‹ rbtfl.
⋯
LeadersConflictsDefenceAIEnergySpaceStartupsTradeMineralsShippingMoneyDebtFood all trackers ▸

Baseten raises $1.5bn at $13bn valuation as AI inference demand hits 1 billion daily calls

Altimeter, Conviction and Spark lead the largest-ever pure-play inference infrastructure round; 20x revenue growth in a year and 87 clusters across 18 clouds position Baseten as the production layer beneath AI applications

Startups·AI· active Whose Money·The Long Game ·5 takes · ·rbtfl upd Jun 26, 2026

Summary

AI inference startup Baseten closed a $1.5bn Series F on 22 June 2026 at a $13bn post-money valuation, led by Altimeter Capital, Conviction Partners and Spark Capital. Sands Capital, Wellington Management, Battery Ventures, Australian VC Blackbird, D.E. Shaw Ventures, Durable Capital, Greylock and IVP also participated. Total capital raised now exceeds $2bn. The company, founded in 2019 by CEO Tuhin Srivastava and co-founders Amir Haghighat, Pankaj Gupta and Philip Howes, provides cloud-native infrastructure for deploying and scaling production AI models, processing over 1 billion inference calls per day across 87 clusters spanning 18 cloud providers. Revenue grew 20x year-over-year in the twelve months before the close. The raise was structured across two tranches at $13bn and $11bn, reflecting rapid re-rating as inference workloads scaled through H1 2026. Headcount is tripling in 2026. As foundation models commoditise and application builders compete on cost-per-token, inference infrastructure companies like Baseten, Supabase and AWS Bedrock are absorbing the margin that model providers once captured.

The split

US tech press frames Baseten as the production-layer winner in a market where model weights are increasingly commoditised. The 20x revenue growth figure is treated as a hard signal, not hype. Blackbird's participation, noted by European tech media, is read as evidence that inference infrastructure is no longer purely a US venture thesis. No significant critical voice has emerged; the main skeptical thread from specialist AI commentary is whether Baseten can defend margin against hyperscaler-native inference options from AWS, Google and Azure as they lower per-token pricing.

By the numbers

  • $1.5bn, Series F round size.
  • $13bn, post-money valuation (structured across two tranches).
  • $2bn+, total capital raised to date.
  • 1 billion+, inference calls processed per day.
  • 87, global clusters across 18 cloud providers.
  • 20x, revenue growth year-over-year before close.
  • June 22, 2026, round announced.

Why it matters

Baseten's raise marks inference infrastructure, not model weights, as the new frontier capital target. As OpenAI and Anthropic monetise at the model layer, the companies winning the deployment layer, the actual compute routing, batching and serving stack, are capturing a structurally different margin pool. A $13bn valuation for a 7-year-old infrastructure company with no consumer product signals that investors expect inference costs to become the dominant operating expense for AI applications, and that whoever controls the serving layer has pricing power.

What to watch

  • Whether Baseten files for an IPO in 2026 or 2027 alongside Supabase and Databricks.
  • How hyperscalers respond with native inference pricing cuts targeting Baseten's core market.
  • Competitor trajectory: Together AI, Fireworks AI and Modal are all pursuing similar positions.
  • Whether the two-tranche valuation structure (at $13bn and $11bn) signals LP caution at the top of the AI infrastructure cycle.