Should American Enterprises Work With Open-Source Chinese Models? | Only 10% of Neo-labs survive

By 20VC

Categories: VC, Startup

Summary

In 3 years, 99% of AI workflows will run on open-source models, not frontier models—and the smartest model is often the cheapest when measuring outcome cost, not token cost. Companies must shift from evaluating price per input to price per outcome to understand true AI economics.

Key Takeaways

  1. Measure AI model economics by outcome cost, not token consumption. A sophisticated model using 1,000 tokens for perfect code review beats a cheap model burning 50M tokens with errors—the full-system cost favors quality despite higher per-token rates.
  2. Expect rapid model specialization where enterprises build custom models via post-training on commodity open models for their high-volume specialized tasks, keeping them internal rather than distributing them.
  3. Open-source models will dominate commodity task execution while frontier models remain too expensive for most use cases. The opportunity lies in the middle tier: post-trained commodity models for specialized workflows.
  4. Current company structures and teams lack the infrastructure to execute post-training and model implementation at scale. Building internal AI capability requires organizational redesign, not just tool adoption.
  5. Frontier AI labs calling open-source models 'Chinese models' is a deliberate scarcity tactic to create fear. Ignore FUD around open-source—the market trend is toward accessibility and commoditization of AI intelligence.
  6. Only 10% of AI-focused labs survive long-term, making model differentiation through post-training and internal specialization critical for startups building on AI infrastructure.

Related topics

Transcript Excerpt

I see a world where the smartest model is actually the cheapest. People are thinking about outcomes in AI and they're looking at 20 30 50 and they're saying that's ludicrous. That's crazy. That is underestimating by an order of magnitude how massive a transformation this is going to be. 8 and 10 billion is the new 1 billion. Today we have CTO and co-founder of Factory Eno Reyes. He is one of the most articulate and insightful thinkers about the value stack of AI that I've interviewed. Factory is one of the leading companies that specialize in autonomous software development and ENO is incredible in this show today. There's going to be a lot of notes taken in this discussion. >> Two of the largest companies that provide models today have explicitly said we are going to go after every single…

More from 20VC