Baseten’s CEO on why every company will want to own its AI

September 15, 2026

AI Summary

5 min read

Baseten’s CEO on why every company will want to own its AI

When Tuhin Srivastava started Baseten in 2019, the company's co-founders had no idea that serving models—what they considered the least important facet of their business at the time—would eventually become the entire company. By late 2024, that bet had paid off so dramatically that Baseten raised at a $5 billion valuation in January 2025 and nearly tripled to $13 billion by June. The reason, Srivastava says, is that the market for AI inference—running models in production rather than training them—is growing faster than almost anyone inside it expected.

What inference actually requires

Inference is what happens when you get a ChatGPT response back, or use any AI-powered product. For closed frontier models like GPT-4, the provider handles everything from the chips to the API. But for open-weight models or custom post-trained models, companies need an inference provider to solve three distinct layers of problems.

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

Never miss an episode of Sources with Alex Heath

Get every new episode summarized in your inbox — free, ~5 minutes to read.

No spam. Unsubscribe anytime.

What you'll learn

1 Baseten’s CEO on why every company will want to own its AI
2 (00:00) **What is AI Inference?** - Tuhin explains inference as the process of serving models in production, distinguishing it from training.
3 (02:47) **The Three-Layer Inference Stack** - Tuhin breaks down Baseten's full stack from heterogeneous compute through orchestration to inference-adjacent primitives.
4 (05:10) **Why Not Just Use AWS or Azure?** - Tuhin explains why specialized inference clouds exist alongside hyperscalers.
5 (08:34) **The Software Layer is What Differentiates** - Tuhin argues that while compute is necessary, the software stack creates real value.
6 (10:12) **The Continual Learning Loop** - Tuhin introduces the concept of a closed loop where companies own their AI improvement cycle.
7 (12:56) **The Recent Capability Shift in AI** - Tuhin and Alex discuss how AI has become genuinely useful in the last 6-8 months.

+ Full timestamped outline available in the app

Show Notes

Tuhin Srivastava is the CEO and co-founder of Baseten, an AI cloud startup recently valued at $13 billion. He tells me why every company will want to own its AI and what’s driving the explosion in demand for inference.

We discuss the rise of agents using browsers and virtual machines to get things done and Baseten’s acquisition of Blaxel to power that shift.

We also talk about the data center backlash, why companies are embracing Chinese open models, his plans for Baseten’s new research lab, and why he thinks inference becomes the only market left after AGI.

Thanks to the show's premier sponsors: Atlassian, Granola, and Mercury.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit sources.news/subscribe