General Compute Bets on Non-Nvidia Silicon to Scale Inference Cloud

General Compute has raised $100 million in Series A funding led by Abstract Ventures and Blackbird Ventures, as TechCrunch first reported. The startup, which reached an $800 million post-money valuation, is positioning itself as a distribution partner for chipmakers competing with Nvidia.

While many neoclouds have focused on GPUs, General Compute is specializing in inference silicon. The company has entered a $400 million deal with SambaNova—the first debt financing of inference chips—and has added Positron and Cerebras to its supplier portfolio. To avoid the overhead of building massive data centers, the company utilizes smaller power sources, rents space in existing facilities, and repurposes former bitcoin-mining sites.

CEO Finn Puklowski told TechCrunch that demand for compute remains high, dismissing fears of a chip-led collapse as "massively overstated." By 2027, Puklowski intends to deploy 75 megawatts of computing capacity and estimates he can access 200 megawatts of power. The company provides "bare metal" raw servers to cloud companies that offer high-speed inference to developers.

This strategy comes as the neocloud sector faces headwinds. TechCrunch reports that CoreWeave's stock has hit a 12-month low, and Elon Musk has predicted consolidation and "guppy-level death" for over 200 neoclouds. However, Puklowski suggests Musk is attempting to drive down GPU prices.

James Palmer, a principal at Blackbird, noted that firms like Cerebras, Positron, and SambaNova require partners to navigate physical scaling. SambaNova CEO and co-founder Rodrigo Liang told TechCrunch that while the market focuses on speed, the critical factor is whether providers can make the economics of AI cloud compute profitable.