AI infrastructure startup Baseten raises $1.5bn at up to $13bn valuation as Blackbird makes its biggest-ever bet

The AI infrastructure startup raised $1.5bn in a Series F, with Australia’s Blackbird VC putting in what may be the largest single outlay yet by an Australian firm.

Baseten has raised $1.5bn in a Series F round that values the AI infrastructure startup at up to $13bn, a number that arrives barely 18 months and four fundraises into the company’s current growth spurt.

The round was led by the US investors Sands Capital and Wellington Management, and it carries a notable footnote from the other side of the world: Australia’s Blackbird VC says it made its biggest-ever investment in the deal.

The company is California-based but was co-founded by Australians, which is part of why Blackbird’s involvement carries a local angle as well as a financial one.

The firm declined to specify how much it put in, though reporting around the round notes its contribution may be the largest single outlay to date by an Australian VC firm, a record claimed with the appropriate hedging given the undisclosed figure.

What Baseten sells is the unglamorous plumbing beneath the AI boom. The company provides software and infrastructure that let other companies customise and run their own AI models, pitched as a cheaper alternative to leaning on providers such as OpenAI and Anthropic.

Instead of renting intelligence from a frontier lab, customers use Baseten to deploy and operate models themselves, which is increasingly where the recurring costs of AI actually accumulate.

The engine behind the growth is inference, the stage at which a trained model stops learning and starts producing outputs, the answers, images, and predictions that users actually pay for. B

aseten said its revenue grew 20-fold over the past year, driven by demand for exactly that workload. As models move from training runs into everyday production use, the bill for serving them at scale has become its own market, and a crowded one.

The economics explain the rush. Training a model is a one-off, capital-heavy event; inference is a continuous cost that scales with every query a deployed model handles, which means it only grows as adoption spreads.

For companies running AI in production, shaving the price of each inference call compounds quickly, and that is the saving Baseten is selling.

It is also why investors are treating infrastructure as the more durable bet, on the logic that demand for serving models outlasts the hype around any single model.

Baseten is competing in a field that has drawn heavy capital this year. The same demand for cheaper, faster inference has powered rivals chasing inference optimisation, while GPU-cloud upstarts have pulled in their own rounds, among them a fresh GPU-cloud raise aimed at the same customers.

Others are attacking the problem in silicon, with at least one startup building a dedicated inference chip with Anthropic in mind. The throughline is that running models, not just building them, is now where the money and the engineering are converging.

That convergence has reshaped who is willing to write the cheques. Infrastructure providers have become some of the most sought-after names in the sector, attracting the sort of large equity bets that until recently were reserved for the model labs themselves.

Baseten’s jump to a $13bn valuation, on the strength of inference revenue rather than a flagship chatbot, fits that shift neatly.

For the startup, this is the fourth capital raising in 18 months, a cadence that says as much about investor appetite as it does about the company’s own needs.

The next test is whether 20-fold revenue growth holds as inference workloads multiply and the field gets more crowded, and whether a $13bn price tag looks prescient or steep once the next round of competitors reports its own numbers.