Advanced AI chips are not just a hot commodity for enterprises, but also a source of great geopolitical tension. After months of back and forth over export terms and restrictions, the White House announced this week that chips such as Nvidia’s top-of-the-line product, the Blackwell chip, would not be made available to China — and may soon be restricted from all non-U.S. companies as well.
It’s this second threat that could have serious repercussions for global enterprises, even those with U.S. headquarters.
In initial comments aired on CBS’ 60 Minutes program on Sunday Nov. 2, President Trump declared that “the most advanced [AI chips] — we will not let anybody have them other than the United States.” White House Press Secretary Karoline Leavitt also addressed the topic on Nov. 4, stating that the Blackwell chip and other advanced chips would not be sold to China at this time, although she did not comment on other global exports.
Specialized AI semiconductor chips — GPUs and TPUs — fuel the immense computing power required for today’s most advanced AI models , which are increasingly used at all layers of business operations and across nearly every industry. For global enterprises with operations across multiple geographical regions, inhibiting the use of AI chips overseas could lead to the following:
-
Complex compliance issues.
-
Two-tiered operational capacity.
-
Impeded performance in markets where the chips were formerly permitted.
If CIOs want to maintain performance levels across international jurisdictions, while adhering to new regulations, they will need to be ready to act — fast.
When AI is Confined by Geographical Boundaries
Restrictions over AI chip exports are not new, with the Biden administration first limiting certain exports through the U.S. Department of Commerce in September 2021. But the main target has always been China, with looser terms and loopholes for allied and neutral countries.
“Politically, the position has already been that the U.S. won’t sell China its most advanced chips — or even the second line for a while,” said Alexander Harrowell, principal analyst in advanced computing at Omdia. “Over the summer this became more like a ‘no-tier market’ when first the U.S., then China, restricted even the second-tier China-compliant products,” he said.
As Harrowell explained, there have been lobbying efforts to walk back those restrictions and reopen the Chinese market to American chip manufacturers; now it looks like instead of relaxed rules, the White House may be looking to expand those restrictions to cover more geographical territory.
The Impact of AI Chip Restrictions on Enterprises
To understand how the restriction of physical exports of chips can impact tech workloads overseas, the use case of the People’s Republic of China (PRC) is a helpful one.
“If you have facilities inside the PRC, this will definitely be a problem,” said Harrowell. “There aren’t many non-U.S. options, at least not that provide a genuine replacement without a lot of work.”
Baron Fung, industry analyst at Dell’Oro Group, concurs that companies operating in both the U.S. and a restricted region could face real conflict. This is because legal restrictions around “exports” don’t just apply to the initial selling of the product, but also where the product is then moved to and used. A U.S.-based company that purchases a Blackwell chip in California cannot then ship it to Beijing for use at a local Chinese base.
This is where two-tier capacity comes in, with workloads in restricted territories being forced to use reduced-capability AI chips that meet current export terms. A company with on-premises infrastructures in both America and China would not be able to use a uniform AI product across its business, unless it defaulted to the lower-capacity chip — which could cause broader harm to its ability to compete in the market.
Should the White House move forward with limits on any use of advanced AI chips outside of the U.S., global enterprises would face this challenge for all international operations.
Working Around, Not Against, AI Restrictions
CIOs looking to stay compliant and competitive within this context of legislative volatility have a few different options, with some offering short-term fixes and others requiring greater investment, but also greater long-term protection.
Two-Tier Operations
The first is to embrace the mixed-capacity model that has been used previously with the Chinese market, to varying success. While domestic operations use advanced chips like the Blackwell, international operations would leverage second-tier chips like H20s.
“A two-tier market already exists and has done for a while — this was the point of products like NVIDIA A800, H800, H20, L20, and AMD MI308,” Harrowell said.
However, he warns that this may not be practical going forward. When even the H20 was banned from export to China during the summer, no one operating outside of China really wanted to settle for Nvidia’s H20 stock — “Why would you, when you could buy the real thing?” Harrowell pointed out. Therefore, Nvidia and other advanced chip manufacturers have had less incentive to produce these second-tier chips, especially since the Blackwells command higher price points.
Fung described the production of two chip tiers as being a “significant risk” for chip manufacturers, since the rules are changing so rapidly and yet companies need to make these production decisions on much longer timelines. Even if they could find a buyer for these less competitive products, they cannot guarantee that they will legally be allowed to sell them.
“Developing a platform can take more than a year and, given the volatile nature of the current administration trade policies, there is high risk that a product that was compliant in the design stage may be non-compliant when ready for shipment, wasting R&D resources” Fung explained.
As a result, the supply of H20s and other second-tier chips may quickly dwindle as manufacturers divert resources only to first-in-class chips like Blackwells. This could leave CIOs floundering to find enough chips that meet export regulation, should the White House formalize new global restrictions on Blackwell exports.
Cloud-based AI Operations: A Long-Term Strategy
Even if a two-tier system is legally permitted, global operations could suffer if one region is forced to utilize weaker AI capabilities. For longer-term proofing, experts recommend that CIOs rethink the geography of their computing: By hosting their cloud architecture within the U.S., powered by advanced AI chips, companies can then extend those capabilities abroad without needing the same AI capabilities on-premises internationally.
The GPU clusters and model-training environments should be kept in America, while local or regional data centers overseas deploy smaller models for inference, personalization, or compliance with local data laws. This makes it simpler for CIOs to ensure regional compliance, without surrendering AI capability; global teams will still be able to legally access advanced AI capacity through the cloud. Where possible, Fung recommends partnering specifically with a cloud provider that has presence in both regions, to assist with compliance.
The challenge with this will be ensuring that this international access falls within the regulations, especially if changes are announced quickly without formal documentation.
“The rules issued by the U.S. Bureau of Industry and Security back in January 2025, in the last week of the Biden administration, provided very detailed rules about cross-border cloud provision, within-group transfers, trade in AI models, and the like,” said Harrowell. “But I have no idea to what extent those rules still have any validity.”
In the absence of clarity, CIOs must remain vigilant and agile when it comes to their AI workloads. This way, no matter what new geographical limitations are imposed, they can rise to meet the challenge.
