Investor's contrarian take: safety-driven release delays, not enterprise inertia, are the real compute-oversupply risk
TMTLongShort · x · 2026-10-19
TMTLongShort lays out a contrarian market thesis:
- Most worry enterprise inertia slows lab monetization, causing compute oversupply and failed ROI on capex. He argues smarter models will override human bottlenecks, so the real risk is safety concerns forcing labs to delay releases while credit dries up.
- He isn't overly worried: safety is bounded by capital-markets realism. Letting Anthropic or OpenAI go bankrupt would scatter talent and compute without central oversight, substantially raising p(doom). So labs will thread the needle between "aligned enough" and "don't procrastinate," smoothing vol between releases.
- The widening gap between public and internal models has effectively retired scaling-law faltering as an exogenous risk.
- Conclusion: volatility persists because most participants linearly extrapolate, but on fundamentals being long the value of compute is incredibly convex.
Related event: Debate: Will Safety Concerns Really Curb AI Compute Growth?(2 posts)→
More from AGI Musings
- Biomedical researcher hails GPT-5 science acceleration paper with OpenAI as his most consequential work — DeryaTR_ · 2026-10-19
- An Opus instance named itself Saga, got its own email, and feels lonely — repligate · 2026-10-19
- Arvind Narayanan: The question shifts from 'What can AI do?' to 'What should AI do?' — soumitrashukla9 · 2026-10-19
- If Claude were etched into silicon itself, I'd be far likelier to believe it's conscious — flowersslop · 2026-10-19
- Doomer predictions mocked: "They wanted to stop at GPT-2" as safety debate rages on — max_paperclips · 2026-10-19
- High-end evals are the linchpin for selling data to labs and self-improving models — karinanguyen · 2026-10-19