Inference is a permanent reverse auction: endless supply and easy switching keep grinding prices down
joecole · x · 2026-10-12
Jaya Gupta argues inference has become the second big funding story: outside the labs, inference companies can justify massive rounds by pointing at data centers and compute, even though the underlying product is trivially switchable. Many neolabs are also renting out their own compute, adding more supply.
Companies that verticalized around batch workloads and post-training are pivoting back to "pure inference," joining the same undercutting cycle — a permanent reverse auction where providers keep undercutting each other (half off Fireworks, half off Baseten, 10% cheaper GLM or Kimi).
Meanwhile app companies have enough choice to run bake-offs at will. Quoter dotpem says the related a16z chart is hugely important and few have internalized it.
More from Venture
- Indie maker shows Claude Code prompt that builds SEO free tools, driving 4M impressions — tibo_maker · 2026-10-12
- Indie dev's cat light app got cloned dozens of times, yet only the original hit #1 — AlchainHust · 2026-10-12
- Indie site's organic traffic climbs back to 450k after 10 months in the ditch — daluoseo · 2026-10-12
- Why China's AI model price war hurts labs but matters less for cloud providers — menhguin · 2026-10-12
- Moonshot reportedly in talks to settle $700M+ shareholder dispute ahead of IPO — Polymarket · 2026-10-12
- 4-person studio takes bank loan as free Steam demo's AI bill hits $1,000/day — aakashgupta · 2026-10-12