☁️Quake AI Builds Cloud for AI Agents on 22,000 Nvidia GPUs
Your next API consumer might not have a face
TL;DR
Quake AI, an Nvidia Cloud Partner running 22,000 GPUs, is building infrastructure for AI agents instead of people. Its bet: agents want queryable APIs and utility-style compute, not browsers. If agents already outnumber human visits on the web, this design choice matters.
Quake AI is betting that its main customers won't be people. The Nvidia Cloud Partner says its users will be mostly AI agents, and it's tailoring its infrastructure and APIs to fit them. Its CFO put it bluntly: agents don't use vision to interact with systems. If you're building agent workflows, this is the pitch to pay attention to. The CEO's argument is that making an AI act like a human to click through a site burns insane amounts of compute. A queryable API, he says, would use about a thousandth of that. Quake wants agents to consume compute as a utility, on homogeneous hardware where performance and uptime expectations are predictable. The numbers behind the bet: TollBit reported roughly one AI bot visit for every 31 human visits to a site by the end of 2025. Quake runs 22,000 GPUs as an Nvidia Cloud Partner. Gartner, meanwhile, forecasts that AI services won't generate enough revenue to cover planned IT infrastructure capex through 2030, and it expects provider consolidation. What to watch: whether Quake's API-first model wins real agent traffic, or whether it becomes one more specialist provider that gets absorbed.

Key Points
Quake AI operates an estate of 22,000 GPUs as an Nvidia Cloud Partner, giving it serious capacity for a provider still focused on specialized AI workloads.
Keep reading — it's free
Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.
Already a member? Sign in
Comments
Be the first to comment