Baseten has raised $150 million in a Series D funding round, pushing its valuation to $2.15 billion. The round was led by BOND, with new participation from CapitalG, Premji, and Kevin and Elizabeth Weil of Scribble, alongside returning investors Conviction, 01A, IVP, Spark, and Greylock. The company has now raised more than $285 million since its founding six years ago.
Baseten’s focus is on inference, the step where trained models are deployed and used in real-world applications. Unlike training, which has received much of the attention in recent years, inference is about delivering speed, reliability, and scale when AI systems interact with end users.
“Inference is the foundation for the next generation of technology companies,” said co-founder and CEO Tuhin Srivastava. “Every breakout AI application depends on fast, reliable, and cost-effective inference, the same way the last 15 years of companies depended on the cloud.”
Baseten’s Inference Stack is designed to manage those demands. The company says it provides faster runtimes, cross-cloud capacity for resiliency, and developer-friendly tools, claiming performance gains of 40–50% in head-to-head tests against competitors.
Customers in Production
That infrastructure has become critical for a growing group of customers. Healthcare startup Abridge, which generates more than a million clinical notes weekly, uses Baseten to keep its systems responsive for tens of thousands of clinicians. CEO Shiv Rao said the company depends on “inference that is both fast and trustworthy,” describing Baseten’s infrastructure as an important part of scaling safely across health systems.
OpenEvidence, another customer, runs billions of model calls each week to provide up-to-date medical information to clinicians. Its co-founder Zachary Ziegler noted that “if you see a doctor today, chances are that they are leveraging OpenEvidence and Baseten,” adding that the infrastructure has held up under what he described as “life-or-death” reliability requirements.
Even in non-healthcare settings, speed and reliability matter. Clay, an AI go-to-market platform, uses Baseten to launch new features more quickly. Co-founder Kareem Amin said the partnership gives his team “the confidence [capabilities] will perform for customers,” calling Baseten “a critical piece” of their delivery.
Product Expansion
Baseten has also been broadening its offering. Alongside its core Inference Stack, the company has introduced Baseten Model APIs and Baseten Training, tools designed to help developers deploy models more efficiently and tailor them to specific use cases. By combining inference infrastructure with developer tooling, Srivastava argues, Baseten allows teams to “focus on how they use the models and the applications they build—while giving us the boring piece.”
For investors, the company’s ability to simplify inference has been central. “Inference is a major bottleneck, and solving it is key to unleashing the full potential of AI,” said Jay Simons, general partner at BOND, who has joined Baseten’s board. He compared the company’s trajectory to Atlassian, where he previously served as president: “Baseten is on a similar trajectory. They are years ahead in both product and adoption.”
CapitalG partner Jill Chase highlighted the role Baseten plays in shortening development cycles: “Baseten doesn’t just support AI innovation—it accelerates it,” she said, describing the platform as an enabler for teams building products that reach customers quickly.
Scaling the Business
The company has grown its team from about 30 employees a year ago to more than 100 today. Srivastava said the fresh capital will be used to expand engineering and customer-facing teams, as well as to continue investments in model performance and infrastructure. “It feels like a land grab right now,” he said. “This fundraise is about solving two core problems—building a killer go-to-market team, and hiring incredible engineers, who are both expensive and hard to find.”
Baseten began focusing heavily on inference in 2022, just before the launch of ChatGPT spurred widespread interest in generative AI. Since then, it calls itself a behind-the-scenes provider for companies building model-driven products.
For Srivastava, that role is deliberate: “The smart companies realize infrastructure isn’t their differentiator. They should focus on how they use the models and the applications they build.”
