The hyperscalers have collectively raised their 2026 capital spending guidance to $725 billion, up 77 percent from 2025, pushing combined free cash flow to roughly $4 billion, the lowest since 2014 when their revenues were one-seventh of today’s scale.
The numbers are staggering even for companies accustomed to scale. Microsoft guided to $190 billion in capex for calendar 2026, well above the $152 billion analyst consensus, while citing $25 billion in price inflation from memory chips and components. Meta raised its forecast to $125 to $145 billion, Alphabet to $83 to $93 billion, and Amazon held at $200 billion. Those commitments are mostly directed toward AI infrastructure, including data centers, servers, GPUs, networking equipment, and software. The four companies are now investing more in physical hardware in a single year than many countries spend on infrastructure.
The cash flow impact is immediate and visible. Analysts project Alphabet’s free cash flow to fall 90 percent, Meta’s to turn negative in 2027 and 2028, and the group as a whole to reach its lowest point since 2014. That is a remarkable shift for corporations that built their reputations as asset-light cash machines. The change reflects how deeply AI has transformed their operating models. Cloud revenue growth remains strong, with Alphabet’s cloud up 63 percent to $20 billion, Microsoft’s AI business at a $37 billion annual run rate, and Amazon’s AWS showing the strongest growth since 2022. But those gains are now being consumed by capex that runs ahead of revenue recognition.
For SF readers, the hyperscaler capex wave is the financial stress test behind the AI infrastructure race. Startups, GPU clouds, data center developers, and model labs all depend on this spending continuing unabated. When Microsoft, Alphabet, Amazon, and Meta commit $725 billion to data centers and chips, it creates a massive tailwind for the entire ecosystem. The risk is that public-market patience has limits, and tolerance for negative free cash flow is not infinite. If investors start to question the return profile of these investments, the funding cycle could tighten faster than expected.
The AI capex burden is becoming Big Tech’s new margin tax. Historically, these companies generated free cash flow margins above 25 percent. Now they are trading those margins for capacity that they hope will generate future revenue and competitive advantage. The logic holds as long as cloud AI revenue keeps accelerating and the market believes the spend is necessary to stay in the race. But the math is unforgiving. A company that spends $190 billion on capex while generating $37 billion in AI revenue run rate needs the revenue to compound quickly to justify the outlay. If growth slows or competition erodes pricing power, the cash flow hole becomes harder to fill.
That exposure lands hardest on downstream startups. GPU cloud operators, data center builders, and model-infrastructure providers all assumed hyperscaler spending would remain a reliable demand signal. If public-market scrutiny forces a capex slowdown, those businesses face a revenue cliff. Infrastructure startups with long sales cycles and fixed costs are particularly vulnerable. A model lab that depends on AWS Trainium or Google TPUs sees its economics shift if the hyperscalers decide to prioritise short-term cash flow over capacity expansion. The entire AI startup ecosystem is now priced for hyperscaler aggression. A return to capital discipline changes the game for everyone downstream.
The market reaction so far has been mixed but telling. Alphabet and Meta shares held up better than Microsoft and Amazon after earnings because investors saw clearer paths to monetisation in search, advertising, and cloud backlog growth. That suggests the market is starting to reward execution over spending scale. Founders should read that as a warning. The hyperscalers can tolerate negative free cash flow longer than most. But if investor patience fades, the ripple effects will hit the startup economy first.
Also read: Vietnam’s AI propaganda blueprint is a preview of how states will weaponize the creator economy • AWS overheating in Virginia is a reminder that cloud reliability is now a physical problem • SoftBank’s OpenAI margin loan is the most leveraged bet in the history of artificial intelligence