An Australian who left Sydney for an Electrical Engineering degree in Los Angeles is now running one of the most valuable AI infrastructure companies on the planet, and Blackbird has just backed him for the third time, this time with its largest single investment into a startup round, ever.
Baseten, the San Francisco-based AI inference platform that lets companies deploy, optimise and run open-source and custom models in production, has raised $1.5 billion USD in Series F funding. The round was led by Altimeter Capital, Conviction and Spark Capital, co-led by Sands Capital and Wellington Management, with participation from Blackbird, IVP, Greylock, 01A, Durable Capital Partners, Verified Capital, Battery Ventures and D.E. Shaw Ventures.
The financing was completed across two tranches, valuing the company at $11 billion and $13 billion USD, respectively. It is Baseten's fourth raise in 18 months and takes the total capital raised past $2 billion USD.
For the Australian context, Baseten was co-founded by Tuhin Srivastava and Philip Howes. Howe’s attended the University of Sydney and co-founded Shape Analytics with Srivastava. More on that below.
Srivastava grew up in Sydney, attended Normanhurst Boys' High School, and began his career as an investment banker at Macquarie Group before moving into machine learning and founding the company in 2019 alongside Amir Haghighat, Phil Howes and Pankaj Gupta.
Srivastava is also a venture partner at Athletic Ventures, keeping a foot firmly in the ANZ ecosystem even as Baseten scales out of San Francisco.
A relationship a decade in the making
The firm has now backed Srivastava and Philip Howes three separate times.
It first invested in his people-analytics startup Shape Analytics in 2015, which was later acquired by Reflektive, then participated in Baseten's $300 million Series E in January, and has now returned for the Series F.
According to Niki Scevat, this is Blackbird's largest single cheque into a single company raise, with more than $200 million committed.
For context, a Blackbird fund in 2012 was $30 million.
Why inference, why now
Baseten sits at the layer of the AI stack that turns trained models into working products, handling the GPU orchestration, autoscaling, observability and developer tooling that sit between a model and an end user. The thesis is that as open-source models close the gap on closed APIs, enterprises increasingly want to run custom, post-trained models they own rather than renting intelligence from a handful of frontier labs. Leading AI application companies now direct between 30 and 50 per cent of their model spend toward customised and post-trained models, according to Baseten.
Revenue has grown roughly 20x year-on-year, and the platform now processes more than one billion inference calls a day across 87 clusters and 18 clouds.
Customers include Cursor, Notion, Lovable, Abridge, Clay, Mercor and OpenEvidence.
What the money is for
The capital will fund compute, software and talent, with Baseten planning to triple its headcount this year across engineering, research, operations and go-to-market.
"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, CEO and co-founder of Baseten. "It's how they build intelligence they own, on data that's theirs, optimized for the customers they serve."