Nvidia Executives Debate Open vs. Closed AI Models as Strategic Choice for Startups

Nvidia's Nader Khalil and Sydney Sykes took the stage at TechCrunch Disrupt 2026 to frame one of the most consequential decisions facing startup founders today: whether to build on open or closed AI models. The session positioned this choice not as a technical or philosophical preference, but as a core business decision with implications that ripple across nearly every aspect of a startup's trajectory.

The Core Business Trade-offs

Founders choosing a proprietary frontier model can move quickly with established API access, while building on an open model offers greater control over data and deployment at the cost of managing more infrastructure decisions. The choice affects multiple dimensions: cost, infrastructure requirements, margins, differentiation, speed to market, and operational control.

The decision becomes more complex because model capabilities and economics are shifting rapidly. Founders must consider whether their current choice locks them into a particular vendor relationship or technology stack, and whether that lock-in creates defensibility or merely operational dependency. If two models can deliver similar results, the question shifts: does lower cost win, or does the ability to own more of the stack create defensibility worth the infrastructure maintenance burden?

The Closing AI Capability Gap

Open models have advanced quickly, with Nvidia reporting that 145 papers accepted at ICML 2026 cited its Nemotron open models and datasets, spanning robotics, autonomous vehicles, and biomedical research. At the same time, proprietary frontier labs continue advancing model capabilities, creating a market where the distinction is increasingly less about whether open models are useful and more about where each approach makes commercial sense.

This convergence has reshaped the strategic debate. Nvidia CEO Jensen Huang argued at GTC 2026 that the future is not proprietary versus open, but proprietary and open together—a framing that reflects the reality that many companies are now combining open and proprietary models rather than treating them as mutually exclusive.

Where Competitive Advantage Actually Lives

The open-versus-closed debate ultimately raises a deeper question about competitive defensibility: if competitors can access the same proprietary API, differentiation must come from proprietary data, workflow, distribution, customer relationships, product experience, or specialized technology. Choosing an open model does not automatically create a competitive moat either—it provides flexibility and control but requires founders to make decisions about deployment, optimization, and infrastructure that competitors might not face.

Nvidia is investing heavily in the open ecosystem, launching Nemotron 3 Super in March 2026 as an open 120-billion-parameter model designed for agentic workloads, with companies already combining it with proprietary models rather than viewing the two as opposing choices.

The Panelists and Their Perspectives

The two moderators bring complementary expertise to the debate. Nader Khalil, Nvidia's Director of Developer Tech, approaches the discussion from the builder and infrastructure perspective. Before joining Nvidia, he co-founded Brev.dev, an AI infrastructure company acquired by Nvidia in July 2024, which was built around simplifying GPU infrastructure access across public cloud, private cloud, and on-premises environments. His background positions him to discuss how developers can avoid lock-in while maintaining flexibility across compute environments.

Sydney Sykes, Nvidia's Global Head of VC Partnerships, brings the venture ecosystem perspective into the discussion. Together, the pairing creates room to examine the open-versus-closed choice from two angles: what developers need to build and what companies need to become investable, scalable businesses.

Why This Matters Beyond Startups

The debate has implications across multiple constituencies. For founders, the choice shapes margins, the fundraising narrative, and product roadmap evolution. For investors, understanding where value sits in the AI stack helps distinguish genuine defensibility from a thin product layer dependent on someone else's model. For line-of-business leaders, the decision affects procurement, security, data control, infrastructure costs, and the ability to change vendors later if circumstances shift.

That Nvidia—which supplies the GPU infrastructure underlying both open and closed AI models—is framing this as a central startup decision reflects a deeper shift in how AI capability is structured. The company has a commercial stake in both sides of the architecture debate, making its willingness to present the trade-offs neutrally significant. Nvidia's role in the AI supply chain is foundational; GPU architecture became the default compute substrate for neural network training, meaning the company's position remains central regardless of whether startups choose open or closed models.

The Broader Strategic Context

This panel framing emerges at a moment when the open-versus-closed question has become entangled with supply chain access, regulatory considerations, and geopolitical strategy. Open models can be deployed on-premise in jurisdictions with restricted access to US cloud services, while closed models typically require API access through US-headquartered providers subject to export regulations. For defense and security applications specifically, this distinction carries heightened importance: open weights allow local deployment without ongoing vendor dependency, which is why governments and large enterprises are increasingly evaluating open-source AI for infrastructure use cases where control and sovereignty matter.

The timing and positioning of this debate at TechCrunch Disrupt reflects how deeply the open-versus-closed question has moved from academic or ideological territory into operational business strategy. For startups making product decisions today, the choice will have compounding effects over months and years as the AI landscape continues to shift.