Founders are stuck in a paradox. Build fast with proprietary APIs or gain control with open weights? The choice defines your margins, infrastructure, and defensibility. Nvidia is forcing the industry to stop treating this as a binary. At TechCrunch Disrupt 2026, Nader Khalil and Sydney Sykes will dissect the hybrid reality facing builders today.
Khalil brings the engineer’s view. As Director of Developer Tech, he led Brev.dev before Nvidia acquired it in July 2024. He knows the pain of locking into single compute sources. Sykes offers the investor lens. As Global Head of VC Partnerships, she sees which architectural choices actually scale. Together, they’ll argue that the future isn’t proprietary versus open. It’s both. Jensen Huang made this clear at GTC. Now, startups must operationalize that ambiguity.
The gap is closing. Nvidia reported that 145 papers at ICML 2026 cited its Nemotron open models. These aren’t just academic curiosities. They power robotics, autonomous vehicles, and biomedical research. Yet frontier labs keep pushing capabilities higher. The question shifts from utility to commercial fit. Does lower cost win if results are similar? Or does data control matter more? Owning the stack creates defensibility but also maintenance debt.
Your model isn’t your moat. Competitors can access the same API. Differentiation must come from proprietary data, workflow, or distribution. Open models offer flexibility but demand deployment expertise. Nvidia’s Nemotron 3 Super, a 120-billion-parameter model for agentic workloads, shows the trend. Companies combine it with proprietary tools rather than choosing one side. This hybrid approach may be the only sustainable path. Register for Disrupt 2026 in San Francisco this October 13-15. Save up to $200 before prices rise on September 25 at 11:59 p.m. PT. The clock is ticking on both your decision and your wallet.
