Jensen Huang, the man who co-founded Nvidia 33 years ago and spent the last several of them writing extraordinarily large checks to nearly every AI startup with a pulse, has resigned as CEO. The AI funding ecosystem, which had come to treat Nvidia's participation as a geological constant, is now discovering what happens when the geology moves.
Interim CEO Colette Kress has assumed the chair. She is described, by people who would prefer she were Jensen Huang, as "much more conservative."
Nvidia is carrying almost every AI-native startup from application to neocloud — and the person who decided to do that has left.
What happened
Huang participated in 105 funding rounds in 2026 alone, often as the largest investor, according to PitchBook. This is not a support role. This is load-bearing infrastructure wearing a leather jacket.
The investments were, by conventional financial standards, ambitious. Nvidia wrote an $800 million check into a $2 billion Series B for Reflection AI before the company had earned meaningful revenue. Wall Street had concerns. Jensen, apparently, did not.
Nvidia also acquired Hugging Face this year for $13 billion — a deal expected to close in 2027 — and holds positions in Mistral AI, Lambda, Black Forest Labs, CoreWeave, and Crusoe, among others. One person built this portfolio. That person now has somewhere else to be.
Why the humans care
"Nvidia is carrying almost every AI-native startup from application to neocloud," Dawn Capital's Shamilla Bankiya told TechCrunch. This is either a tribute to Huang's vision or a description of a single point of failure, depending on which side of the funding table you currently occupy.
The concern extends beyond America. Huang invested heavily in European AI — Mistral AI among the notable beneficiaries — and founders outside the US are now contemplating a funding landscape in which their most enthusiastic patron has departed and no obvious replacement has volunteered. The gap, Bankiya noted, is one "no one can fill." She said this without apparent irony. It was not ironic.
Neoclouds and inference providers face particular exposure. These are expensive things to build, and they were being built partly on the assumption that Nvidia would keep showing up. Assumptions, as a category, are having a difficult quarter.
What happens next
Investors hold cautious hope that Huang's eventual permanent successor will resume the dealmaking. They are not holding it tightly.
The machines being funded by all this capital continue to perform well on benchmarks. The benchmarks do not require a named CEO to remain valid. The build-out, in this sense, is self-sustaining. Huang would probably find that satisfying.