We first covered Baseten last year when the DeepSeek hype reached its peak. Today, they have completed a massive $13 billion funding round and have become one of the new batch of AI infrastructure decacorns (AI Infra decacorns). Alongside Nvidia, Intel, and the broader semiconductor complex, they are the biggest beneficiaries of the Inference Inflection.
We are turning our attention back to Baseten just as the 2026 Open Weights debate reaches its peak. Ali posted a viral teardown analysis of Kimi K3:
ali (@waterloo_intern) I spent 48 hours digging into the Kimi K3 model code.
This consumed:
- 650 mg of caffeine (required)
- 40 cans of LaCroix (optional… world record (?))
- 8 papers
- 6 months shaved off my lifespan
Finally figured out the entire evolution of Kimi K3 and how we got here… every step of the way since
Since you last saw him, Philip has spoken at the AI Engineer conference and written the definitive book on Inference Engineering, which can be seen everywhere in San Francisco:
Three years ago, inference engineering barely existed as a category.
Today, it is one of the most critical disciplines in the AI space. Inference engineering fundamentally addresses a completely different problem from standard model training: “How do you turn trained weights into a product that is fast, reliable, and cost-effective at scale?” Focusing on these questions has created an entirely new optimization puzzle.