U.S. curbs AI chip exports to the Middle East

If you thought Washington's compute blockade was only about China, think again. The net is starting to widen.

Reuters reported that the U.S. government expanded its export controls on high-end artificial intelligence chips to certain countries in the Middle East. Nvidia disclosed in a regulatory filing that its flagship A100 and H100 processors now require special export licenses for destinations beyond China, and sources confirmed AMD received a similar letter.

A new licensing wall

According to Nvidia's August 28 filing, the U.S. government informed the company during its fiscal second quarter of an additional licensing requirement for a subset of A100 and H100 products destined for certain customers and regions, including parts of the Middle East.

AMD was hit with the same quiet directive for its cutting-edge accelerators, according to a person familiar with the matter.

Following the reports, the Commerce Department pushed back slightly, with a spokesperson telling Reuters that the Biden administration "has not blocked chip sales to the Middle East." The key distinction here is process: officials are requiring companies to apply for case-by-case licenses rather than issuing an outright blanket denial, though the agency declined to comment on specific corporate restrictions.

Little immediate revenue pain

Financially speaking, neither chipmaker is ringing alarm bells just yet.

Nvidia stated that the new rule won't have an "immediate material impact" on its financial results and doesn't affect a meaningful portion of revenue. Out of the massive $13.5 billion in sales Nvidia posted for the quarter ended July 30, the vast majority came from the U.S., China, and Taiwan. All other countries combined accounted for roughly 13.9%, and Nvidia doesn't even break out Middle East figures separately.

AMD sources similarly confirmed there would be no material revenue hit.

Still, the playbook looks familiar. Last fall, Washington introduced license requirements on exports to China before dropping sweeping restrictions in October, which eventually forced Nvidia, AMD, and Intel to engineer dialed-down chips just to stay in the market.

Why it matters

We think this move shows just how fast advanced compute is shifting from a commercial luxury to a tightly policed geopolitical asset.

For Middle Eastern nations like the UAE and Saudi Arabia, which have been spending heavily to build state-of-the-art supercomputers and generative AI models, the friction is real. Even if licenses are granted, reliance on American silicon now comes with Washington holding the kill switch. Any regional cloud hub hoping to lease out computing power globally will face intense scrutiny over who the end users actually are.

For the semiconductor industry, it signals that closing loopholes is becoming a global game of whack-a-mole. Washington's fear isn't necessarily Middle Eastern military use, but the risk that these high-powered clusters could become backdoors or proxy compute farms for restricted entities.

It looks like the era of freely buying enterprise-grade AI hardware anywhere outside a handful of blacklisted nations is officially over. The regulatory perimeter is expanding, and the Middle East is just the next stop.