Google Cloud hits $20B, but the real story is what’s holding it back

Google Cloud just crossed the $20 billion mark in quarterly revenue for the first time. That’s a big number, and it’s driven almost entirely by AI workloads. But reading between the lines of the earnings report, the more interesting story is what didn’t happen: they could have grown even faster if they had the infrastructure to support it.

The company explicitly called out capacity constraints as a limiting factor. That’s not unusual for a hyperscaler in the middle of an AI boom, but it’s worth unpacking. When a cloud provider says they’re “capacity-constrained,” they’re basically saying demand is outstripping their ability to build out data centers and GPU clusters fast enough. The result? Customers get put on waitlists, or they go elsewhere.

I’ve seen this play out before. During the early days of cloud computing, AWS and Azure both hit similar walls. The difference now is the scale. AI training and inference require massive, specialized hardware—think Nvidia H100s and custom TPUs—that can’t be spun up overnight. Google’s been investing aggressively in custom chips like the TPU v5, but even that isn’t enough to keep pace with the frenzy.

The $20B figure itself is impressive, no doubt. It’s a 35% year-over-year increase, and it puts Google Cloud firmly in third place behind AWS and Azure, but narrowing the gap. What I find telling is that the growth rate would have been even higher if not for those constraints. In other words, the demand is real, and it’s not just hype.

But here’s the catch: capacity constraints are a double-edged sword. On one hand, they signal strong demand, which is good for long-term revenue. On the other hand, they create an opportunity for competitors. If a startup can’t get GPU time on Google Cloud, they’ll spin up on AWS or Azure instead. And once they’re in, switching costs are high.

Google’s response has been to ramp up capital expenditure. They’re building new data centers in places like Ohio, Singapore, and Germany. They’re also doubling down on software optimizations to squeeze more out of existing hardware. But infrastructure takes time. A data center doesn’t go from planning to production in a quarter.

I also wonder about the pricing dynamics. When capacity is tight, prices tend to rise. Google Cloud has been relatively disciplined about not gouging customers, but I’ve heard from some startups that spot pricing for GPUs has become unpredictable. That’s a pain point that larger enterprises can absorb, but smaller players feel acutely.

Looking ahead, I expect Google to keep investing heavily. The AI arms race is real, and no one wants to be the cloud that runs out of compute. But the real test will be whether they can scale fast enough without sacrificing reliability or pricing sanity. Crossing $20B is a milestone, but the capacity story tells me the next few quarters will be more about execution than celebration.

For now, Google Cloud is in a good spot—growing fast, with clear demand. But the constraint narrative is a reminder that even the biggest players can’t always keep up. And that’s not a bad thing for the industry; it keeps everyone on their toes.