Baseten raises $2 billion at $30 billion valuation for AI inference infrastructure

What happened

Bloomberg reported on October 30 that Baseten is in talks to raise about $2 billion across two tranches. The first tranche would value the AI inference startup at $26 billion; the second would value it at $30 billion. Existing investor Sands Capital is expected to lead. Bloomberg's own headline framed the raise as roughly doubling the $13 billion valuation Baseten hit in June, when it closed a $1.5 billion Series F.

None of this is official. The reporting relies on people familiar with the talks, the deal has not been announced, and terms can move before signing. Axios reported in September that talks were happening at a $26 billion valuation; that figure is now the floor, and the step to $30 billion is the new information. Bloomberg did not name who else joins Sands at the top of the book, and did not say whether the two tranches close concurrently or on a staggered schedule.

Why inference matters

Training is the expensive part of AI that everyone talks about: months of GPU time to build a model. Inference is everything after that — the work of answering a single request once the model exists. Every time a user asks a chatbot something, something has to load the model, run the text through it, and stream tokens back. That something needs GPUs, scheduling software, caching, and the ability to scale up under traffic spikes and down when demand fades. Each query costs compute, and the bill lands on whoever serves it.

Baseten's business is exactly that layer. The company, founded in 2019 by a team that previously worked together at Gumroad, operates a serverless inference platform that takes open-weight models and its own tuned versions, runs them on GPUs across multiple clouds, and bills customers for the compute they use. A company that wants a specific model without building a serving team rents the whole arrangement. Bloomberg tied the new round directly to growing demand for this kind of infrastructure.

The valuation context

The comparison that makes the round striking is Baseten's own history. In June 2026 the company closed a $1.5 billion Series F at a $13 billion valuation. If the top tranche of the new round lands at $30 billion, Baseten is worth about 2.3 times more four months later. Nothing about the product changed by a factor of two in that window — the market's estimate of the category did. Inference, long the unglamorous half of the AI stack, is being repriced as the expensive half.

The two-tranche structure is itself a signal. Pricing the first tranche at $26 billion and the second at $30 billion lets Baseten reward the investors who commit early while still selling later money at a higher mark — useful when a valuation is climbing faster than the paperwork can keep up. Inference-layer funding has been a running theme across deal coverage for the past three months, and this round is the largest data point yet.

Who is investing

Sands Capital, an existing Baseten backer, is expected to lead the round. The company's cap table already includes Altimeter Capital, IVP, and Spark Capital from earlier rounds. None of those names are new, which is worth noting: this is an insider-led round at a higher price, a structure that signals conviction from investors who have seen the company's numbers up close.

Neither Bloomberg nor the available summaries name who else joins Sands at the top of the book, and the timeline is equally underspecified. It is unclear whether both tranches close at the same time or in sequence, and whether the $30 billion mark is guaranteed or contingent on the second close. For a raise of this size the participant list and the tranche mechanics will matter as much as the headline valuation.

Competitive landscape

Baseten is not alone in the inference lane, and the nearest comparison is instructive. Rival Fireworks AI raised $1.5 billion in July at a $17.5 billion valuation. If Baseten clears $30 billion, it lands roughly two-thirds higher on a round only a few months newer. That gap is either a bet on Baseten's specific customers or a sign that the whole category is being repriced at once — and the round's timing suggests the latter.

At the low end, serving open-weight models is close to a commodity skill, and the big cloud providers can bundle inference into deals they have already signed. The differentiation an inference specialist sells is engineering: batching requests so GPUs stay busy, caching repeated prompts, routing each job to the cheapest hardware that can handle it, and taking model deployment and autoscaling off a customer's plate. Baseten also pushes its own open decision models through its Base Labs arm, which keeps it in the model conversation as well as the serving one.

What it means for the AI infrastructure market

The round is a market signal before it is a company event. A $2 billion raise at a $30 billion valuation for a company that serves other companies' models says investors expect the token volume explosion to keep compounding. Reasoning models think before they answer, and that thinking is billed like any other output — a single question can generate many times the tokens of a plain model. Agents make it worse in a commercially good way: an agent that runs for twenty minutes, calls tools, checks results, and retries a failed step is generating tokens the entire time, turning one user action into hundreds of model calls.

That math is why capital is pouring into the layer even as inference prices keep falling. Cheaper tokens mean more usage, and more usage means more infrastructure spend — the classic repricing of a commodity that is growing into a bottleneck. The implications reach across the market: GPU capital is the constraint on every inference provider, hyperscalers will keep bundling inference into enterprise deals, and the neocloud shakeout that some founders have predicted will separate providers with real serving economics from those without. For anyone building on top of these services, the practical read is simpler: pricing has been falling for two years and the money pouring in suggests that trend is not about to reverse.

Caveats

This is a reported round, not a closed one. Bloomberg's story relies on people familiar with the talks; Baseten has not confirmed it, and terms can change before signing. The investor list beyond Sands Capital is unnamed, and the mechanics of the two tranches — whether they close together, in sequence, or at all — are unspecified. The most useful thing to watch is whether both tranches close at the reported prices: a single-tranche deal at $26 billion would tell you the top end did not hold. Treat the numbers as a negotiation in progress, not a completed transaction.