OpenAI's 'overcommitted' 2025 compute now looks like genius, can run inference near zero margin
illscience · x · 2026-09-27
Discussion of OpenAI's competitive room: it depends on inference margins and capacity — if willing to run hot at near-zero gross margin, they have plenty of room. Commenters also note that 'overcommitting' to compute supply in 2025 now looks prescient, while criticizing complicated product messaging and brand bifurcation (Codex, ChatGPT, Sora).
Related event: OpenAI's Product Sprawl and Compute Edge Spark Debate(2 posts)→
More from Infra
- Doubling to 4× RTX 5060 Ti 16GB for local MoE inference: is it worth it? — Pear_Virtual · 2026-11-16
- Analyst: Anthropic IPO won't save neoclouds — their business model was broken all along — RihardJarc · 2026-11-16
- Starlink replaces 1 Mbps VSAT across North Star's 36-vessel offshore fleet — XFreeze · 2026-11-16
- Oxide's second-gen Cloud Computer ships with Micron memory and storage; Micron invests — Sethwinterroth · 2026-11-16
- Training-free page-sparse decode in llama.cpp: 1.72× faster 128k decoding on an RTX 5090 — topppggg · 2026-11-16
- AMD MI455 beats Nvidia Rubin on raw specs, claims viral chip debate — zephyr_z9 · 2026-11-16