Joules-to-GDP: across 44 occupations, highest-value-per-joule AI tasks aren't the highest-value ones
sprooos · x · 2026-11-05
Stanford researchers introduced Joules-to-GDP (J2G), a framework measuring how efficiently AI converts energy into economic value and identifying which tasks, models, and hardware yield the best return. It lands amid energy-limit concerns over whether AI investments will pay off; the authors argue for targeting the highest-value workloads and building models and chips that do more per joule.
Analyzing tasks across 44 occupations (finance, engineering, law, admin, etc.) from OpenAI's GDPval, three findings:
- The highest economic-value tasks aren't the highest value-per-joule tasks: CAD design earns 10% more per task than website building, but website building yields 27% more value per joule
- The most powerful model isn't always the best use of energy (details in the full report)
J2G offers a quantified lens on the AI-ROI debate, relevant to both compute economics and model selection.
Related event: Stanford's Joules-to-GDP Framework Quantifies AI Energy-Economic Efficiency(5 posts)→
More from Infra
- Mach 2 Medium sets new Pareto frontier for Qwen 3.8 compression: 1.5% loss at 26.7GB, beating Unsloth — pbaylies · 2026-11-05
- Redditor finds Tiel-Coder-35B MoE runs faster than Qwen 3.8 locally with 256K context — zxspectrun · 2026-11-05
- 63% of planned US data centers sit in states whose governors pledged bans or pauses — Catholic_Indian10 · 2026-11-05
- OpenAI's Stargate to consume up to 40% of global DRAM output via Samsung/SK hynix deal — TheZachMueller · 2026-11-05
- Mooncake author presents fleet-wide elastic KV cache across Cloudflare on Kastan Day — michellechen · 2026-11-05
- Baseten Router cuts coding-task costs 72% with only 0.8pt pass-rate drop vs Opus 5.5 — baseten · 2026-11-05