CVPR 2026 Compute Report: Accepted papers had 2x the GPU access of rejected ones
CSProfKGD · x · 2026-11-02
CVPR 2026's Program Committee released a Compute Report based on mandatory compute disclosures from 16,034 submitted papers, quantifying how GPU access relates to acceptance.
Key findings:
- Accepted papers reported a median of 4 GPUs vs 2 for rejected papers — a two-fold advantage
- Accepted papers logged 67% more development time (2,000 vs 1,200 median hours), despite similar 7-month project timelines
- Countries whose researchers deploy more GPUs per paper see consistently higher acceptance rates; GPU generation matters less than how intensively hardware is used
- Per capita, AI research productivity concentrates in small hubs: Singapore submitted 64.6 papers per million people, Hong Kong 53.2, vs just 6.6 for the US and 6.1 for China
Program Chair Vladimir Pavlovic said the goal was to let evidence speak on whether infrastructure drives AI research outcomes.
More from Infra
- Prime Mega MoE extends DeepSeek's kernel for training, up to 34% faster end-to-end MoE training — cfregly · 2026-11-02
- AI datacenters priced at ~$50B per gigawatt, but build speed is the metric that matters — JOBhakdi · 2026-11-02
- Reddit asks what happened to 'too cheap to meter' as OpenAI quietly cuts quotas — rolyataylor2 · 2026-11-02
- $1T+ US Mega-Project Pipeline: Texas Eyed as the Future Hub for AI Intelligence Deployment — NinaDSchick · 2026-11-02
- One RTX 3090 Ti runs a 552B DeepSeek MoE at 35 tok/s — full local LLM guide — uniVocity · 2026-11-02
- Musk says robotics needs more DRAM than Earth produces, yet Micron trades at 4x forward PE — sudoraohacker · 2026-11-02