Baseten funding hits $US1.5b as Blackbird makes record bet

Baseten, an AI infrastructure startup co-founded by Australians Tuhin Srivastava and Amir Ghavi, has raised $US1.5 billion (about $2.3 billion) at a $US13 billion valuation (about $20 billion). Blackbird said the deal included its largest investment so far, putting an Australian venture firm inside one of 2026’s bigger AI infrastructure financings.

The financing matters because Baseten sits below the model race most consumers see. The company runs inference systems, the layer that serves trained models inside live products. In that layer, customers care less about launch theatrics and more about latency, cost, reliability and whether the application keeps working when usage jumps. Baseten’s revenue grew 20-fold over the past year as demand for inference capacity rose, according to Reuters. Model providers can win attention with new releases, but enterprise spending tends to follow when a system is cheap enough to run every day.

This is Blackbird’s bet: the invoice matters after the demo works.

For local readers, the Australian connection changes the read. Baseten’s founders are Australian, and Blackbird is the most visible local investor in the deal. Michael Tolo, a Blackbird partner, told Reuters the investment was “a signal of conviction.” He also said AI was seeing “the biggest shift” so far in unit economics and competitive leverage. The round ties Australian founders and venture capital to the paid, production side of the AI market, rather than the more familiar launch-day race among model labs.

Blackbird had put that argument in writing before the new round. In its investment note on Baseten’s Series E, the firm said task-specific open-source models could cut inference bills by 60 to 80 per cent compared with some workloads on OpenAI and Anthropic models. If customers see savings in that range, tools for deploying, tuning and managing those models become part of the budget conversation instead of back-end plumbing.

The timing is unusually fast. The new funding comes five months after a $US300 million Series E, after investors re-priced Baseten as AI spending shifted from trials to production systems. A company moving from one mega-round to another that quickly is a sign of how aggressively late-stage capital is chasing infrastructure attached to real usage. Private-market valuations still need caution, and a $US13 billion mark can compress if revenue growth slows, cloud costs bite or customers consolidate around bigger vendors.

For local founders, that is the practical signal. Capital is chasing the less glamorous parts of AI. For teams building around model serving, evaluation or cloud cost controls, that is a more useful signal than the headline valuation. It points to the infrastructure investors think can turn AI adoption into repeat spending.

Blackbird is using its biggest cheque on infrastructure under the current AI product wave, not on a consumer app or a general model lab. That does not settle whether Baseten’s valuation holds. It does show how closely Australian capital is now tied to the global inference race. It also gives Blackbird a test case for whether Australian venture capital can keep writing meaningful cheques as AI infrastructure rounds move into Silicon Valley scale.