Open or closed AI? How founders are choosing what to build on at TechCrunch Disrupt 2026
TechCrunch TechCrunch Events ● Covered by 3 sources
Founders are no longer picking one AI model and moving on. At Disrupt 2026, the big question is how much of the stack to own, and when flexibility beats lock-in.
Based on reporting by TechCrunch, TechCrunch Events — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
AI startups are treating model choice less like a one-time architecture call and more like a moving target. Open models are getting better, frontier APIs keep improving, and some products now mix several models instead of betting everything on one. That shift gives founders more room to build, but it also forces harder choices about cost, control, and how much flexibility to leave on the table.
That tension will be front and center at TechCrunch Disrupt 2026. On the Builder’s Stage, Mo Jomaa of CapitalG, Together AI’s Vipul Ved Prakash, and Pathway’s Zuzanna Stamirowska will talk through why companies are going multi-model, how they balance cost against performance, and when open models can beat proprietary ones. The point isn’t just technical elegance. Model choice affects operating costs and can shape how fast a startup can adopt better tools as they show up.
A different session asks a bigger question: what should a startup own at all? Manos Koukoumidis of Oumi will lead a Real World AI Stage discussion on whether companies should rent, customize, or build. His session will use audience polls, startup scenarios, and a practical framework to weigh frontier APIs, customized open weights, and owning AI outright. That matters because building more of the stack can bring control and differentiation, but it also takes more time, talent, and resources.
Nvidia’s Nader Khalil and Sydney Sykes will take a similar angle from the Builders Stage, focusing on the trade-offs between frontier APIs and open-weight models and how those choices flow into product strategy and long-term differentiation. And then there’s the hardware layer. Anna Goldie and Azalia Mirhoseini of Ricursive Intelligence will look at how AI is already helping design chips, and what a more open ecosystem could mean for the infrastructure under all these products.
Disrupt itself is set to be a fairly crowded proving ground: 200-plus sessions, six industry stages, roundtables, breakouts, more than 10,000 expected attendees, 250-plus speakers, and 300-plus exhibiting startups, all at Moscone West in San Francisco from October 13-15. The subtext is simple. In AI, flexibility is becoming a strategy, not a luxury.
My take — AI-written commentary, not fact-checked reporting
This is the right argument to have, because the real moat in AI is rarely “we picked the cleverest model.” It’s usually knowing when to rent, when to tweak, and when to stop pretending you need to own the whole stack. The open-versus-closed debate is still useful, but startups that worship it like a religion usually end up paying for the ceremony.
Read more about this at: TechCrunch