Alphabet Posts Negative Cash Flow for the First Time Since 2004 as Market Begins to Assess AI Costs
Alphabet's stock dropped over 7% after the company raised its 2026 capex guidance to $205 billion and reported negative free cash flow for the first time since its IPO. The SOX index fell 17% in July, the worst month since 2022.
Alphabet reported negative free cash flow of $5.9 billion in the second quarter of 2026, the first negative reading since its IPO in 2004. The quarter's capex doubled from the previous year to $44.9 billion, a figure that alone surpassed the company's $39.1 billion operational cash generation for the period. The full-year capex guidance was raised to up to $205 billion, compared to the previous range of $180 to $190 billion.
Alphabet's stock fell 7% on Thursday following the release, marking its worst day in over a year. The contagion spread. According to Bloomberg, in a report published this Sunday, investors are 'no longer giving a free pass to spiraling expenses' and have begun to demand monetization or at least visibility on monetization.
The Philadelphia Stock Exchange Semiconductor Index has fallen 17% in July, heading for its worst month since June 2022. On the same day that Wall Street repriced the sector, the Nikkei 225 dropped by 1,811 yen, with Tokyo Electron alone accounting for about 448 yen of that decline.
The Size of the Commitment
The combined capex of Alphabet, Microsoft, Amazon, and Meta reaches $724 billion in 2026, with projections of $950 billion in 2027. This is almost double what the four companies spent in 2024. Alphabet maintains $240 billion in cash and investments, and its free cash flow over the past twelve months is still positive at around $53 billion. The company is not bankrupt. What has changed is the perception of return architecture.
Alphabet's cloud revenue grew 82% in the quarter to $24.8 billion, and operating income increased by 30% to $40.8 billion. This is the official argument from management to justify the $205 billion: demand for AI cloud services is growing faster than installed capacity, and every dollar of capex today generates multiples of revenue tomorrow. Sundar Pichai reiterated this narrative during the earnings call, and CFO Anat Ashkenazi presented a utilization ramp model in which each new gigawatt enters contracted capacity before becoming energized.
The problem is that this architecture only holds if demand continues to grow at the pace of 2024 and 2025. If a single quarter of slowdown occurs, the ratio between capex and free cash flow turns into an accounting wound.
The European Counterpoint
In the same period that Wall Street pressed Alphabet, SAP announced that its current cloud backlog grew by 27% year-over-year to €22.9 billion at the close of the second quarter. Cloud revenue rose by 22% to €6.3 billion. This is the European company that has most clearly managed to transform investment in AI into signed contracts.
SAP does not spend what Alphabet spends. But it spent enough to transform Joule, its enterprise copilot, into a contractable product, and the results are evident in the backlog and not in current revenue, the same signal investors sought but could not find in Alphabet's release. SAP's operating income guidance was adjusted to between €11.8 billion and €12.2 billion due to recent acquisitions of Dremio and Prior Labs, a strategic choice the market interpreted as a tolerable investment.
The difference between the two narratives illustrates Wall Street's new demand: capex is acceptable when the contractual ramp supports the bill.
What This Means for Corporate Clients
An investor revising the capex thesis also exposes the corporate client. A company that signed a three-year contract with Google Cloud in 2024 is questioning whether the next price increase will come in 2027 to cover the amortization curve of the data center. A CIO at a bank in São Paulo, Tokyo, or Frankfurt is now considering a multi-cloud architecture not just as operational redundancy but as a shield against the risk of cost pass-through.
The window in which hyperscalers subsidized capex with their own balance sheets to win workload has shrunk. On the supplier's side, the new mantra is 'demonstrate a path to return'. On the client’s side, the new mantra is 'do not lock in more than 40% of the workload with a single provider'.
The coming days will bring results from Microsoft (Wednesday), Meta (Wednesday), Apple, and Amazon (Thursday). If any of them repeat Alphabet's exercise, raising capex without presenting proportional contractual backlog, the 17% drop in the SOX will seem like a warm-up.