do more with less while simultaneously betting that the infrastructure required to run AI will define competitive advantage for the next decade.

Bottom Line

Meta's $145 billion bet is simultaneously a statement of confidence in AI's continued growth and a reminder of how high the stakes have become. For the companies building AI products, this level of infrastructure investment shapes what's technically and economically possible. For investors, it raises real questions about return timelines. For the rest of the AI industry, the message is unambiguous: the infrastructure race isn't slowing down — it's accelerating.