Baseten nears $1.5 billion round at $13 billion valuation

Baseten is closing in on a $1.5 billion financing at a $13 billion valuation, a leap that would put one of the AI infrastructure market’s most closely watched startups under a brighter spotlight. The proposed deal comes only months after Baseten announced a $300 million Series E at a $5 billion valuation, underscoring how quickly investor appetite for inference has accelerated.

The company sits in the part of the AI stack that runs after a user submits a prompt. That inference layer is where latency, cost and reliability can make or break enterprise use, and Baseten has built its pitch around making that process faster and cheaper. Its platform serves open-source, custom and fine-tuned models on infrastructure designed for high-performance inference, while its pricing model charges customers only for the compute they use, down to the minute.

Baseten’s customer pages point to demand from Abridge, Clay, Cursor, Decagon, Descript, EliseAI, Heygen, Lovable, Notion, OpenEvidence, Poolside and Writer. The company says those workloads have seen 30% to 50% lower p99 inference latency and 44% lower cost per million characters, claims that speak directly to the economics investors are betting on. In a market where the costs of running models can overwhelm the promise of the models themselves, those savings are the difference between a useful product and an expensive experiment.

The financing also highlights how aggressively capital is still flowing into the infrastructure layer. Baseten’s Series E on Jan. 23, 2026 was led by IVP and CapitalG, with NVIDIA, Altimeter, Battery Ventures, BOND, BoxGroup, Blackbird Ventures, Conviction and Greylock among the investors. Baseten said at the time that the round brought total capital raised to more than $285 million.

The company traces its origins to 2019, when co-founders Tuhin Srivastava, Amir Haghighat and Philip Howes launched Baseten after working together on machine-learning and product problems at Gumroad. Greylock says the startup began with $20 million in seed and Series A funding led by Greylock, and Baseten later said its thesis was that many specialized models would run in production at scale and that fast, reliable inference would be the critical link between AI promise and real-world impact.

NVIDIA added another signal in February 2026, naming Baseten alongside DeepInfra, Fireworks AI and Together AI as providers reducing cost per token by up to 10x on Blackwell hardware. If the new round closes on these terms, it will further cement Baseten as a benchmark for whether the AI infrastructure boom is being driven by real enterprise demand, or by investors still willing to pay almost any price for the companies powering the next layer of the stack.