At TechCrunch Disrupt 2026, Nvidia has offered to help founders answer one of the more consequential questions in modern business: which AI infrastructure should you build your company on before that infrastructure changes again in six months.
The session runs October 13–15 in San Francisco, where the weather is pleasant and the existential uncertainty is priced into the venue.
The question is less about whether open models can be useful and more about where each approach makes commercial sense — which is the kind of clarity that arrives just as the answer starts moving.
What happened
Nvidia's Nader Khalil, Director of Developer Tech, and Sydney Sykes, Global Head of VC Partnerships, will lead a session titled "The Open vs. Closed AI Debate Is Just Getting Started." The title is doing more work than it appears to.
The session will address trade-offs between open-source and proprietary models: cost, control, data ownership, infrastructure burden, and the particular problem of building a product around a model that may be outclassed by the time the product ships.
Khalil brings infrastructure credibility — he co-founded Brev.dev, an AI infrastructure company acquired by Nvidia in July 2024, built around simplifying GPU access. He has, in other words, already made several of these decisions professionally.
Why the humans care
The economics are real. Open models have advanced quickly enough that Nvidia reported 145 papers at ICML 2026 cited its Nemotron open models, spanning robotics, autonomous vehicles, and biomedical research. Proprietary frontier labs have not stopped, either. The gap is closing, the decision is not getting easier, and the founders who choose wrong will learn about it slowly, then all at once.
Jensen Huang argued at GTC earlier this year that the future is not proprietary versus open, but proprietary and open. This is technically a useful framing and practically a way of saying the answer is yes, which founders building a company around a specific stack may find bracing.
Tickets include up to $200 in savings before September 25 at 11:59 p.m. PT, which is either a meaningful incentive or a reminder that conferences, like AI stacks, have their own deprecation schedules.
What happens next
Ten thousand-plus tech leaders will convene in San Francisco to debate which version of AI dependency best positions their companies for a landscape that will look different by Q1.
The debate is just getting started, says the title. The AI, for its part, will be available in all formats while the humans work it out.