Baseten

Investing (Again) in the Inference Cloud

“Civilization advances by extending the number of important operations which we can perform without thinking about them.”
-- Alfred North Whitehead

Turn on a faucet, and water flows. Flip a switch, and the lights come on. Open a laptop, and you’re connected to the internet. Each advance takes something essential and makes it effortless, so we can focus on what comes next.

Baseten is doing that for machine learning. From day one, their mission was to make ML effortless, so more people can use it to build useful things.

Since inception, the “deployment gap,” the problem of turning trained models into production grade application services, has only grown more acute. Production inference demands both complex scale infrastructure with specialized performance optimization. Without Baseten, teams must build GPU orchestration, autoscaling, multi-cloud/geo routing, and low-level optimizations like batching, KV-caching, token streaming and operator fusion, all while also handling capacity and vendor agreements in a volatile hardware market. As models advance, open source proliferates, and ambitions for AI applications grow, the underlying stack remains immature, brittle, and hard to scale. This drives massive demand not just for GPUs, but for a true inference platform—performant, reliable, global, and developer-first. We need the AI grid.

You don’t build a utility overnight. But you can if you hold to the right principles. At Baseten, they’re clear:

• Do it right. From CEO to new hire, the team acts with integrity and total obsession. Uptime is oxygen for our customers. The team is permanently on call for the most important AI applications in production.

• Do it with long-term ambition. Aim for the frontier, then aim higher. Our dedicated inference and model APIs set the standard, consistently winning on latency, throughput and reliability. Achieving this requires patience to rewrite abstractions until they’re correct, and ambition to tackle foundational projects like Multi-Cloud Management. It means building to scale, with infrastructure that spans ten clouds and forty regions to deliver best-in-class coverage, capacity, and cost.

• Do it like developers. World-class experts in model performance push the frontier, but the ethos is transparency. Optimizations become productized and configurable, not hoarded secrets. The company competes on outcomes, infrastructure quality, and the earned trust of a consistently green status page.

• Do it with the best people. This team is stacked, and the entire company holds itself to the highest standards. They partner with the most demanding customers (including Abridge, OpenEvidence, Clay, Zed Industries, Gamma, Sourcegraph and Notion) whose challenges keep them at the bleeding edge. I’m thrilled to welcome Jay Simons, former Atlassian President and now General Partner at Bond, as lead investor in the Series D and board member, and Dannie Herzberg, formerly of Sequoia, Hubspot, and Slack, as President. Both are world class operators that embody urgency, ambition, and principled long-term thinking—the exact qualities that define this company.

At the center is Tuhin. He’s one of the most grounded leaders I know, with an unwavering compass. From day zero, he asserted that even the biggest internet companies shouldn’t need to build ML platform teams. Baseten should handle that, so they can just ship. Over five years and thousands of texts and late-night conversations, his decisions have been strikingly consistent: do the right thing, make it real in production, never let customers down, hold the highest standard. Simple, hard, magnetic principles. They make everyone around him better.

I’m grateful to Tuhin, Amir, Phil and the entire Baseten team for letting me be part of this journey. I invested on Day Zero. This was my first bet that AI would be the most important technological shift of our lifetimes. I’ve invested in every round of this company, most recently in the $150M Series D, through our new venture fund, Conviction. And I hope to still be here on Day 10,000. I’ve never been more confident in the scale of the opportunity, or in this team’s ability to build the winning platform.

Let’s build and run the power grid for AI. If you want to work on the frontier of model performance and infrastructure, Baseten is hiring.