Big Tech's AI Tab Hits $650B for 2026
Amazon, Alphabet, Meta, Microsoft, and Oracle have committed a combined $650 billion in capital spending for 2026, with 75% aimed at AI infrastructure.
The final numbers are in, and they're staggering. With Amazon's $200 billion capex announcement on Thursday, the five largest cloud infrastructure companies have now collectively committed roughly $650 billion in capital expenditure for 2026. That figure — larger than the GDP of Sweden — represents a 36% increase over 2025 levels and dwarfs anything the technology industry has ever attempted in a single year.
The Breakdown
| Company | 2026 Capex | YoY Change | Stock Reaction |
|---|---|---|---|
| Amazon | ~$200B | +60% | -8% |
| Alphabet | $175–185B | +40% | -3% |
| Meta | $115–135B | +87% | +8% |
| Microsoft | ~$105B* | +35% | -11% |
| Oracle | ~$20B+ | +50%+ | — |
*Microsoft fiscal year ends June; estimate covers FY ending June 2026.
CreditSights estimates that approximately 75% of this spending — around $450 billion — goes directly to AI infrastructure: GPUs, networking equipment, servers, and data centers. The rest covers maintenance capex, non-AI cloud expansion, and other infrastructure.
The Market Can't Decide If This Is Bullish
The divergent stock reactions tell the story. When Meta and Microsoft reported last week, the market rewarded Meta's spending plan with an 8% rally while hammering Microsoft's stock 11%. This week, Amazon dropped 8% on its $200 billion forecast — $55 billion above consensus — while Alphabet fell 3% despite otherwise strong Q4 numbers.
The pattern: investors accept AI spending when it comes with clear revenue inflection (Meta's ad targeting improvements, Alphabet's 24% AWS growth). They punish it when the return on investment timeline feels murky.
Amazon's case is the most extreme. Morgan Stanley now projects Amazon will post negative free cash flow of nearly $17 billion in 2026. Bank of America sees a $28 billion deficit. For a company that spent decades conditioning investors to accept thin margins in exchange for free cash flow, that's a genuine paradigm shift.
The Beneficiary Chain
Someone's cashing $650 billion in checks, and semiconductor stocks rallied hard on Friday as the market figured out who benefits. Nvidia surged 8% after CEO Jensen Huang appeared on television declaring AI demand is "going through the roof." AMD gained 8%. Broadcom added 6%.
The spending also flows to power companies, real estate operators, and construction firms. Caterpillar hit a record high last week on earnings that showed its data center infrastructure backlog reached $51.2 billion. Data center REITs, cooling system manufacturers, and electrical equipment companies are all downstream beneficiaries.
But there's a paradox embedded in this spending boom. The same AI capabilities these companies are building caused the SaaSpocalypse that wiped $285 billion from software stocks this week. The hyperscalers are spending $650 billion to build AI systems that may undermine the subscription software model that many of their enterprise customers depend on.
The Bear Case Is Simple
$650 billion is a bet that AI revenue scales fast enough to justify the investment. If it doesn't — if the AI monetization timeline stretches to 2028 or beyond — these companies will have built expensive infrastructure with uncertain returns. The dot-com era offers an imperfect but relevant comparison: massive fiber-optic buildouts in the late 1990s eventually proved useful, but the companies that built them went bankrupt first.
The bull case is equally straightforward: demand for compute is genuinely supply-constrained today, enterprise AI adoption is accelerating, and the companies making these investments generate enough cash to absorb the capex without existential risk. Meta generates $60 billion in annual free cash flow. Alphabet generates $70 billion. Even with aggressive spending, these aren't companies burning through their last dollar.
What's missing from both cases is clarity on unit economics. How much revenue does a dollar of AI capex generate? Nobody can answer that with confidence yet, which is why the market reaction has been so uneven. Until that number comes into focus, expect continued volatility around every earnings report that includes a capex revision.