Open vs Closed AI: Founders Navigate Multi-Model Strategies

Overview

For AI startups, selecting a foundational model is no longer a static architectural decision made once and forgotten. The landscape has shifted dramatically, with open-source models gaining significant performance gains and frontier APIs from proprietary vendors continuing to advance rapidly.

This evolution means that AI products can now call upon a diverse suite of models tailored for specific workloads, rather than committing to a single provider. Founders are increasingly adopting multi-model strategies, which offer greater flexibility but also introduce complex decisions regarding cost allocation, intellectual property ownership, and adaptability as technology evolves.

At TechCrunch Disrupt 2026, taking place October 13-15 at Moscone West in San Francisco, industry leaders are exploring these trade-offs across four key sessions. From application-layer orchestration to the silicon beneath the hood, the event highlights how the choice between open and closed ecosystems shapes both product strategy and long-term business viability.

The Rise of the Multi-Model Approach

The traditional binary of choosing between open and proprietary AI is becoming obsolete for many sophisticated builders. Increasingly, companies are constructing products that leverage different models for different tasks, optimizing for a balance of cost, performance, and flexibility.

This approach allows startups to avoid vendor lock-in and ensures they can rapidly integrate superior models as they emerge. It also affects operational economics, enabling companies to route less complex queries through cheaper, smaller models while reserving high-cost frontier APIs for critical, high-stakes inferences.

Navigating Complexity at Scale

The session “The Real Tokenmaxxing: How the Best AI Companies Navigate a Multi-Model World” on the Builder’s Stage addresses these practical challenges. Featuring Mo Jomaa, partner at CapitalG; Vipul Ved Prakash, co-founder and CEO of Together AI; and Zuzanna Stamirowska, CEO and co-founder of Pathway, the discussion will dissect the mechanics of orchestration.

"Choosing between open and proprietary models assumes a company needs to choose one in the first place. Increasingly, AI products can call on different models for different jobs."

Attendees will learn why leading companies are moving beyond single-model dependencies and how they evaluate when open-weight models can outperform their proprietary counterparts. The conversation aims to provide a framework for balancing the technical capabilities of various models against the business imperatives of speed and cost efficiency.

Deciding What to Own in the AI Stack

Beyond simply selecting multiple models, founders must determine the degree of vertical integration required for their specific business case. This question of ownership—whether to rent infrastructure, customize existing open weights, or build proprietary models from scratch—defines the competitive moat of many emerging AI companies.

Building more of the AI stack offers greater control over differentiation and user experience, but it demands significant talent, time, and capital. Conversely, relying entirely on existing models can limit a company’s ability to innovate uniquely, potentially reducing it to a thin wrapper around commoditized technology.

Practical Frameworks for Founders

Manos Koukoumidis, CEO and co-founder of Oumi, will address this dilemma in his session, “Which AI Should Your Company Actually Deploy: Rent, Customize, or Build,” on the Real World AI Stage.

Using audience polls and real-world startup scenarios, Koukoumidis will provide attendees with three decision principles to evaluate their options:

1. Renting: Leveraging frontier APIs for rapid prototyping and low-infrastructure burden.
2. Customizing: Fine-tuning open-weight models to capture domain-specific advantages.
3. Building: Developing proprietary architectures for maximum control and unique differentiation.

The session is designed to help founders move beyond theoretical debates and apply a concrete framework to their next architecture review, ensuring their choices align with both immediate product needs and long-term strategic goals.

The Hardware-Software Convergence

While software architecture and model selection are critical, the physical hardware underlying AI execution is undergoing its own revolution. Advances in AI are not only changing how applications are built but are also beginning to influence how the chips that power them are designed.

This convergence between model architecture and hardware engineering suggests that the separation between software and silicon is narrowing. For startups, this implies that the efficiency and speed of new AI capabilities may soon depend heavily on custom hardware optimizations that are increasingly driven by AI itself.

AI Designing Its Own Infrastructure

The session “When AI Starts Designing Its Own Hardware” on the Disrupt Stage features Anna Goldie, founder and CEO of Ricursive Intelligence, and Azalia Mirhoseini, founder and CTO. They will explore how AI-driven optimization is reshaping chip design and why model architecture and hardware are becoming inextricably linked.

"Model architecture is only part of the equation. AI performance ultimately depends on the hardware underneath it, and advances in AI are beginning to change how that hardware gets designed."

The discussion will examine what an increasingly open AI ecosystem means for the infrastructure layer. As AI tools begin to assist in designing more efficient hardware, the barrier to entry for high-performance computing may shift, potentially democratizing access to cutting-edge capabilities for startups that can effectively leverage these new hardware designs.

Strategic Implications for Builders

For venture capitalists and founders alike, the trajectory of AI development is characterized by a tension between openness and control. The conversations at Disrupt 2026 underscore that there is no single correct path; rather, success depends on aligning model strategy with business objectives.

As Nader Khalil, Director of Developer Tech at Nvidia, and Sydney Sykes, Global Head of VC Partnerships at Nvidia, will discuss in their session “Building AI Startups Worth Betting On,” the trade-offs between frontier APIs and open-weight models have direct implications for product strategy.

These decisions shape everything from infrastructure costs to the ease with which competitors can replicate a startup’s core technology. By understanding where open and proprietary AI fit within their respective stacks, founders can make informed choices that preserve flexibility while building sustainable competitive advantages.