Google's search revenue is slowing down. AI might be why.

Key points

• Alphabet's search and other segment grew just 12% year-on-year in Q3 2026, down from 19% in Q1 and 17% in Q2, missing analyst estimates for the second straight quarter.
• Google Cloud revenue more than doubled to nearly $31 billion in Q3 2026, with a contracted backlog of approximately $560 billion.
• Alphabet maintained its full-year 2026 capital expenditure forecast of $195 billion to $205 billion, with spending set to rise further in 2027.
• Negative free cash flow hit $9 billion in Q3 2026, three times the estimated deficit and the second consecutive quarter of negative free cash flow in the company's public history.
• Gemini AI models helped advertisers generate more than 1.5 billion campaign assets in Q3 2026, up more than sevenfold from a year earlier.

The number that should worry Google is pretty simple: search revenue growth has fallen in a straight line for three quarters running. Nineteen percent in Q1, then 17%, now 12%. Alphabet missed on the top and bottom lines this quarter, and the stock fell roughly 10% in after-hours trading.

Here's the uncomfortable part. Google says search query growth, meaning the number of times people actually typed something into Google, was the strongest in five years. People are still searching. They're just clicking on fewer ads, or the ads they click on are worth less. A gap between how much a product is used and how much money it makes is a warning sign for any business.

D.A. Davidson analyst Gil Luria put it plainly: the AI fears that circulated from 2023 to 2025 "appear to finally be showing up in Google's business." Alphabet says foreign exchange headwinds drove most of the shortfall, and those will get worse in Q4. Fair point, but it doesn't explain everything.

Is AI eating Google's search business?

Not definitively yet, but the direction is uncomfortable. AI assistants and agents, software that can answer questions and complete tasks on your behalf rather than just returning a list of links, have given people a real alternative to typing queries into a search box. Google makes Gemini and uses it inside search results. The problem is that when an AI answers your question directly, you may never click the ad sitting next to the old-style results.

UBS analysts wrote after the results that the numbers would "bring Search cannibalization fears right back to the forefront." Cannibalization means one product from the same company eating the revenue of another, which is roughly what happens when Gemini answers a question that would previously have sent a user to an ad-laden results page. Mizuho analyst Jordan Klein went further, writing that the numbers "elevate the fear around AI and agentic taking share from Search."

For ordinary users this internal tension is mostly invisible. Google search still works. Ads still appear. Our story on Google paying more than 400 news publishers for AI features, published two days before these results, captured the same underlying question: when AI answers the query, does anyone visit the original source? The earnings figures suggest that problem isn't hypothetical.

Where is Alphabet actually winning?

Cloud. Google Cloud more than doubled revenue to nearly $31 billion in Q3 2026, and its contracted backlog reached around $560 billion. Management said just over half of that backlog should be recognised within 24 months. That's an extraordinary pipeline.

Alphabet CFO Anat Ashkenazi explained the cash crunch directly on the earnings call. Google has to build the systems before customers pay for them. Cash goes out first; revenue comes in later. The $195 billion to $205 billion capital expenditure forecast for 2026 is the cost of staying in that race.

Part of that spending goes toward Google's own TPUs, or tensor processing units, custom chips designed to run AI workloads that compete with Nvidia's GPUs (graphics processing units, the standard chips the AI industry has relied on). Google released its seventh-generation TPU, called Ironwood, in November 2025 and began selling it to outside customers in Q2 2026. We reported this week on how those chips are now leaving Google's own data centres, reaching Oracle's infrastructure with OpenAI already keen to follow.

The honest read here is that Alphabet is running two businesses at once: a giant legacy ads machine that's showing age, and an AI infrastructure play that's genuinely growing fast. Investors bid the stock down because they're not sure the second will replace the first quickly enough. One quarter doesn't decide that question, but it's no longer a theoretical one.

Common questions

Does this affect regular Google users right now?

Not directly. Google search still works the same way it always has. The changes are financial, meaning Google earns less per search than it used to, not a change in what you see on screen.

What is a TPU, and why does it matter here?

A TPU, or tensor processing unit, is a custom chip Google designed specifically to run AI calculations. It's Google's alternative to the Nvidia GPUs most AI companies buy. Selling them to outside businesses turns that chip investment into a new revenue line.

Should I worry about my Google account or data?

Nothing in this quarter's results changes how Google handles your data. The privacy considerations that already existed, such as Google using your search history to target ads, remain unchanged.