US Eases AI Export Controls, Granting UAE License-Free Access to Advanced GPUs
pstAsiatech · x · 2026-07-11
The US Bureau of Industry and Security (BIS) recently moved the UAE to Country Group A:5, allowing approved UAE and US entities to access advanced GPUs without a specific license. This marks a major shift in US AI export control policy, substantially accelerating Middle East AI infrastructure development, though it is not a blanket deregulation.
Policy Details & Impact:
- Replaces old rules: This effectively replaces portions of the revoked Biden-era "AI Diffusion Rule." Instead of fixed national chip quotas, the new rules establish bilateral "trusted compute" corridors based on country and entity risk.
- Reduces uncertainty: Eliminating case-by-case transaction approvals makes it much easier to finance, procure, deploy, and scale large compute clusters—especially when the end-user is a US hyperscaler or an approved UAE operator.
- Strings attached: Entities like G42 and Core42 received special treatment under strict conditions—their authorizations will automatically expire after 270 days unless they become "US companies" or secure new BIS approvals.
- Ecosystem barriers: The current list of approved US entities includes Amazon, Apple, Google, Meta, Microsoft, OpenAI, Oracle, and xAI. This creates a priority lane for the US-led AI ecosystem rather than allowing unrestricted access to Nvidia and AMD chips for all UAE data centers.
Geopolitical Significance:
This move accelerates the restructuring of US-UAE AI relations. The UAE is being rewarded for its cooperation on export controls and technology security assurances, widening the gap between it and regional competitors like Saudi Arabia and Qatar, which have not received the same A:5 status. It also reflects G42's broader shift away from Chinese tech dependencies and towards deeper partnerships with US firms like Microsoft.
Related event: US Eases AI Chip Export Controls for the UAE(5 posts)→
More from Infra
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11
- Can a 7900 XTX 24GB run Qwen locally? Reddit seeks ROCm tok/s benchmarks — thenomadexplorerlife · 2026-09-11
- RTK Terminal Compression Cuts Tokens but Leaves Your AI Coding Bill Unchanged — Bartaseth · 2026-09-11
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11