Summary: Alphabet’s April 29, 2026 report showed Google Cloud crossing $20 billion in quarterly revenue, up 63% year over year, while Sundar Pichai said the business was compute constrained in the near term. Cloud backlog doubled to $462 billion, and Google systems processed about 16 billion tokens per minute versus 10 billion in the prior quarter. The article reads the constraint as supply-side, not cooling demand.
Alphabet’s first-quarter report, out on April 29, 2026, contained a milestone: Google Cloud cleared $20 billion in quarterly revenue for the first time, growing 63% year over year. Yet the sentence that mattered most came from CEO Sundar Pichai on the earnings call. “Obviously, we are compute constrained in the near-term,” he said, adding that cloud revenue “would have been higher” had Google been able to take on the demand in front of it.
Two sentences capture the quarter: the constraint is supply, not demand. When an operator of this scale states publicly that paying customers are queuing for compute it cannot secure, the shortage stops being an industry rumor and enters the financial filings. Every team running AI products on cloud infrastructure should treat that as the most direct capacity signal available.
Reading the Growth: Demand Keeps Outpacing Supply
The topline rate conceals the internal mix. Google Cloud Platform expanded faster than the Cloud division as a whole — a division that also folds in data analytics, AI/ML tooling, and Google Workspace. Products built on Google’s generative AI models posted revenue growth approaching 800% year over year, while Gemini Enterprise rose 40% quarter over quarter.
The usage figures say it even more plainly. Google’s systems now handle about 16 billion tokens per minute, against 10 billion in the fourth quarter — a 60% rise in a single quarter. New customer counts doubled year over year, and existing customers consumed 45% beyond their initial commitments quarter over quarter, which means contracts are signed on estimates that systematically undershoot real consumption. Deal values are climbing alongside everything else: the count of contracts between $100 million and $1 billion doubled year over year, several of them above the billion-dollar mark.
A $462 Billion Backlog With Two Faces
Cloud backlog doubled within one quarter to $462 billion, and Google expects roughly half of it to convert over the following 24 months. Pichai presented the number as proof of differentiation — customers are committing at a scale very few operators could physically absorb. Analysts on the call, however, kept pressing the other face of it: how will constrained capacity be allocated, and when does the constraint ease?
A backlog is at once a reservoir of future revenue and a waiting line: behind every committed dollar stands an enterprise buyer who wants more capacity today and is being told to wait. When every cloud provider is chasing the same chips, power, and data centers, allocation — which customer receives which megawatts and which silicon — outweighs list pricing as a commercial decision. It also explains a pattern now standard across the industry: hyperscalers signing multi-year commitments with chipmakers and data center operators, so supply is spoken for at the negotiating table years before it is energized.
Why Capital Alone Cannot Summon Compute
Pichai built his explanation on two arguments. One, part of the demand is for the infrastructure itself: appetite for TPU hardware and data centers expanded along with everything else, and Google now ships TPUs directly to some customers instead of offering them only through its cloud. Two, capital decisions are governed by return on invested capital — a discipline that keeps funding what Pichai called the cutting edge while making sure no single customer’s commitments dictate the build plan.
Beneath both points sits lead time. A data center takes years to move from site selection through power procurement, interconnection queues, construction, and commissioning, and chip supply is booked out on much the same clock. Demand can double in a quarter; supply cannot follow. That asymmetry is precisely what a capacity-constrained quarter measures — and it is why Pichai could attribute the gap to Google’s building pace rather than to any operational failure: the company simply could not construct fast enough.
The Takeaways for Developers and Product Teams
Three actions follow. First, plan capacity before you need it: if your product runs on any first-tier cloud, never assume peak capacity can be bought on demand — negotiated commitments and reserved quotas belong in pre-launch planning, not post-incident repair. Second, resist reading constrained supply as cooling demand. The shortage sits on the supply side; token volumes, enterprise contracts, and backlog are all still accelerating, so teams that can secure capacity face improving unit economics for their AI products. Third, treat multi-cloud and multi-silicon as capacity insurance rather than architectural taste — a single provider’s allocation queue is a single point of failure for your service.