Google поднял прогноз AI-капзатрат до $205 млрд — Уолл-стрит занервничал

Google, in its quarterly report published on July 22, 2026, raised its full-year capital expenditure forecast to $205 billion — up from the $190 billion the company cited a quarter earlier. Even the lower bound of the new range, $195 billion, exceeds the previous ceiling, and the revision rattled investors on Wall Street.

What Changed in the Forecast

Google raised its upper estimate of 2026 capital expenditures from $190 billion to $205 billion, and the lower bound of the range to $195 billion. The key detail: the new spending minimum turned out higher than the previous maximum, which the company cited just a quarter ago. According to Reuters, Google revised the forecast at the same time as it published its quarterly report, in which revenue growth at its cloud division, Google Cloud, beat analyst expectations. In other words, the business is growing — but its appetite for spending is growing even faster.

• Quarterly report published July 22, 2026
• New capex forecast — up to $205 billion (previous — up to $190 billion)
• Lower bound of the new range — $195 billion, above the previous ceiling
• Size of the revision — in just one quarter — about $15 billion
• Main spending category — AI infrastructure: data centers, servers, chips

Why Did This Scare Investors?

Investors were troubled not by the size of the sum itself, but by the fact that Google effectively admitted it cannot precisely forecast its own spending. A $15 billion revision in a single quarter shows that the company does not control the pace of growth in its AI infrastructure spending. For the market, unpredictability of spending is scarier than its absolute size. On top of that, according to the Financial Times, Google is spending more on infrastructure than it earns — which calls into question the economics of its entire bet on AI.

"From an investor's point of view,

Google has essentially admitted it cannot accurately forecast its own spending" — The Verge column

How the Market Reacted

The market's reaction is already reflected in The Verge's headline: "AI finally got expensive enough to rattle Wall Street." The news came out in the middle of earnings season, when investors are especially attentive to spending forecasts, and AI-sector stocks responded with a decline. The temptation to wave it off — "what's $15 billion among friends" — is understandable, but it's precisely the inability of the largest internet company to name a precise figure for its own spending that hits confidence harder than any single sum.

Where Is This Money Going?

Google's capital expenditures go almost entirely toward AI infrastructure: data centers, servers, and proprietary chips for training and running models. The $205 billion figure is spending for 2026 alone. This scale reflects the broader race among tech giants, who are scaling up computing capacity to meet demand for generative AI — and doing so faster than these investments manage to pay for themselves through revenue.

What This Means

The race for AI infrastructure has crossed a threshold beyond which its scale has begun to worry Wall Street: spending is growing faster than companies can forecast it. For Google, the second quarter of 2026 became the first period in which AI spending turned from a symbol of ambition into a source of investor concern. Going forward, the market will focus less on the announced sums themselves and more on how accurately companies are able to forecast them.

Frequently Asked Questions

How much did Google's capex forecast increase?

Google raised its forecast from "up to $190 billion" a quarter earlier to "up to $205 billion" in its report for the second quarter of 2026. The size of the revision is about $15 billion, and the lower bound of the new range ($195 billion) is above the previous maximum.

Why did the rise in capex worry investors?

A sharp revision of the forecast within a single quarter means Google cannot accurately forecast its AI infrastructure spending. On top of that, according to the Financial Times, the company is spending more on infrastructure than it earns.