PaleBlueDot AI raises $200M Series C at $3.2B for Super Intelligence compute

PaleBlueDot AI said it completed a $200 million Series C led by ComputeCore at a $3.2 billion valuation, with proceeds earmarked for more Super Intelligence compute capacity across its owned GPU clusters, marketplace, and serverless inference stack.

Another company that rents out AI computing just raised at a multi-billion-dollar price while calling the product Super Intelligence compute. If the $5 billion in signed contracts turns into machines customers actually use, that is real demand. If most of that book is still paper, the $3.2 billion valuation is running ahead of capacity that still has to be built.

On Thursday, 1 October 2026, PaleBlueDot AI said it completed a $200 million Series C. PR Newswire carries the release. The page stamps Oct. 1, 2026, 16:41 ET, which is 4:41 p.m. Eastern. The dateline is Palo Alto, Calif. ComputeCore led the round. The release states a valuation of $3.2 billion. A Series C is a later round of private funding, after earlier money, for a company that is already selling. The $200 million is the new money. A valuation is the price the round puts on the whole company. The release does not say whether $3.2 billion is the price before that new money or after it. It names ComputeCore as the lead and does not describe the firm. Those lines are PaleBlueDot AI’s, on the wire.

Who else put money in, as the release states it. B Capital, an existing shareholder, participated, along with other global investors. An existing shareholder already owned a piece of the company before this round. The release does not say how much B Capital put in, and it does not name the other investors. The round follows a $150 million Series B the company says it announced in January. A Series B is the earlier round, the one before this Series C. The release says January. It does not print the year. Those lines are the release.

What the company says it sells. Founded in 2024 and based in Palo Alto, PaleBlueDot AI calls its product the Super Intelligence Infrastructure Platform. The release says that platform brings together three businesses: GPU clusters the company owns, a GPU marketplace, and serverless inference. A GPU is the chip that does the heavy math for an AI model. A cluster is a group of those chips, run as one pool. A marketplace, here, is a way to buy compute from a network of other suppliers, not only from machines the company owns. Serverless inference means a developer can run a trained model on demand, without reserving a whole cluster first. Inference is the step where a trained model answers, as distinct from the earlier work of training it. Super Intelligence, on this page, is the company’s name for that computing product. It is not a finding that a model has crossed a scientific line. The release says the mission is to make intelligence universally accessible. Those lines are PaleBlueDot AI’s.

The machines it says it runs itself. The company owns and operates dedicated GPU clusters, and it says it customizes them for customers with large or specialized workloads. A workload is the job a customer runs on the chips. Dedicated means the cluster is set aside for that use, rather than shared minute by minute with whoever shows up. The release says a B300 cluster in Japan recently earned NVIDIA Exemplar Cloud status, and that the status validates the cluster’s performance against NVIDIA’s reference benchmarks. A reference benchmark is NVIDIA’s own test of how that hardware should perform. B300 is the release’s name for that cluster. Recently is the release’s word. It does not print the day the status was granted. That status line is the company’s claim in the release. The release does not say NVIDIA bought shares in this round.

The other two ways it says a customer can get compute. Through the marketplace, the release says, customers get on-demand access across a global network of supply partners. A supply partner is another operator whose machines the marketplace can offer. The serverless offering, the release says, delivers flexible Super Intelligence compute for developers and enterprises. An enterprise, here, is a company buying the product for its own work. The release says the company’s clusters sit in multiple regions, and that customers span markets worldwide. It says the supply chain was built for efficiency and security, so customers can bring capacity online quickly and run it with confidence. Capacity, here, means more chips available to rent. Those lines are the company’s. The release does not print a count of regions, a count of partners, or a number of days from order to a running cluster.

The contract book, and whose number it is. As of the end of September 2026, the company said it had signed over $5 billion in customer contracts. Signed means a contract is in place. Over $5 billion is the company’s figure for the value of those contracts. The release does not say that money has been collected, and it does not call it one year of revenue. It also says customers in the United States and Japan together accounted for more than half of monthly revenue. More than half is the release’s share. It does not print the monthly revenue in dollars, and it does not say which of the two countries is larger. The same paragraph says that this year the company has diversified its data-center capacity and its customer base. Diversified means the capacity and the customers are spread across more than one place and more than one kind of buyer. Those lines are PaleBlueDot AI’s. The release does not name a customer.

Where the new money goes, and what the chief executive said. Proceeds will fund additional Super Intelligence compute capacity, so customers have more choice in locations and hardware. Hardware, here, is which chips and which machines a customer can pick. Stephen Watts, chief executive, said: “We will continue to broaden our customer base, with a focus on frontier labs, Neolabs and enterprises in the U.S., and we will invest in our full-stack and go-to-market teams.” A frontier lab, in that sentence, is a lab building the most capable models. Neolabs is the release’s word. The release does not define it. Full-stack, here, means people who work across the product, from the machines to the software a customer uses. Go-to-market is the work of selling that product and supporting the companies that buy it. That quotation is his, in the release.

The about box, and the offices the company site prints. The release says PaleBlueDot AI is named for the photograph of Earth that Voyager took in 1990, the picture Carl Sagan called a pale blue dot. It says the company shares a belief that Super Intelligence technology can benefit people. The contact on the release is palebluedotAI@fgsglobal.com. The company site calls the product an intelligence platform and repeats the line about making intelligence universally accessible. It lists offices in Palo Alto, which it calls the U.S. headquarters, at 1881 Page Mill Rd, and in New York, Seoul, Tokyo, and Singapore. Those are offices on the site. They are not a list of data centers, and the financing release does not print the street addresses. The site also describes customized compute: dedicated clusters for large deployments, reserved capacity for planned work, and flexible GPU sourcing across its network. A customer, the page says, requests capacity, the company designs it, and then it is deployed. Those lines are the company’s, on its site.

The picture is the company’s marketplace page for that product. The headline on the frame reads “Reserved GPU Clusters for Enterprise-grade Custom Token Factory Solutions.” A token is a small piece of text, or other input, that a model reads or writes. Token Factory is the company’s name for turning reserved GPU clusters into that work. Under the headline, the frame prints four figures: 150+, 219,595+, 60+, and 25+. The live company site uses those four slots for GPU clusters, GPUs connected, regions, and supply partners. On that live page the GPUs-connected line reads 220,000+. The frame’s second figure reads 219,595+. The other three figures match the live page. The frame does not print a calendar date. It is the product page. It is not a photograph of a data hall, and it does not show a signed contract.

In plain terms, PaleBlueDot AI said on Thursday that it raised $200 million in a Series C led by ComputeCore, at a valuation of $3.2 billion, to add computing capacity. B Capital, already a shareholder, took part, with other investors the release does not name. The company, founded in 2024 in Palo Alto, says it combines clusters it owns, a marketplace, and on-demand inference, and that a cluster in Japan recently met an NVIDIA reference bar. It says that by the end of September it had signed more than $5 billion in customer contracts, and that U.S. and Japan customers together are more than half of monthly revenue. Those contract and revenue lines are the company’s. The release does not say the $5 billion has been collected, does not name a customer, and does not say NVIDIA bought shares in the round.