Big Tech's AI build-out now outruns its cash, and startups will feel the change

For most of the past decade the largest technology companies paid for their ambitions out of pocket. The third-quarter results published between October 21st and October 29th show that era ending. Alphabet reported negative free cash flow of $9.1bn for the quarter. Amazon's free cash flow over the trailing twelve months swung to minus $42.4bn, from plus $14.8bn a year earlier. Meta's capital spending of $34.9bn exceeded its operating cash flow, leaving free cash flow at minus $2.1bn. Microsoft was the exception that pleased investors, beating forecasts as Azure grew by 48%.

Markets were unforgiving. Alphabet's shares fell almost 13% on October 29th, which Fortune described as the worst trading day in the company's history, erasing about $500bn of market value. Amazon fell more than 6% in extended trading after its own report. The headline figures matter less than what lies beneath them: the hyperscalers are now financing the AI build-out with borrowed money and newly issued stock, and that changes the terms on which startups get compute, capital and exits.

From self-funding to capital markets

Alphabet's own release is the clearest illustration. Long-term debt rose from $46.5bn at the end of 2025 to $125.1bn at the end of September. The company raised $29.0bn from debt in the third quarter alone and $85.2bn so far this year. It also took in $12.4bn from issuing common stock during the quarter, including $4.3bn from selling 12m Class A shares under an at-the-market programme that still has $27.6bn available. It bought back no shares, against $11.5bn in the same quarter a year ago. Capital expenditure reached $51.4bn, more than double the $24bn of a year earlier, and the company kept its 2026 range of $195bn to $205bn while saying spending would "increase significantly in 2027".

Meta has moved in the same direction. According to XTB, its long-term debt climbed from $58.7bn at the start of the year to $80.9bn after $24.9bn of bond issuance, and it has repurchased no stock in 2026. It also announced a new non-voting Class C share, to be distributed one for one to existing holders from January 19th 2027, with capital raising, employee pay and acquisitions given as purposes. Meta narrowed its 2026 capital spending range to $135bn to $145bn. Amazon spent $77.3bn on capital projects in the quarter, well above the $58.7bn analysts expected, held its 2026 plan at about $220bn and said it would spend "meaningfully more in 2027".

Microsoft stands apart. It trimmed its calendar 2026 capital spending forecast to $170bn from $175bn, but not for lack of appetite. Its chief financial officer, Amy Hood, said deliveries from third-party GPU providers had slipped, pushing some capacity into the first half of 2027, and that spending would accelerate then. "We still have demand that exceeds supply," she said.

Private stakes, public volatility

A second feature of the quarter deserves attention from venture investors. The hyperscalers now hold private stakes large enough to swing their reported earnings. Alphabet recorded an $8.7bn loss on equity securities, which cut diluted earnings per share by $0.55. CNBC noted that the same line had produced a gain of about $99bn in the second quarter, and that Alphabet's holdings include stakes in SpaceX and Anthropic. Amazon's net income fell year on year partly because the comparable quarter of 2025 included a $9.5bn pre-tax gain on its Anthropic investment.

The consequence is that marks on a handful of private companies are now flowing through the income statements of the world's most valuable listed firms. Every new funding round for a large AI laboratory becomes, in effect, a public earnings event. That creates pressure in both directions: corporate investors have an interest in rounds that hold or raise valuations, while public shareholders, having watched an $8.7bn swing in a single quarter, are likely to ask harder questions about how those stakes are valued and how much more capital is committed to them.

Compute remains scarce, and the neoclouds carry the risk

For startups that buy compute, the message from the calls was that scarcity persists. Microsoft's delays came from outside suppliers; CNBC reported that the company relies on providers including CoreWeave, IREN and Nebius, and that shares in those firms and in Oracle fell after the reports. Alphabet's finance chief, Anat Ashkenazi, said outside capacity "came online only at the very end of Q3". Amazon said its Trainium chips would become available externally only in late 2027.

The independent GPU clouds that grew up to fill this gap now look exposed. CNBC cited a Gartner forecast that by 2030, 80% of neocloud companies will be absorbed or cease operations. Firmus, an Australian provider backed by Nvidia, withdrew its listing plans in early October, and Nscale, which filed to go public in September, had not updated its prospectus for weeks. When hyperscaler customers delay or reshape orders, the smaller providers absorb the shock first, and their own venture backers with them.

What the cloud numbers say about demand

None of this signals weak demand. Google Cloud revenue rose 103% to $30.7bn with an operating margin of 36.2%, and its chief executive, Thomas Kurian, spoke of strong growth from AI laboratories. AWS grew by 39% to $46.0bn, and Andy Jassy said he expects growth above 40% in both 2027 and 2028. Microsoft's Azure grew by 48%. The weakness lay elsewhere: Google Search grew 12%, slower than the 15% analysts expected, and Amazon's advertising and online sales also came in below forecasts.

Much of the cloud growth comes from the same few model developers that the hyperscalers have funded. That circularity is not new, but negative free cash flow makes it more visible. The hyperscalers are borrowing to build capacity that is sold, in significant part, to companies in which they hold equity.

Our view

The quarter marks a structural shift: Big Tech has become a heavy user of the capital markets rather than a provider of surplus cash. For founders this has three implications. First, compute will stay tight into 2027, so long-term capacity contracts and multi-cloud arrangements are a matter of survival, not procurement tidiness. Second, the hyperscalers' generosity as strategic investors may become more conditional. Shareholders who absorbed a $9.1bn cash burn at Alphabet will scrutinise every private stake that moves earnings, and terms tied to cloud spending are likely to tighten. Third, exits are being re-engineered. Meta's non-voting Class C shares give it a fresh acquisition currency that does not dilute control, which could support stock-funded acquisitions of startups even as cash grows scarce.

Venture investors with exposure to neoclouds and infrastructure intermediaries face the sharpest risk. Watch fourth-quarter guidance for 2027 capital spending, the pace of Alphabet's share sales, and whether listing candidates such as Nscale revive their plans.

Sources

Filed under Big Tech. Written by InvestTimes Editorial under our editorial standards. Commentary and analysis for information only, not investment advice.

From self-funding to capital markets

Private stakes, public volatility

Compute remains scarce, and the neoclouds carry the risk

What the cloud numbers say about demand

Our view

Sources

Further reading

Datahorses moves the first line of cyber defence outside the company network

Kalshi's $40bn round asks investors to price growth and a legal fight at the same time

OPay's aborted listing and two hawkish pauses show how narrow the exit window has become