Meta raises 2026 capex to $125–145B

Meta’s AI bill just got even bigger. On April 29, the company raised its 2026 capital-expenditure outlook to $125 billion to $145 billion, up from $115 billion to $135 billion, even though first-quarter capex came in lighter than Wall Street expected. The stock dropped anyway. That tells you what investors are worried about — not whether Meta can spend, but whether this spending race is getting too expensive to justify. What changed here is simple. Meta did not just say “AI remains important.” It widened the top and bottom of its capex range by $10 billion each, and management tied the increase to higher component pricing and extra data-center costs needed for future capacity. In the same quarter, multiyear infrastructure commitments jumped by $107 billion. That is the louder signal than the guidance line itself — Meta is locking in supply, cloud capacity, and buildout plans years ahead. Why did the market hate it if the quarter was strong? Because the operating business looked great. Revenue rose 33% to $56.3 billion. Ad pricing and impressions were both up. Operating income hit $22.9 billion. In other words, investors were not reacting to a weak core business. They were reacting to the idea that even a booming ads machine may now have to pour astonishing amounts of cash into compute just to stay in the race with Alphabet, Amazon, and Microsoft. The awkward part is that Meta’s capex increase came after a quarter when actual capex was only $19.8 billion, below analyst expectations. Normally, missing on spending would calm people down. But Meta’s explanation went the other way — the lower quarter did not mean lower needs, it meant an even more aggressive full-year catch-up. Basically, investors heard: the bottleneck is not ambition, it is timing, pricing, and how fast Meta can get hardware and data centers online. So what is Meta buying with all this money? Mostly the boring but crucial stuff — servers, networking gear, data-center capacity, and cloud contracts. The company has been expanding its own footprint while also signing external cloud deals that stretch into 2027. That matters because AI demand is no longer just about training one flashy model. It is about running inference at scale across Facebook, Instagram, WhatsApp, ads tools, and the Meta AI app. If usage keeps climbing, the infrastructure has to be there first. Where do the reported layoffs fit? They matter, but not in the clean “8,000 jobs were cut to fund capex” way the internet likes to frame things. Meta’s official earnings commentary said it plans to reduce headcount in May while keeping the full-year expense outlook unchanged at $162 billion to $169 billion. That suggests management is trying to protect margins while redirecting spending toward infrastructure, not making a neat one-for-one swap between payroll and GPUs. The tradeoff is real — but it is broader than one layoff number. The supplier angle is where this gets more interesting. A company willing to sign $107 billion more in contractual commitments is good news for chip, networking, power, and data-center vendors. But the catch is that hyperscaler customers are getting more demanding at the same time. If component prices are rising and future-year capacity is getting booked early, every vendor now has to prove not just that it can ship, but that it can ship on time, install fast, and produce measurable utilization. The bottom line is that Meta just made the AI arms race feel more expensive, not less. The business is strong enough to fund it for now. But the burden of proof has shifted. Meta now has to show that this mountain of capex turns into durable product gains, ad gains, and user growth — not just bigger clusters and bigger promises.