Alphabet hikes 2026 AI capex to $195-$205B, despite admitting supply limits and margin pressure

Alphabet’s Q2 earnings came with the kind of headline figure finance teams can’t ignore: $119.8 billion in revenue, up 24% from a year earlier. But the real message showed up right after the numbers, in how Google talked about what comes next. Executives used the earnings call to make one thing painfully clear: AI growth is no longer just driving revenue, it is also driving bigger infrastructure bets, and those bets will cost money in the near term.

The biggest financial tell: Google raised its 2026 capital expenditure forecast to between $195 billion and $205 billion, from a previous outlook of up to $190 billion. That higher spending is tied to the company’s view that demand for AI infrastructure continues to exceed available capacity. Finance chief Anat Ashkenazi said the higher capex reflects an accelerated rollout of computing capacity. Translation for decision-makers: Google is not waiting for the market to loosen. It is trying to buy down the supply squeeze by building and expanding faster, even if that means sacrificing margins while it catches up.

This is the core tradeoff Google is making. The company prioritized long-term AI growth over short-term profitability by doubling down on its AI buildout. Ashkenazi also described a strategy of leaning more heavily on third-party cloud providers while Google builds out its own infrastructure. That mix is designed to keep pace with customer demand, but it comes with a warning label: Ashkenazi said the approach will create “modest margin pressure in the near term.” The logic is straightforward: spending more now to keep customers close and keep the growth engine running. The company expects the partnership-heavy path to help it “keep growing our customer base and capture greater overall value.”

If that sounds like pure growth at all costs, Alphabet gave an additional data point on where the growth is actually landing: Google Cloud is becoming the AI story’s primary engine. Across Alphabet’s businesses in Q2, Google Cloud delivered the strongest performance. Cloud revenue jumped 82% year over year to $24.8 billion, well ahead of analyst expectations, and cloud backlog reached $514 billion. Backlog matters because it is a forward-looking signal of demand, especially in a world where enterprises are building AI applications and need capacity commitments.

Ashkenazi reinforced the supply-constrained theme by saying Google is still in a supply-constrained environment. He added that Google is seeing “very strong demand both from external cloud customers as well as across the business.” That helps connect the capex increase to real-world customer pressure, not just corporate ambition. The company also gave a glimpse into how it is prioritizing its infrastructure: about 60% of Google’s infrastructure spending during the quarter went toward AI servers, with the rest invested in data centers and networking equipment. For operators and board members, this is a resource allocation statement as much as an earnings update, because it shows where engineering and spend are being concentrated to turn demand into delivered compute.

Then there is the consumer side of the equation, where Google is trying to translate infrastructure scale into product pull. The earnings call framed Gemini as closing in on ChatGPT, with a specific milestone: the Gemini app now has 950 million monthly active users, up from about 650 million last October and more than 750 million earlier this year. CEO Sundar Pichai also said daily active users have tripled over the past year. Those numbers are not just marketing. They are evidence that Google’s AI assistant strategy is accelerating at a time when competition with OpenAI and Anthropic is intensifying.

Google also signaled it is working on the economics of running AI, not only the adoption curve. This week, it introduced three new Gemini models, including Gemini 3.6 Flash. Google says the model improves coding performance while using fewer tokens, which would reduce the cost of deploying AI applications. That matters to anyone watching the industry because “AI demand” is only half the story. The other half is whether the unit cost of serving models comes down enough to make growth scalable. Gemini’s usage milestone plus a push toward cheaper deployment is Google’s attempt to keep both its top-line momentum and its infrastructure spend on a sustainable trajectory.

Put it all together and the strategic stakes become obvious. Alphabet is operating on a belief that AI compute demand will not politely wait for the supply chain to catch up. It is raising 2026 capex to $195 billion to $205 billion, using a combination of owned capacity and third-party cloud, accepting “modest margin pressure” in the near term, and pushing Gemini toward the consumer scale needed to reinforce the ecosystem. For peer companies and boards across tech and cloud, the lesson is not just “Google spent more.” It is that the winners in AI are likely the ones that treat capacity as a competitive product, then back it with distribution, both for developers through Cloud and for users through Gemini.