New US chip export rules now cover renting computing power, not just selling chips
The US widened its AI chip controls to include compute rental. Investors say global demand holds, but timing of hardware purchases may slip.
The short version
- The rules now reach compute leasing, not only chip sales.
- Overseas data centers serving Chinese firms with large compute need prior US permission.
- Five countries, including those in Southeast Asia, now need a license to import advanced chips.
- Investors expect delayed equipment purchases and shipments, not a drop in demand.
- Analysts keep a buy rating on the AI sector.
Why it matters
Access to computing power decides who can train and run AI models. These rules change where that power can be rented and who can rent it. Most readers will feel it later, as shipment delays, hardware availability or cloud access, rather than overnight.
What the new rules change
As UDN reported, the US Commerce Department's Bureau of Industry and Security (BIS) published new export rules for AI chips. Earlier controls focused on selling chips. The new ones also cover compute leasing, meaning renting out processing power in a data center.
An overseas data center that provides large amounts of computing power to Chinese companies must now get US permission first. Five countries, including those in Southeast Asia, must also get a license to import advanced chips.
Who feels it first
Analysts quoted by UDN say the rules mainly squeeze the overseas computing capacity China had planned to add. Chinese-made chip platforms can already handle several stages of building and running models. But they are held back by HBM (high-bandwidth memory, a fast memory type AI chips depend on) and by weaker software ecosystems, so they cannot quickly fill the gap.
GPU rental firms in Southeast Asia that serve Chinese customers could see fewer of their machines rented. Banks have started adding export-control clauses to loan talks, which may push back some equipment purchases. Taiwan's supply chain faces higher export-check costs and the risk of later delivery dates.
Why investors stay calm
The investors' argument is that demand for AI computing power still outruns supply. In their view the rules mostly change when hardware is bought, for example when a delivery location changes and a shipment slips.
They say they would only cut their outlook if major US cloud providers lowered their spending plans, equipment acceptance dragged on and payments weakened. Until then, they keep a buy rating on the sector.
Opinion versus measurement
"Demand is unaffected" is a forecast from analysts, not a measured result. The rules are new, and their real effect will show up in delivery schedules, rental rates and spending plans. Planning for some friction, without assuming a lot of it, is the sensible stance.
What to watch
- Whether major US cloud providers change their spending plans.
- Whether Southeast Asian GPU rental firms report lower rental rates.
- Whether Chinese chip platforms close the memory and software gaps.
Try it
If you depend on rented GPUs or a regional cloud, ask your provider how the new rules affect where your workloads run and whether delivery times may change.
Sources and links
- ReportingUDN: the article we sourced
First reported by UDN. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.