Baseten Raises $1.5 Billion Series F, Its Fourth Round in 18 Months

A fourth raise, and a fivefold repricing, in a year and a half

Baseten has raised $1.5 billion in a Series F financing that values the AI inference company at up to $13 billion—the fourth time in 18 months it has gone back to investors, and the clearest sign yet that the plumbing beneath AI applications has become one of the most fiercely contested corners of the technology economy.

The round, announced June 22, was led by Altimeter Capital, Conviction, and Spark Capital, with Sands Capital and Wellington Management participating as co-leads. IVP, Greylock, 01A, Blackbird, Durable Capital Partners, Verified Capital, Battery Ventures, and D. E. Shaw Ventures also contributed alongside existing backers. Unusually, the capital came in across two tranches priced at $13 billion and $11 billion—a structure that hints at just how quickly the company's valuation has been moving. As recently as January, Baseten's Series E raised $300 million at a $5 billion valuation, itself only three months after a Series D. The San Francisco company, founded in 2019, has now raised more than $2 billion in total.

What Baseten actually sells

Baseten builds the systems software that runs AI inference—the moment a trained model is put to work answering a query, generating an image, or transcribing a call—at production scale. Its platform manages the full workload: GPUs, autoscaling, observability, billing, and developer tooling, spread across what the company says is 87 clusters and 18 clouds. That multi-cloud architecture, Baseten notes, is "itself a core part of what customers are buying," letting companies chase available GPU capacity without being locked into a single provider.

The customer list reads like a directory of the current AI boom: Cursor, Notion, Lovable, Harvey, HubSpot, OpenEvidence, Abridge, Decagon, Clay, and Mercor among them. These are firms building products where, as Baseten puts it, "intelligence is core to the user experience."

The growth figures are the reason investors are willing to reprice the company every few months. Revenue has climbed roughly 20x year-over-year and inference volume 40x. The platform now processes more than one billion inference calls every day. Reports pin Baseten's annualized revenue run-rate near $600 million by early 2026, up sharply from a fraction of that a year earlier. Management says the new capital will fund talent, compute, and enterprise go-to-market, with headcount set to triple this year.

Why inference economics are the real story

The most telling detail in Baseten's disclosure is not the valuation but a shift in how AI companies spend. According to the company, leading app-layer firms now direct 30 to 50 percent of their model spend toward custom and post-trained open-source models rather than closed frontier APIs. Open-weight models have grown capable enough to serve as genuine alternatives, and post-training them on proprietary data has become, in Baseten's framing, an act of ownership rather than a cost-cutting hack.

That is the thesis underwriting the raise. If the AI value chain's early years belonged to whoever trained the biggest model, the emerging phase rewards whoever can serve millions of specialized models cheaply, quickly, and reliably. Inference is a recurring, volume-driven cost that scales with usage, which makes it structurally more attractive—and more defensible—than the episodic economics of training runs.

"The future of AI will be built on millions of specialized models, and the companies building the best ones know that post-training has become existential," said Tuhin Srivastava, Baseten's CEO and co-founder. "The companies closing this loop fastest are using Baseten to build a system that compounds in value over time."

His investors echo the compounding language. "The inference market is compounding faster than almost anything we've seen, and some of the most sophisticated AI companies in the world are standardizing on Baseten to power it," said Apoorv Agrawal, a partner at Altimeter Capital, who called the company's trajectory "hard to overstate."

Baseten is not alone in chasing this layer—Together AI, Fireworks, Modal, and the major cloud providers all want the same workloads—and a $13 billion price tag on a company still measured in hundreds of millions of revenue leaves little room for a slowdown. The two-tranche pricing suggests even some participating investors wanted a cushion.

What to watch

Three things will determine whether this valuation holds. First, gross margins: inference is capital-intensive, and Baseten's economics hinge on squeezing more throughput from scarce GPUs. Second, retention among its marquee customers, several of whom are young companies whose own spending could swing hard in either direction. Third, the pace of open-weight model improvement—the trend that makes Baseten's post-training pitch compelling could also commoditize it. With a war chest this large and a hiring plan to match, the company has bought itself room to answer those questions. The next raise, if the pattern holds, may not be far behind.