AI inference startup Baseten is in talks to raise $1 billion at an $11 billion post-money valuation — more than double the $5 billion valuation it announced just three months ago. The massive leap reflects the insatiable demand for raw compute power to run open-source AI models.
By the end of the first quarter of 2026, Baseten's annualized revenue had exploded from $200 million at the start of the quarter to roughly $600 million — a 20-fold increase from the same period in 2025. The growth is so rapid that some investors have reportedly floated informal offers valuing the company at nearly $15 billion. For context, competitor Modal recently raised at a $4.65 billion valuation with $300 million in annualized revenue, while Together AI closed a round at $7.5 billion this spring with annualized revenue north of $1 billion.
Founded in 2019, Baseten rents Nvidia AI servers to developers, letting them fine-tune, customize, and run inference on open-source large models. In December 2025, it acquired Parsed, a startup that helps enterprises customize open-source models. Baseten charges on a pay-as-you-go basis — either by the number of tokens consumed through API calls or by the time chips are rented. To date, it has raised $585 million from investors including Nvidia, CapitalG, IVP, and Spark Capital.
But even with soaring revenue, inference providers face margin compression. The insatiable GPU appetite of OpenAI and Anthropic is driving up spot prices for chips, and tight supply means Baseten — which essentially subleases GPUs from major cloud providers — could see its profits squeezed. Meanwhile, most large enterprises still prefer to use closed-source models directly from OpenAI or Anthropic, limiting the market for open-source inference.