The AI Stack Decision: Open or Closed?

Why This Choice Matters More Than Ever

When building an AI product, founders face a fundamental question: Should we build on a proprietary frontier model or leverage open-source models? It's not just a technical decision—it's a business strategy that can shape the entire trajectory of a company.

"The question is increasingly less about whether open models can be useful and more about where each approach makes commercial sense."

I've spent considerable time observing how startups approach this decision, and what strikes me most is how quickly the landscape shifts. Just a few months ago, the choice seemed clear-cut—proprietary models for advanced capabilities, open models for flexibility. But today's reality is far more nuanced.

The Rise of Hybrid Approaches

Nvidia's own CEO Jensen Huang recently argued that the future isn't proprietary versus open, but rather proprietary and open. This shift in perspective is not just philosophical—it's practical.

Consider how companies like Brev.dev have positioned themselves around simplifying access to GPU infrastructure across different environments. That kind of infrastructure approach allows developers to deploy AI software across public cloud, private cloud, and on-premises infrastructure without locking themselves into a single compute source.

This is precisely what we're seeing in the market today: companies combining open models with proprietary ones rather than treating the two approaches as mutually exclusive. It's a hybrid reality that requires deeper strategic thinking.

Open Models: The New Reality

Open models have advanced rapidly. In July, Nvidia reported that 145 papers accepted at ICML 2026 cited its Nemotron open models and datasets. That's a significant indicator of how quickly the open ecosystem is maturing.

But here's where it gets interesting: while open models provide flexibility and control over data, they also introduce new challenges. The economics can shift dramatically depending on workload and scale. And if you're building a product that needs to be defensible, open-source models don't automatically create a moat.

For developers and students, this means understanding how technical decisions today directly connect with business models being built around them. It's not just about what technology works—it's about how you can leverage it strategically.

Proprietary Models: The Traditional Approach

Proprietary models continue to push capabilities forward, but they come with their own set of trade-offs. Companies that rely on proprietary APIs have to think harder about differentiation—because competitors can access the same underlying technology.

The challenge for startups is not just in choosing between open and closed, but in understanding where value actually lives in the stack. If you're building a product that's heavily dependent on model performance, your competitive advantage may lie elsewhere: in proprietary data, specialized workflows, distribution channels, customer relationships, or unique product experiences.

This is why I've seen so many successful startups invest heavily in building around their specific use cases rather than simply leveraging someone else's AI stack. It's a strategic choice that can make all the difference in how investors perceive your company and how customers interact with your product.

Strategic Implications for Founders

For founders, this decision impacts everything from margins to fundraising stories to product roadmaps. If you're building a product where the core value is in how you use AI rather than what AI you use, then choosing an open model might be more strategic.

But if your business model relies heavily on proprietary capabilities or specific data controls, a proprietary approach may be necessary. The key is understanding that neither approach is inherently superior—it's about aligning with your business objectives.

I've seen companies make these choices based on their immediate needs, only to find that the landscape has shifted by the time they're ready for the next phase. That's why it's crucial to think not just about what you're building today, but how you'll adapt as technology evolves.

The Investor Perspective

From an investor standpoint, understanding where value sits in the stack is critical. A product layer built on someone else's model isn't necessarily defensible unless it's tied to something unique—whether that's data, distribution, or customer relationships.

This is particularly important for early-stage companies that are trying to establish their competitive moat. If your differentiation doesn't come from the underlying AI technology itself, you need to be clear about where else you're creating value. Otherwise, investors may question whether your company can scale beyond its initial proof of concept.

The reality is that we're entering a new era in which the boundaries between open and closed are blurring. Companies that can navigate this hybrid landscape while maintaining clear strategic focus will likely be best positioned to succeed.

Building for Tomorrow

What makes sessions like the one featuring Nader Khalil and Sydney Sykes at TechCrunch Disrupt so valuable is that they're not just about technology—they're about strategy. They're helping founders understand the trade-offs they face, especially when it comes to infrastructure choices that can have lasting implications.

The AI landscape isn't static anymore. As we move forward, we'll see more companies building hybrid stacks, using open models for some functions while maintaining proprietary solutions for others. This isn't just about convenience—it's about maximizing the strengths of each approach while minimizing their weaknesses.

Ultimately, this decision is not about choosing between open and closed—it's about choosing how to build a sustainable business that can adapt as the technology evolves. That's a strategic move that every founder should consider carefully.