The Compute Compact
Compute is the one input to advanced AI that is physical, measurable, and hard to hide — the sharpest oversight lever, or the deepest source of concentration.
Listen
The Compute Compact: Physicality as a Lever for AI Governance
A ~13-minute audio interview about the report with Dr. Krzysztof Pietroszek, President of FAIR Labs. Produced with www.allais.com.
Summary
You cannot hide a gigawatt. That makes compute the best lever we have — if the access fight doesn't decide everything first. Every other input to advanced AI — data, algorithms, talent — is easy to move, copy, or hide. Compute is not: chips are manufactured in countable facilities, clusters draw metered power, and training runs leave a physical footprint. That makes compute the most promising substrate for governance in Volume 04's toolkit. It is also the most concentrated layer of the entire AI stack, which means the fight over who gets access to it will shape who gets a say in everything downstream.
The seventh volume of the Frontier Series treats compute — chips, clusters, energy, interconnect — as AI's most governable input and its most concentrated one. The report builds the case for compute-indexed thresholds as the backbone of Volume 04's governance ladder, confronts the security logic and limits of export controls, and makes the affirmative case for a publicly funded compute commons so that oversight capacity does not live only inside the firms being overseen. It closes with a decoded map of the compute stack for readers who have never had to think about a wafer fab before.
How to cite this report
The Compute Compact: Compute is the one input to advanced AI that is physical, measurable, and hard to hide — which makes it either the sharpest oversight lever or the deepest source of concentration. FAIR Labs Frontier Series, Vol. 07. Fair Artificial Intelligence Research Labs, 2026. Available at https://fairlabs.ai/research/frontier-vol07-the-compute-compact
FAIR Labs (Fair Artificial Intelligence Research Labs) is a nonpartisan 501(c)(3) research institute in Washington, DC. Its reports may be reproduced with attribution for non-commercial purposes. For interviews, briefings or data requests, use the contact form.