Eric Jang's three signals of AI overinvestment: compute repricing, open-source catch-up, robot cost halving
ericjang11 · x · 2026-10-15
After an investor event in NYC, Eric Jang shared three quantities he watches to judge AI overinvestment:
- Compute contract repricing: below-market compute contracts signed in 2024/25 will roll off and reprice higher. What refreshed H100/H200 cluster contracts look like in 2027 has big impact on depreciation and lifecycle assumptions, worsened further if not enough powered land comes online in 1-2 years.
- Speed of Chinese & open-source catch-up: Companies like Ant and OpenAI need to survive to keep the frontier going. It's not enviable — compute FLOPs must geometrically scale while margins may not, and you're one Zuck-style open-source raid away from losing the frontier.
- Doubling time of robot capabilities: Citing @robocurve, the cost for frontier LLMs to complete a robot subtask halved every 3 weeks, from $2.36 in July to $0.38 in September 2026.
More from AGI Musings
- A decade of degrees rendered useless: one person's take on skills devaluing — Paimaamu · 2026-10-15
- Fast Search Fundamentally Changes What an LLM Is, From Library to Live Interface — dioscuri · 2026-10-15
- Aiden Bai: performance engineering is now just a matter of tokens — aidenybai · 2026-10-15
- PMs should build more themselves: how agents collapsed weeks of design into hours — thechrisperry · 2026-10-15
- Memia #2026.41: AI agents scan 3,956 articles to curate 323 frontier signals — ben_r · 2026-10-15
- Memia #2026.40: Gemini 4 leaks, protein watermarking, AI red phone and open USD — ben_r · 2026-10-15