# Compute Governance

Compute is a strategic resource; governing it shapes who builds AI. This sub-article expands [Section 3.2 of Philosophy](../philosophy.md), [Hardware & Software](../hardware-software.md), and [Recent Topics](../recent.md).

## Why Compute Matters
- Frontier AI requires massive, concentrated compute.
- Compute is measurable, trackable, and harder to hide than models or data.
- This makes it a **governance lever** for safety and national strategy.

## Instruments
- **Export controls** on advanced chips (e.g., NVIDIA restrictions to China).
- **Cloud licensing** and compute thresholds for training.
- **Energy reporting** and data-center siting rules.
- **Compute registries** to track large training runs.

## Debates
- Who should have access to frontier compute?
- How do we govern compute without concentrating power further?
- Environmental and equity implications.

## Why It Matters
Compute governance is central to the frontier-AI policy debate and to national-security discussions. See [Frontier AI](frontier-ai.md).

## Sources
- US Executive Order on AI (2023).
- Compute governance research (e.g., OpenAI, Epoch).

*Back to [Hardware & Software](../hardware-software.md) · See [Recent Topics](../recent.md).*