General Compute, an AI inference cloud startup, secured a $400 million loan from Upper90, a technology investment firm. It could be the first deal to offer inference-specific chips as collateral: chips built to run already trained AI models quickly and efficiently, rather than the more expensive chips used to build the models in the first place.

The funding is the latest sign that markets are responding to concerns about the pricing of AI tools and tokens by turning to infrastructure that runs open source models at a lower price than the newest LLMs from frontier labs.

Founded by CEO Finn Puklowski, General Compute raised a $15 million seed round in May to build an inference neocloud around silicon from SambaNova, a chipmaker backed by Intel. (Neoclouds are designed specifically for AI workloads, unlike the general-purpose infrastructure offered by traditional hyperscalers like AWS or Azure.)

The company’s SN50 chips are designed for inference. They are energy efficient and do not require expensive water cooling systems, meaning they can be deployed more quickly than GPUs in a wider variety of data centers. General Compute says the new chips will provide 16 times faster inference than GPU-based clouds.

The challenge is getting a lot of these chips, especially when it’s a startup.

Upper90 co-founder and CEO Billy Libby, a former Goldman Sachs quant trader, had a playbook for this: In 2021, his firm funded GPU purchases by energy-focused data center startup Crusoe, which he believes was the first loan against value for advanced chips.

Traditional lenders avoided these types of deals at the time due to the risks and uncertainties around GPU depreciation. But as CoreWeave turned chip-backed loans into a business model and then the basis of a successful initial public offering (IPO), this type of financing has become common.

“When we funded Nvidia GPUs as the first group to do so, the market was inefficient,” Libby told TechCrunch. “We could really put something together as an early entrant and, in a sense, be compensated for the risk.”

Now that GPUs are comparatively well understood and perhaps overbought, Upper90 is turning to companies like General Compute to take advantage of the next wave of the AI ​​boom. “We think open source models are going to be important, and last year we looked for a player that was in inference,” Libby said. “Not everyone needs a supercomputer, but inference and artificial intelligence do.”

That thesis has been strengthened, with companies that provide access to open models, such as OpenRouter and Fireworks, generating new rounds with enormous valuations. New models like Kimi’s K3, as recently as this week, have proven to compete with the latest releases from Anthropic and OpenAI in coding tests. And new chipmakers like Groq and Cerebras have attracted interest from both buyers and public markets.

General Compute’s ability to access chips outside of the Nvidia ecosystem is important for the same reason. TensorWave, another AI infrastructure company, is making a similar bet in partnership with AMD. As more alternatives to Nvidia emerge, compute vendors that are not tied to agreements with Nvidia may have an advantage in providing cost-effective inference.

“There are a lot of chips that are starting to scale that have incredible [total cost of ownership]or that they can operate much faster than Nvidia, but there aren’t too many buyers for them,” Puklowski said. “By teaming up with Upper90, this isn’t just, ‘a cool startup got some money to buy some computing.’ “This is the first sign that capital is organizing and the fragmentation of Nvidia’s monopolistic dominance.”