# The five biggest cloud builders are spending every dollar they earn on AI and then some > Source: > Published: 2026-07-23 23:43:04+00:00 *Microsoft, Amazon, Alphabet, Meta, and Oracle have pledged a combined $725 billion in AI infrastructure spending for 2026, and for the first time their collective free cash flow is converging on zero , a moment Wall Street has been waiting to test.* Amazon's trailing free cash flow has fallen from roughly $38 billion to $1.2 billion. That's a 95% collapse in a single year, driven almost entirely by the company's $200 billion capital expenditure commitment for 2026, the largest infrastructure bet in its history. Morgan Stanley projects Amazon's full-year 2026 free cash flow at negative $17 billion. Bank of America's number is more alarming still: negative $28 billion. These are not rounding errors. They are a signal. The five largest cloud builders, Microsoft, Amazon, Alphabet, Meta, and Oracle, are collectively spending at a pace 70% faster than their cash earnings are growing, according to figures reported by CNBC and tracked by analysts at Morgan Stanley. The $725 billion figure is already being revised upward: Morgan Stanley's latest estimate now puts total 2026 hyperscaler capex closer to $805 billion, with forecasts for 2027 approaching $1.1 trillion. Alphabet's Q1 2026 free cash flow fell 47% year over year to $10.1 billion. The direction across the group is consistent. The build is outrunning the business. Not all five are equally exposed. Meta and Microsoft are generating enough operating cash to fund large portions of their buildout internally, even as margins tighten. Meta raised its 2026 capex guidance to between $125 billion and $145 billion, blaming higher memory-chip costs and additional data center construction, but its advertising business still throws off cash at a rate that cushions the hit. Microsoft, guided to roughly $190 billion in capital spending, has Azure's recurring enterprise contracts as ballast. Amazon is in a different position. AWS grew at its fastest rate in 15 quarters in Q1 2026, and advertising revenue topped $70 billion for the trailing year. Those are real bright spots. But the gap between what Amazon earns and what it's committing to spend has become structural rather than cyclical. To fill it, the company, like its peers, is turning to the bond market. Morgan Stanley found that AI-related issuers had raised approximately $236 billion in debt by the end of May 2026, roughly four times the pace of the same period a year earlier. The bank projects the full-year total will approach $570 billion. Hyperscalers are building the infrastructure of the AI economy on borrowed money. Oracle, the smallest of the five by revenue, carries the most asymmetric risk. Its inclusion in the group at all reflects how aggressively it has committed to AI infrastructure, including a stake in the Stargate data center project alongside SoftBank and OpenAI. But Oracle's operating cash flow base is far narrower than Amazon's or Microsoft's. PIMCO has projected that by 2026 and 2027, capex across the five hyperscalers will collectively reach approximately 94% of operating cash flow. That's essentially the entire earnings stream reinvested before a dollar reaches shareholders. The Nasdaq registered that concern plainly. In the week ending June 27, 2026, the index fell sharply as investor confidence in the sustained build-and-they-will-come thesis cracked for the first time since 2024. Alphabet dropped between 6% and 10% across two sessions. Amazon shed 4%. Apollo Global Management's Torsten Sløk had already flagged the risk: a timeline mismatch between when the capex is committed and when the revenue to justify it arrives. ## What cash-negative cloud providers mean for everyone downstream If you're a startup or an enterprise buying AI compute from any of these companies, the capex crisis looks abstract right now. Your contract still works. Your API still responds. But the structural dynamics are worth watching because they don't stay abstract forever. Concentration risk is real and increasingly documented. OpenAI and Anthropic are two of the largest purchasers of hyperscaler AI infrastructure at this moment, and neither is profitable at scale. If AI startups can't monetize their products at the pace needed to service their cloud commitments, hyperscalers face a revenue collection problem on top of a capital allocation problem. PIMCO and CreditSights have both flagged this as an emerging systemic consideration, not a tail risk. For enterprise buyers, the sharper questions aren't about which accelerator a cloud provider offers. They're about capacity reservation terms, what the contract says when you need to exit it, whether your region is actually available when demand spikes, pricing escalation clauses, and what happens to your stack if a provider's balance sheet pressure forces infrastructure de-prioritisation. A vendor burning cash at this scale has different incentives than a vendor generating it. That's a procurement reality, not a prediction of collapse. The Q2 earnings reports from these five companies, most expected in late July and early August 2026, will be the first real stress test. Revenue growth from AWS, Azure, and Google Cloud needs to show a credible acceleration toward the returns these buildouts imply. If it does, the investment thesis holds and the debt gets absorbed. If Q2 shows another round of capex increases alongside flat or declining free cash flow guidance, the June selloff will look like a preview rather than an overreaction. The number that matters most isn't the capex figure itself. It's whether the revenue curve is actually bending up to meet it. **Also read:** [New York becomes the first US state to ban new hyperscale data centers and other states are watching](https://startupfortune.com/new-york-becomes-the-first-us-state-to-ban-new-hyperscale-data-centers-and-other-states-are-watching/) • [Amazon brings Alexa Plus to UK browsers and the AI assistant wars get harder to ignore](https://startupfortune.com/amazon-brings-alexa-plus-to-uk-browsers-and-the-ai-assistant-wars-get-harder-to-ignore/) • [Genesis AI is in talks to raise $500 million as VCs bet on robotics software over hardware](https://startupfortune.com/genesis-ai-is-in-talks-to-raise-500-million-as-vcs-bet-on-robotics-software-over-hardware/)