Pomegra
Baseten raised $1.5 billion in a June 2026 Series F led by Altimeter Capital, Conviction, and Spark Capital, reaching a valuation of up to $13 billion after 20x revenue growth and one billion daily inference requests.
Key Takeaways:
- The round valued Baseten at up to $13B across two tranches, up 160% from a $5B Series E closed just five months earlier.
- Annualized revenue reached approximately $600M in Q1 2026, up from $200M at the end of 2025.
- Inference volume grew 40x year-over-year; the platform now processes over one billion requests per day.
The Round in Full
Baseten closed a $1.5 billion Series F on June 22, 2026, structured across two tranches priced at $11 billion and $13 billion respectively. Altimeter Capital, Conviction, and Spark Capital led. Sands Capital and Wellington Management co-led, with additional participation from IVP, Greylock, Battery Ventures, D.E. Shaw Ventures, Durable Capital Partners, Blackbird, and others.
The deal arrived five months after a $300 million Series E in January 2026 - backed by Nvidia and CapitalG at a $5 billion valuation. That round now reads as either conservatively priced or simply overtaken by events.
What Does Baseten Actually Build?
Baseten builds infrastructure for AI inference - the production step where trained models generate outputs for real users at scale. The company offers dedicated compute for open-source, fine-tuned, and custom models, alongside model APIs for prototyping and a training pipeline for teams building specialized systems.
Its commercial proposition centers on inference at up to 30% lower cost than closed-source API providers, with 99.99% uptime and global deployment. Customers include Cursor, Clay, Abridge, Lovable, Mercor, and OpenEvidence - a range of developer tooling, healthcare AI, and enterprise software companies that share a common need for high-throughput, customizable model serving.
Why Did Investors Move This Fast?
The pace is the story here. Baseten raised a $300 million round five months ago and returned for more than four times that amount. The numbers make the logic legible.
Annualized revenue hit roughly $600 million in Q1 2026, up from $200 million at year-end 2025 - a tripling inside a single quarter. Inference volume grew 40x year-over-year. One billion requests per day is a number that most infrastructure businesses never reach. Investors tracking that trajectory had a narrow window before the next round priced even higher.
The structural dynamic helps too. Open-source and closed-source models have converged in capability for a broad range of workloads, pushing competition away from model quality and toward inference economics. AI infrastructure companies that can serve models fast, cheap, and at customizable precision are capturing margin that used to belong to the model providers themselves.
What Does a $13B Valuation Imply About the Last One?
At roughly $600 million in annualized revenue, the $13 billion valuation carries an approximately 21x forward revenue multiple. Elevated, but not without precedent in infrastructure categories where customer switching costs are high and revenue predictability is strong.
The two-tranche structure at $11 billion and $13 billion suggests the deal was assembled incrementally, with different investors pricing different levels of confidence about near-term execution. The gap between tranches also reflects a known risk: GPU inference pricing has fallen steadily as compute supply increases, which compresses margins even when volume climbs. Growing 40x on volume is only accretive if the per-unit economics hold.
Strategic Context
Baseten now competes in a market where cloud hyperscalers run subsidized inference to retain platform spend, and a growing number of purpose-built inference providers are targeting the same open-source model workloads. The company's differentiation rests on customization depth and multi-cloud flexibility - arguments that resonate more with AI-native startups than with large enterprises still negotiating hyperscaler contracts.
Its customer list skews toward companies building at the application layer: coding assistants, clinical AI, recruiting tools. These are businesses whose unit economics depend directly on inference cost. That alignment creates stickiness but also makes Baseten's revenue sensitive to the fortunes of a concentrated segment of the AI startup market.
Use of Capital
Baseten plans to triple headcount in 2026, with investment concentrated in engineering, research, operations, and go-to-market teams. The company has not disclosed specific hiring counts by function.
Outlook
The AI inference market is expanding faster than the model layer in operational complexity, and Baseten has built revenue scale early in that cycle. The $1.5 billion raise provides runway to compete on infrastructure depth against providers with much larger balance sheets. The central question for the next 12 to 18 months is whether unit economics improve with scale or erode as inference commodity pricing falls across the industry. A tripling of headcount adds cost well before it adds revenue. How that equation resolves will determine whether the $13 billion valuation was a clear-eyed read - or a generous one made easy by exceptional short-term momentum.
