- Baseten is reportedly nearing a $1.5 billion round at a $13 billion valuation
- The jump comes five months after its $300 million Series E at $5 billion
- Conviction and Spark Capital, both prior investors, are reportedly returning to co-lead the new round
Baseten, the San Francisco-based AI inference platform, is reportedly closing in on a $1.5 billion funding round that would value the company at up to $13 billion, according to a report from The Wall Street Journal.
The reported round comes just five months after Baseten, founded by Tuhin Srivastava, Amir Haghighat, Phil Howes, and Pankaj Gupta, announced a $300 million Series E at a $5 billion valuation led by IVP and CapitalG. That followed a rapid step-up to a $2.15 billion valuation just months prior.
The new round is reportedly structured as a split-priced deal, with some investors entering at an $11 billion floor and others paying the full $13 billion headline price, a structure increasingly used by late-stage AI companies raising large sums quickly.
“If cloud was the foundation that enabled the last generation of great technology companies, inference is the foundation for the next,” Srivastava says.
The latest financing is reportedly being co-led by a powerful syndicate including Conviction, Spark Capital, Sands Capital, Altimeter Capital, and Wellington Management.
Why investors are suddenly obsessed with inference
The raise underscores how investor attention is shifting from building large language models toward the infrastructure needed to run them.
Baseten has said industry estimates put inference at two-thirds of all AI compute by the end of 2026, up from one-third in 2023, a shift the company has positioned itself to capture by helping customers run their own models in production rather than relying solely on a handful of frontier labs.
Baseten has spent six years building toward what it calls a multi-model future, betting that AI’s next phase belongs not just to a few giant general-purpose models but to thousands of specialised models built by individual companies for their own use cases.
The startup works by providing developers with systems software that runs the full workload behind an AI application, covering GPUs, autoscaling, observability, billing, and developer tools, so customers can focus on their models rather than the infrastructure beneath them. Its runtimes are open rather than tied to proprietary weights, and its infrastructure spans multiple clouds, which the company says lets customers avoid being locked into a single vendor.
That openness is central to how Baseten positions itself against rivals
Together AI, another fast-growing inference and training platform, and Fireworks AI, which is reportedly in talks to raise at a $15 billion valuation of its own, both compete for the same enterprise customers.
“Baseten is quickly becoming the default infrastructure. In a world where every ambitious AI team wants to run many models and fully own its IP, Baseten gives them the freedom, reliability, and economics to do that at scale,” says Sarah Guo, general partner at Conviction, when the firm backed Baseten’s Series E.
Baseten counts Cursor, OpenEvidence, Abridge, Notion, Clay, Mercor, and Lovable among its customers.
The raise lands amid a broader surge of capital into AI infrastructure. Fireworks AI’s reported $15 billion talks and Together AI’s rising valuation point to a market where investors increasingly see the systems running AI models as valuable as the models themselves.
Baseten had raised $585 million across its prior rounds as of its Series E, according to the company. If the reported $1.5 billion round closes, that would push total funding past $2 billion.
