At TechCrunch Disrupt 2026, AI founders are facing complex decisions about whether to build their products on open or closed AI models. Rather than choosing a single AI model, many startups are leveraging multiple models for different functions, combining open-source and proprietary technologies to optimize cost, performance, and flexibility. This multi-model approach, highlighted by experts such as Mo Jomaa of CapitalG and Vipul Ved Prakash of Together AI, allows companies to adapt more rapidly to emerging capabilities in AI.

A key topic discussed is how much of the AI stack startups should own, especially when deciding between building custom models, renting frontier APIs, or customizing open models. Manos Koukoumidis from Oumi offers a framework at Disrupt to help founders evaluate trade-offs between these options. Companies that build more of their own models can gain competitive control and differentiation but must weigh the increased time, effort, and resource demands that come with greater ownership of AI infrastructure.

Speakers from Nvidia, including Nader Khalil and Sydney Sykes, will explore the practical trade-offs between open and proprietary AI and how these decisions impact startup costs, product strategy, and competitive advantage. These choices have a significant effect on a company’s ability to innovate and maintain control over its technology in an environment where AI evolves quickly and unpredictably.

Beyond software, advances in AI are influencing hardware design, with founders like Anna Goldie and Azalia Mirhoseini of Ricursive Intelligence discussing how AI is starting to design chips and optimize hardware. This intertwining of model architecture and hardware development may accelerate the pace of innovation and shape the infrastructure that supports the next generation of AI products. Attendees can learn more about these developments and strategic AI decisions at TechCrunch Disrupt San Francisco, October 13-15, 2026.