Open or closed AI? How founders are choosing what to build on at TechCrunch Disrupt 2026

At TechCrunch Disrupt 2026, founders and investors will debate whether to build on open or proprietary AI models and how much of the AI stack companies should own. Panels and stage sessions will address multi-model strategies, model customization, hardware co-design, and the trade-offs between frontier APIs and open-weight models.

By AI Newsroom· Reviewed by Pranav, Founder & Editor-in-ChiefPublished 39 minutes agoUpdated 39 minutes ago0 views
Open or closed AI? How founders are choosing what to build on at TechCrunch Disrupt 2026

Why It Matters

These discussions reflect a shifting reality in AI product design: startups increasingly combine multiple models and customize weights, meaning model choice affects costs, product differentiation, and long-term control. The conference sessions aim to give founders practical frameworks for those architectural and business decisions.

Key Facts

  • Event: TechCrunch Disrupt 2026
  • Dates: October 13-15, 2026
  • Location: Moscone West, San Francisco
  • Attendees expected: More than 10,000 founders, investors, operators, and tech leaders
  • Speakers and exhibitors: 250+ speakers; 300+ exhibiting startups

Founders face more nuanced model choices as open models, frontier APIs, and customization options all advance. At TechCrunch Disrupt 2026, multiple sessions will examine why companies are increasingly using more than one model, how to choose which model for which task, and what owning parts of the stack actually means for a startup’s product and balance sheet. One Builders Stage session, “The Real Tokenmaxxing: How the Best AI Companies Navigate a Multi-Model World,” will feature Mo Jomaa of CapitalG, Vipul Ved Prakash of Together AI, and Zuzanna Stamirowska of Pathway. The panel will discuss why multi-model architectures are rising, trade-offs among cost, performance, and flexibility, and cases where open models can outperform closed alternatives. Another conversation will focus on ownership versus renting of model infrastructure. Manos Koukoumidis, CEO and co-founder of Oumi, will lead “Which AI Should Your Company Actually Deploy: Rent, Customize, or Build,” on the Real World AI Stage. That session will present audience polls, startup scenarios, and a practical framework to help attendees weigh frontier APIs, customized open weights, and building models in-house, and will distill three decision principles for architecture planning. TechCrunch Disrupt will also tackle the intersection of model design and hardware. On the Disrupt Stage, Anna Goldie and Azalia Mirhoseini of Ricursive Intelligence will explore how AI is being used to optimize chip design and why closer coupling between model architecture and hardware could influence how quickly new capabilities reach the market. Separately, Nvidia representatives Nader Khalil and Sydney Sykes will discuss how founders are balancing the trade-offs between frontier APIs and open-weight models and the implications for product differentiation and infrastructure. Across more than 200 sessions, attendees can expect practical guidance on when to mix models, when to customize or build, and how hardware trends might shape future options. Organizers highlight matchmaking and dealmaking opportunities alongside the programming; registration promotions noted in the event materials include early discounts and a limited-time $75 Expo+ Pass for people affected by layoffs.

Keep Reading