General Compute Raises $100 Million, Defying Neocloud Pessimism

In a week that was anything but favorable for the "neocloud" sector — those companies that rent out compute capacity purpose-built for artificial intelligence — a startup called General Compute announced it had raised $100 million in a Series A funding round. The timing alone is worth pausing over: shares of publicly listed CoreWeave had just sunk to their lowest level in twelve months, and Elon Musk was talking about $40 billion in debt to purchase compute power, predicting a wave of attrition and consolidation that would wipe out more than 200 companies in the sector.

A Contrarian Read on the Market

Finn Buklovsky, the company's co-founder and CEO, doesn't seem worried. He believes talk of an impending collapse driven by a chip crisis is "greatly exaggerated," and that Musk's statements serve his own interest in driving down prices for the graphics processing units he wants to buy. Buklovsky points to the reality of his day-to-day work: a steady stream of weekly calls from parties requesting any available compute capacity — whether graphics processing units, tensor processing units, or any other chip. He adds that the optimistic financial filing Anthropic submitted for its public offering has helped ease short-term concerns.

Specializing in Inference Chips

What sets General Compute apart is that it never competed with everyone else for the same resources. While other companies race for Nvidia units, the company chose to build its cloud on chips specialized in inference — that is, running already-trained models rather than training them. It struck a $400 million deal to purchase new chips from SambaNova, described as the first debt financing for inference chips, and later added Positron and Cerebras to its list of suppliers.

Lean Infrastructure Instead of Massive Data Centers

The second approach that distinguishes the company is avoiding the costly, complex undertaking of building giant data centers from scratch. Instead, it has relied on three practical pillars:

• Tapping smaller power sources that can be secured more quickly.
• Leasing space inside data centers owned by other companies.
• Converting abandoned cryptocurrency mining sites into platforms for running AI models.

This choice means the company won't soon rise to the ranks of the largest cloud providers and won't compete with giants like Amazon and Microsoft. But in return, it isn't shackled by the constraints of building from the ground up: when it acquires chips, it already has somewhere ready to run them. Buklovsky expects to deploy around 75 megawatts of compute capacity during 2027, with access to as much as 200 megawatts of power — more than the company could finance or source hardware for.

One Contract That Opened the Funding Door

The primary driver of the round was a contract with a major customer generating $100 million in annual recurring revenue. The round was led by Blackbird Ventures and Abstract Ventures. The numbers here are striking: the company was valued at $60 million following a seed round in May, and is now valued at $800 million after this round — an increase of nearly thirteen-fold in just a few months. Even so, the founder acknowledges that the fundraising job isn't over, and that securing both debt and equity capital will remain "a full-time job."

The Company's Place in the Value Chain

General Compute's main customers are cloud companies specializing in ultra-fast inference services for developers. The company provides them with bare metal servers with no software layer on top, so they can build their own commercial offerings on it. James Palmer, one of the round's leads from Blackbird, believes demand for inference on open-weight models — and for companies to own their own intelligence — is growing rapidly, and that makers of alternative chips need distribution partners capable of handling the complexities of the physical world: power, real estate, and installation.

The Real Benchmark: Profitability, Not Speed

The equation the company is betting on is economic more than technical. As Rodrigo Liang, CEO of SambaNova, summed it up: everyone in this market is selling speed, but the real question is whether you can make money from it. In a sector where capital investment is accelerating at a pace that worries analysts, a focus on cost viability — rather than record-breaking figures — may be what determines who survives and who disappears in the coming wave of consolidation.

✦ بقلم فريق دروب أيديا