Open-source models already power most production AI workloads, investor says
RihardJarc · x · 2026-07-24
Rihard Jarc argues that most real-world production AI workloads already run on non-SOTA models, and that open-source models have been essential to that reality.
- He says restricting access to open-source models would be a mistake and could leave the market effectively controlled by Anthropic and OpenAI.
- In his view, long-term value would shift away from data centers and semis toward the model layer, because if only two major buyers of compute remain, even chip vendors lose leverage.
- The quoted thread he responds to makes the complementary point that once AI quality is good enough for a use case, marginal frontier gains do not always justify switching to the most advanced model in production.
Related event: Investors Warn Restricting Open-Source AI Risks Oligopoly(2 posts)→
More from Infra
- Analyst: Ant has the most fungible AI workload of any model provider today — BenBajarin · 2026-10-26
- Anthropic S-1: training infra edge cut per-task cost 40% and doubled revenue growth — BenBajarin · 2026-10-26
- Rewriting scaling law study: optimal synthetic data share rises from 10% to 32% with compute — heghbalz · 2026-10-26
- US states halt new data centers as 72% of Americans oppose local builds — YvesMulkers · 2026-10-26
- $5M Seed Round Can't Afford One DeepSeek Training Run: GPU-less AI Startup's Three Paths — FinanceYF5 · 2026-10-26
- HBM demand shifts: low-stack inference plans shelved, Rubin to go 12-high only — BenBajarin · 2026-10-26