Shares of CoreWeave, a publicly traded neocloud that lends out computing power for AI, have fallen to 12-month lows, Elon Musk is paying down $40 billion in debt to buy computing power, and predicts that “more than 200 neoclouds will suffer guppy-level extinction and consolidation.” Not exactly an ideal week to announce NeoCloud’s Series A.
Finn Puclowski, co-founder and CEO of General Computing, remains unfazed. He is bullish on CoreWeave’s potential and believes Musk is just speaking his mind to lower the price of the GPUs he wants to buy. Concerns about a chip-driven collapse are “highly exaggerated. Every week I have people calling me and asking me to use their TPUs, GPUs, and other chips to get more computing power.”
Equally important, Anthropic’s bullish S-1 “takes away today’s fears.”
General Compute has found a niche building clouds from specialized inference chips developed by Nvidia rivals. The company signed a $400 million deal to buy new chips from SambaNova, its first debt financing for an inference chip, and now adds Positron and Cerebras to its portfolio of chip suppliers.
The signing of a major customer with $100 million in annual recurring revenue paved the way for the company to raise $100 million in Series A funding led by Blackbird Ventures and Abstract Ventures. General Computing, valued at $60 million after a seed round in May, is now worth $800 million post-money to investors.
The startup’s strategy is to zig where other companies zig. While everyone else was chasing GPUs, they were focused on chips built for inference. And instead of taking on the complexities of building huge data centers, they’re using smaller power sources, renting space in other companies’ data centers, and converting former Bitcoin mines into places to run their AI models.
This approach means General Compute won’t immediately rise to the top of neoclouds or compete with hyperscalers. However, it also does not have the limitations associated with building a data center from scratch. If General Computing has a chip, there’s a place to connect it. Puklovsky expects the company to install 75 megawatts of computing power in 2027, and suspects it will have 200 megawatts of power available. “It’s probably more than we can put the money into and get the hardware.”
Still, planning has its limits. Mr. Puclovskiy’s fundraising work isn’t done yet. “Equity is going to be a full-time job for me, both in terms of debt and equity,” he told TechCrunch.
The startup has found a major customer in a cloud company that specializes in providing fast inference to developers. General Compute provides bare metal, or raw servers with no software layering, to bring its products to market. Blackbird President James Palmer, who co-led the round, said part of the company’s confidence in General Computing is due to its investment in inference cloud company Baseten.
“We’ve been following their journey for the past 10 months, observing the incredible growth story[Baseten]has led, open-weight inference, and the demand for companies to own proprietary intelligence,” Palmer told TechCrunch. When it comes to general computing, “Cerebras, Positron and SambaNova all need distribution partners who can scale them and navigate the complexities of the physical world,” he said.
Executives from all three chipmakers praised General Computing to TechCrunch, noting the importance of the company’s focus on making the economics of AI cloud computing work. “Everyone in this market is selling speed,” said Rodrigo Liang, CEO and co-founder of SambaNova. “The real question is, can you make money doing it?”
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
