Inference Platform Baseten Raises $150M Series D at $2.15B

AI inference platform Baseten closed a $150M Series D, reportedly at a $2.15B valuation, to scale its AI inference services.

Three key facts

The money goes to running models, not building them. Baseten doesn't train large models — it deploys other people's models and makes them run faster and cheaper. It's the most practical link in shipping AI: a powerful model is useless if it can't run affordably and reliably.

Inference cost is the make-or-break for AI apps. As apps shift from chat to agents — calling a model dozens or hundreds of times per task — inference speed and unit cost directly decide whether an AI business makes money. That's exactly the spot Baseten occupies.

A $2.15B valuation. While the model-layer giants burn cash at war, capital is betting heavily on the "middle layer" — the companies that turn models into usable, affordable services.

WangDou's Take

The "sell water" business of inference is being fought over by investors — a $2.15B valuation for a company that builds zero models of its own, and just makes other people's run faster and cheaper. The model-builders up top are bleeding cash, competing over who loses more; the ones quietly making money are below, handing out shovels and collecting tolls. In a gold rush, the safe bet was never the prospector.

Source: Crescendo AI — Latest VC Investment Deals in AI Startups (2026)