Baseten is set to raise about $2bn at a valuation of up to $30bn, more than double the $13bn price set in June. Bloomberg reporters Rebecca Torrence and Bailey Lipschultz disclosed the plan on Friday, citing people familiar with the talks. The deal would rank among the largest private rounds for AI infrastructure this year. For business users, it signals that running models in production has become the most contested part of the AI stack.
Two tranches and a June baseline
The proposed round is split into two tranches, one valuing Baseten at $26bn and the second at $30bn. Sands Capital, an existing investor, is set to lead the financing. Bloomberg reported the structure, while TNW noted it had not independently verified the report. The baseline for comparison is the Series F announced on 22 June, a $1.5bn round led by Altimeter Capital, Conviction Partners and Spark Capital. Sands Capital co-led that June round, which Bloomberg valued at $13bn.
Baseten sells software for running AI models in production, a stage known as inference. The product sits between a trained model and a working application, handling scaling, latency and cost control. This week the company released joey-1, described as an open decision model priced at a penny per million tokens. That pricing move shows how inference vendors compete on unit economics as well as reliability. Founded in 2019, Baseten said at the time of the June round that its revenue had grown 20 times in a year.
Investor interest now concentrates on inference, the work of running models after training is complete. Fireworks AI, described by Bloomberg as Baseten closest competitor, raised $1.5bn at a $17.5bn valuation in July. The two financings put large private capital behind competing inference platforms within weeks of each other. Bloomberg also reported on 22 October that venture firms have been using tranched rounds to push up AI startup valuations. In such a structure, investors buy into the same company at two different prices.
What this means for AI buyers
For companies deploying AI, a better-funded inference layer usually means wider model choice and faster path to production. Baseten approach abstracts hosting, scaling and optimization, which lowers the engineering effort needed to keep assistants, search and automation running. Small firms gain access to production-grade serving without building their own infrastructure team. Large enterprises get additional vendor capacity for high-volume workloads and for testing models such as joey-1 at low token cost.
The figures still require caution because the new round is reported, not closed or independently confirmed. A tranched valuation sets two prices for the same company, so the headline $30bn applies only to part of the capital. Buyers should separate fundraising momentum from service terms on uptime, latency, data handling and support. The news by itself does not prove lower inference prices or superior performance versus Fireworks AI. Vendor checks should focus on measured cost per million tokens and behavior under peak load.
The marker to watch is whether the $2bn round closes at $26bn and $30bn and who joins Sands Capital. Confirmation would lock in Baseten as one of the most valuable private inference providers alongside Fireworks AI. A delay or revised terms would suggest limits to tranche-driven pricing. Either outcome will shape vendor options for production AI in 2027.
