General Compute will begin offering Cerebras wafer-scale AI hardware on its cloud platform in Q1 2027, under a multi-year agreement the startup calls its "largest single hardware commitment" to date.

The company declined to disclose the deal's value or the scale of deployment. General Compute will finance the systems itself and sell capacity as a managed inference service. The deal is the startup's first major hardware commitment since it raised $400 million in debt financing in July.

Why is General Compute buying the hardware?

CEO Finn Puklowski framed the arrangement as a balance-sheet play for customers. He said: "The chips that win inference are not going to come from one vendor, and most customers cannot put a wafer-scale system on their own balance sheet. That is the gap we exist to close."

"We buy the hardware, and our customers get Cerebras speed on a contract they can actually sign," Puklowski said. "Agentic coding is where that speed is worth the most right now, so that is where we are starting."

Cerebras CTO and co-founder Sean Lie emphasized latency in agent workloads. He said: "In AI, speed is productivity. An agent that takes hundreds of steps to finish a task is only as fast as its slowest step. Working with General Compute puts Cerebras speed in front of the developers building these agents, on a platform they already trust."

How does Cerebras fit the existing stack?

General Compute already runs a multi-vendor inference platform. Its portfolio includes:

• SambaNova GN50 chips, used for decoding calculations
• AMD MI300X GPUs, handling the remaining workload phases for inference
• Nvidia GPUs, which the company will use for prefill

According to a LinkedIn post from Puklowski, pairing Nvidia GPUs with Cerebras for prefill "gives a major step up in reducing the cost of delivering inference — meaning more intelligence per dollar."

What capacity backs the deployment?

Puklowski and CTO Jason Goodison founded the company, which raised $15 million in seed funding in May 2026. A whitepaper shared at that time stated the company holds "options on 15 megawatts of air-cooled power, which is enough to support the Q4 2026 architecture and the growth beyond it, at colocation facilities that exist today."

Second Cerebras win this week

The agreement marks Cerebras' second customer deal announced this week. AI cloud startup Gimlet Cloud said it will deploy 100MW of Cerebras wafer-scale compute for inference workloads on its platform. The first data center under the Gimlet Cloud agreement is expected to come online later this year.

The two deals signal growing demand for wafer-scale inference capacity among cloud providers willing to own the hardware on behalf of customers who cannot or will not finance it themselves.