Alphabet CEO Sundar Pichai just gave the clearest hint yet that Google is rationing its TPU supply with a very specific mission in mind: AGI. On Alphabet’s second quarter earnings call, Pichai said the company’s “first priority is making sure we are allocating what we need to compete at the frontier in terms of AGI development.” In other words, TPUs are not just another rentable compute product. They are a strategic input, managed to protect the work Google believes underpins everything it does.

That priority matters because it directly answers the tension investors and analysts raised. When Goldman Sachs analyst Eric Sheridan asked how Alphabet balances TPU demand from customers against its own compute needs, Pichai framed TPU allocation as baseline-driven by what it takes to keep AGI development moving “at the frontier.” Then Mark Shmulik of Bernstein asked about how compute gets divvied up across Search, the cloud business, and model training. Pichai said the baseline starts with the same AGI development needs, then Alphabet also “prioritizing our core product areas like Search, YouTube, et cetera, as well as Cloud.” So the allocation logic is not “sell everything, then hope the models get enough compute.” It is “ensure the frontier work has first claim, then allocate the rest to keep the businesses running strong.”

If you are an operator or board member, this is where the second-order implications kick in. When compute is scarce or expensive, “customer demand” is not a simple growth lever anymore. It becomes a tradeoff between near-term revenue and the long-term performance curve of AI models, which then affects Search, ads, cloud adoption, and your credibility as an AI infrastructure provider.

Alphabet is also publicly wrestling with that constraint. CFO Anat Ashkenazi said the company now plans to spend between $195 billion and $205 billion for the financial year, up from a previous estimate of $180 billion to $190 billion. The reason is brutally practical: Google “can’t get all the kit it needs” in a “supply-constrained environment.” That is the setup for the next move Pichai described. Alphabet plans to “expand the use of third-party capacity in Q3 as a bridging strategy while we build out more internal capacity.”

This is not a minor procurement tweak. It changes how Google delivers customer commitments during a period when demand for AI infrastructure is rising fast. Pichai said buying bridging capacity will help Google support “very, very large customers of ours on Cloud” through “this extraordinary moment.” He also argued the economics: while the short-term cost over a few months may be very high, in the lifetime of multi-year cloud deals it can be “highly ROI positive.” That is the corporate translation of a common engineering reality: if the compute supply chain is tight, you either delay service, risk losing momentum with enterprise clients, or pay for external capacity to keep the pipeline and performance intact until internal capacity catches up.

The cloud business is clearly pulling its weight in the numbers. Google Cloud revenue jumped 82 percent year over year to $24.75 billion in the quarter, and it delivered an $8.8 billion profit representing 214 percent growth. Alphabet said the growth came from “strong demand for AI infrastructure and AI solutions.” Search revenue grew 17 percent and YouTube ads grew 13 percent. And in a sign that demand is not just speculative, Google has $514 billion of cloudy backlog on the books, meaning customers have signed up for services they have not yet consumed.

Meanwhile, the fear campaign around generative AI did not materialize in the simple form some pundits predicted. When generative AI arrived, some suggested it could threaten Google’s search ads business. Pichai pushed back directly on that concern by saying Google’s AI Mode for search is “driving an incremental increase in Search queries overall.” He also pointed to hardware and cost efficiency: thanks to engineering and hardware optimizations, Alphabet reduced the cost of AI Mode responses to the lowest level since launch, even as it brought more advanced AI capabilities. Again, the strategy link is obvious. Lower per-query inference cost helps scale AI features without sacrificing the economics that keep ads, cloud usage, and overall margins stable.

And yet the financial reality underneath all of this is complicated. Alphabet’s quarterly revenue landed at $119.8 billion, up 24 percent year over year, and operating income hit $40.8 billion, up 34 percent. But free cash flow landed at -$5.9 billion, the first time Google has not had spare cash to splash since 2004. Investors reacted accordingly, sending the company’s shares down by four percent in after-hours trading.

So here is the crux for executives at competitors and partners: Alphabet’s posture implies a two-track strategy. Track one is frontier model development, with TPU allocation prioritized for AGI competitiveness. Track two is scaling customer-facing AI deployments and cloud throughput using a bridging strategy of third-party compute, justified as ROI positive across multi-year deals. When you combine that with massive planned capex ($195B to $205B) and a supply-constrained environment, you get a clear message: the winners in AI infrastructure are not just buying chips. They are building a compute allocation and delivery model that can survive scarcity, protect model progress, and still keep enterprise customers from looking elsewhere.