Baseten secures $1.5 billion at a valuation of up to $13 billion for AI inference.

      Baseten is nearing completion of a $1.5 billion funding round that values the company at as much as $13 billion. The structure of this round is nearly as striking as its size. It features a dual-tier system, with some investors entering at an $11 billion valuation and others at $13 billion, as stated by the company to the Wall Street Journal. This approach has recently been employed by several AI startups to enhance their headline figures. The funding is co-led by Altimeter Capital, Conviction, Spark Capital, Sands Capital, and Wellington Management.

      Baseten offers the less glamorous aspect of the AI stack. Established in 2019 in San Francisco, the company delivers the necessary software and computational resources for businesses to conduct inference—the phase in which a trained model provides answers to queries. It leases capacity from about 20 cloud providers and integrates its own inference software, allowing clients to deploy and optimize models based on their data.

      Much of its operations utilize more affordable, open-source models instead of cutting-edge systems from OpenAI or Anthropic. This is the core of their strategy. “Open-source models are getting very, very good,” said CEO Tuhin Srivastava to the WSJ, noting that customers are increasingly combining open and proprietary models based on the complexity of the tasks.

      A rapid revaluation

      Investors are quickly backing this thesis. The latest developments in the EU tech scene, along with insights from our wise founder Boris and some questionable AI art, are available for free every week in your inbox. Sign up now!

      Baseten was valued at $5 billion as recently as January, following a $300 million round supported by IVP, CapitalG, and Nvidia. This new financing more than doubles that valuation in approximately five months. As of September 2025, the company's worth was around $2.15 billion.

      The driving force is demand. Clients such as Cursor, Mercor, and OpenEvidence leverage Baseten to operate models, with the company indicating that some have significantly reduced costs by transitioning to open-source options; one reportedly completed a task for roughly 30% of the cost of a proprietary equivalent.

      The picks-and-shovels business

      This funding occurs amid a surge in AI inference, where the infrastructure that enables efficient model operation has become highly sought after in the industry. Cerebras, a maker of inference chips, went public earlier this year, and Fireworks AI is targeting the same model-serving market.

      It also takes place during a price competition. Many popular open-source models now originate from China, including DeepSeek and Moonshot AI, and Nvidia has even unveiled an open family named Nemotron. With OpenAI reportedly considering substantial price reductions to remain competitive, this pressure poses challenges for leading labs while, theoretically, benefiting those who assist companies in utilizing cheaper alternatives.

      The risk aligns with the opportunity. Baseten is being valued as a beneficiary of a cost reduction, even as this reduction pressures profit margins across the industry. Assigning a $13 billion valuation to a company at the heart of a price war inherently suggests a belief that the providers of essential tools will outlast the rush for gold.