General Compute, an AI inference cloud startup, has secured a $400 million loan from Upper90, a tech investment firm. This deal marks the first known instance of using inference-specific chips as collateral, representing a shift in focus away from more expensive model-building GPUs towards cheaper and more efficient inference hardware.

The company's SN50 chips are designed for rapid and power-efficient inference tasks, offering 16 times faster performance compared to GPU-based clouds. Founded by CEO Finn Puklowski and CTO Jason Goodison, General Compute has secured significant backing from SambaNova Systems' silicon, aiming to build a neocloud specifically tailored for AI workloads.

Upper90’s Billy Libby believes that as markets respond to concerns over the cost of AI tools, there is growing interest in infrastructure that can run open-source models more economically. This approach could democratize access to advanced AI functionalities beyond the reach of traditional hyperscalers like AWS or Azure. With new chipmakers like Groq and Cerebras entering the market, there are increasing alternatives to Nvidia’s dominance.

‘By getting together with Upper90,’ says Puklowski, ‘this is not just a cool startup raising money; it’s the first signal of capital organizing itself against Nvidia’s monopolistic grip.’