❯ Baseten seeks about $2 billion in two-tranche financing for its inference business

Two tranches, two proposed pricesBloomberg reported on October 30 that AI inference provider Baseten was pursuing about $2 billion in financing across two tranches at valuations of $26 billion and $30 billion. The financing remains under discussion; the higher price is not a completed valuation for the entire round. An accessible report identifies existing investor Sands Capital as the expected lead, rather than a formally announced, completed transaction.

Running models for product teamsFounded in 2019, Baseten serves developers putting AI into commercial products. It handles deployment, inference optimization, capacity allocation and operations: customers supply open or proprietary models and receive outputs through APIs, with deployment in their own cloud also available. Its product offering includes dedicated inference, model APIs and self-hosting. Customers pay for the compute and software needed to operate models rather than ownership of the models.

A succession of capital raisesEarlier this year, Baseten announced a $300 million Series E at a $5 billion valuation, backed by IVP, CapitalG, NVIDIA and others. That announcement named customers including Abridge, Cursor and Notion and said inference volume had increased 100-fold in the preceding year. Reporting on the proposed round also cites June financing of $1.5 billion at a $13 billion valuation, putting the latest proposed prices after several successive raises.

Usage must support the infrastructureCoding assistants, medical documentation and enterprise search generate inference demand whenever users interact with them. Baseten competes with providers such as Fireworks to lower latency and improve hardware utilization for these production workloads. A larger capital base could support capacity and engineering, although specific uses for this proposed round have not been disclosed. The business must turn rising model usage into revenue that covers the infrastructure required to deliver it.

▮ SIGNALInference providers’ access to capital increasingly depends on turning customers’ recurring usage into profitable compute delivery.