20VC: Will OpenRouter Sell for $10BN to Stripe? | Why Chinese Open Models Are Beating America—and What Happens Next | Why Enterprises Are More Fearful of Anthropic and OpenAI Than China | Is the Routing Layer Becoming a Commodity with Alex Atallah

In July, OpenRouter launched 70 new models — roughly one every ten hours. The company sits at the center of the LLM ecosystem, routing inference requests across hundreds of models from dozens of providers. Co-founder and CEO Alex Atallah joined Harry Stebbings to discuss why the routing layer is harder to commoditize than it looks, why enterprises fear frontier model labs more than Chinese open-weight models, and why he believes the AI market will be "the biggest market in human history" with no single winner.

Why the Inference Provider Layer Is Not a Commodity

A common narrative holds that the layer hosting open-weight models — companies like Fireworks, Together, and others — will be commoditized and squeezed by hyperscalers. Atallah disagrees, pointing to a structural reason: Nvidia does not want customer concentration. "One of Nvidia's top priorities is not having customer concentration. They want lots of customers to all have separate allocations of GPUs. They want the heterogeneity of the market." This creates room for specialized inference providers to innovate on serving models efficiently. Atallah notes that benchmarks for a single model like Kimi K3 vary significantly across providers, and those results change constantly. "These models are very emotional. They're very non-deterministic." OpenRouter continuously re-benchmarks every five minutes and shifts traffic