Work Backwards From First Principles

Asked at Stanford GSB’s View From The Top series (published 25 April 2024) how he makes decisions, Huang said he goes back to first principles “all the time, even in the company today.” His method, in his words: “You figured out what is the end result that you’re looking for and you work backwards to achieve it.”

Why re-derive. On the Lex Fridman Podcast (#494, March 2026) he explained why last year’s plan goes stale quickly: “these AI model architectures are being invented about once every six months,” while “system architectures and hardware architectures” change “kind of every three years.” If the workload keeps changing underneath the hardware, the roadmap has to be re-derived from the problem, not extended from the last version.

In practice. This is the reasoning behind several plays in this chapter: co-designing the whole system, shipping on an annual rhythm, and sizing a bet by what would have to be true for it to work.

Doesn’t transfer without the depth. First-principles reasoning is only as good as the reasoner’s grasp of the underlying physics and economics. Huang is an electrical engineer who has run the same company for 33 years. In most organizations, “first principles” becomes a slogan for ignoring hard-won experience.