AI Founders Face a Growing Dilemma Over Model Architecture

At TechCrunch Disrupt 2026, industry leaders will address the shifting architecture of modern AI. The central challenge lies in determining how much of the technology stack a company should own versus what it should rent. For many, the answer is a multi-model approach, where different architectures handle specific workloads to balance performance and cost efficiency. Experts including Mo Jomaa of CapitalG, Together AI’s Vipul Ved Prakash, and Pathway’s Zuzanna Stamirowska will explore how this flexibility impacts product strategy.

Beyond software, the physical layer is also transforming. Anna Goldie and Azalia Mirhoseini of Ricursive Intelligence are set to discuss how AI is increasingly involved in designing the very hardware it runs on. This convergence of model architecture and chip optimization suggests that the infrastructure powering the next generation of startups will be as fluid as the software itself. As thousands of attendees gather in San Francisco, the focus remains on building resilient businesses that can adapt to rapid technological turnover without being locked into a single vendor’s roadmap.