Industry

GPU financiers shift to inference chips in $400M deal

General Compute, an AI inference cloud startup, secured a $400 million loan from Upper90, marking what may be the first deal to use inference-specific chips as collateral. The financing signals growing market demand for cost-efficient AI infrastructure that runs open-source models, as investors seek alternatives to expensive GPU-based systems.

Neura News

Neura News

Neura Market Editorial

July 17, 20264 min read
GPU financiers shift to inference chips in $400M deal

AI inference cloud startup General Compute has secured a $400 million loan from tech investment firm Upper90, using chips designed specifically for inference as collateral. The deal marks a shift in how AI infrastructure is financed, moving beyond the Nvidia-dominated GPU market toward specialized hardware for running AI models.

The loan is believed to be the first to put up inference-specific chips as collateral. General Compute, which raised a $15 million seed round in May to build its inference neocloud, will use the financing to deploy SambaNova SN50 chips across a wider range of data centers.

A New Kind of Collateral

Traditional lenders have long avoided chips-backed loans due to risks around GPU depreciation. Upper90, however, has a history of pioneering such deals. In 2021, the firm financed GPU purchases by energy-focused data center startup Crusoe, believed to be the first loan against the value of advanced chips.

Billy Libby, co-founder and CEO of Upper90, said the market was ripe for a similar bet on inference hardware. "When we financed Nvidia GPUs as the first group to do that, the market was inefficient," Libby said. "We could really put together something as an early participant, and kind of get compensated for the risk."

Now, Upper90 is turning to inference companies like General Compute to ride the next wave of the AI boom. "We think open source models are going to be important, and we went and looked for a player last year that was in inference," Libby added. "Everyone doesn’t need a supercomputer, but they do need inference and AI."

Why Inference Chips Matter

Inference chips are built to run already trained AI models quickly and efficiently, at a lower cost than the training chips that have dominated the market. General Compute’s SN50 chips are designed specifically for inference, are power-efficient, and require no water-cooling. The company claims they can be deployed more quickly than GPUs across a larger variety of data centers, and that they provide 16 times faster inference than GPU-based clouds.

Finn Puklowski, CEO of General Compute, framed the deal as a turning point for the industry. "By getting together with Upper90, this is not just, ‘a cool startup got some money to buy some compute.’ Like, this is the first signal of capital organizing itself and the fragmenting of Nvidia’s monopolistic dominance."

The #1 Newsletter in AI

Stay ahead of the AI curve

The most important updates, news, and content — delivered weekly.

No spam. Unsubscribe anytime.

The Rise of Alternatives to Nvidia

The financing signals that markets are responding to concerns over the price of AI tools by turning to cheaper open source infrastructure. Open source models are growing stronger, and companies like OpenRouter and Fireworks, which provide access to open models, have raised new rounds at huge valuations. Kimi’s K3 model competes with Anthropic and OpenAI on coding benchmarks.

More alternatives to Nvidia are emerging. New chipmakers like Groq and Cerebras have drawn interest from acquirers and public markets. TensorWave is making a similar bet on a partnership with AMD. Compute providers not locked into Nvidia deals may have an advantage in cost-efficient inference.

Puklowski noted that the market is ripe for such a shift. "There are a bunch of chips that are starting to scale that have amazing [total cost of ownership], or that can operate much faster than Nvidia, but there’s not too many buyers for them," he said.

A Business Model Validated

CoreWeave, an AI cloud provider, made chips-backed loans a business model and used it as the basis for a blockbuster IPO. Upper90’s 2021 deal with Crusoe was the first of its kind, and the firm is now applying the same logic to inference hardware.

General Compute’s ability to access chips outside Nvidia’s ecosystem is key to its strategy. The startup builds neoclouds, which are purpose-built for AI workloads, unlike general-purpose infrastructure from hyperscalers like AWS or Azure. With GPUs now comparatively well understood and perhaps over-bought, the market is opening up for specialized inference hardware.

The $400 million loan from Upper90 gives General Compute the capital to scale its infrastructure and compete with larger players. Puklowski said the deal represents more than just financing—it is a signal that capital is beginning to organize around a post-Nvidia landscape.

Related on Neura Market

More from Neura News

Product Launch

Acer Unveils Veriton RI110 Mini Workstation for Local Agentic AI

Acer unveiled the Veriton RI110 AI Mini Workstation on September 2, 2026, in Berlin. This compact desktop, featuring an Intel Core Ultra X7 processor and Intel Arc B390 graphics, supports local inference of AI models up to 120 billion parameters. It is designed for hybrid agentic AI workloads, combining local processing with cloud resources, and includes the Qubi Claw software suite for secure, autonomous AI tasks. The system offers up to 96 GB of LPDDR5X memory, 4 TB of SSD storage, and extensive connectivity options including OCuLink, Wi-Fi 7, and dual LAN ports. Availability begins in North America in Q4 2026 and EMEA in Q1 2027.

Sep 2·4 min read