Many U.S. companies are experiencing long delays in getting export licenses to sell their goods in China. Most of these products, however, are already available in China. Furthermore, 82 percent of these firms say the products stuck in Commerce Department review already have a Chinese or other non-American supplier standing by to take the order anyway.
That finding sits inside a flash survey that the U.S.-China Business Council published earlier this month.” USCBC polled member firms in technology, industrial manufacturing, energy, and healthcare in July. Thanks to the Commerce Department, as USCBC points out, the United States is “ceding market share to foreign competitors while reducing the profits available for research and development.” The picture that emerges from the report is not a system straining to protect sensitive American technology from a rival power. It is a system straining, full stop, and the strain is landing overwhelmingly on exporters rather than on the goods that the Commerce Department is trying to keep out of Chinese hands.
South Korea already supplies the proof. In 2023, Washington granted Samsung and SK Hynix validated end-user status, letting them import American chipmaking equipment into their Chinese fabrication plants without a fresh license for every shipment. In 2025 it revoked that status ; the companies spent the rest of the year applying for individual annual licenses instead, a slower and revocable substitute that finally came through in December. By August, the Japan Times reports, both companies were quietly testing chipmaking tools from China’s Advanced Micro-Fabrication Equipment, a hedge against the next disruption and, more pointedly, against the risk that Washington stops supplying spare parts for the American machines already running in their Chinese factories.
Neither firm has decided to deploy the Chinese tools at scale. Both now have the option. That is what a slow, revocable license regime buys its own allies: not compliance but a second supply chain that doesn’t run through Washington.
Ninety-five percent of the companies in the USCBC report named lengthy license reviews as their top operational headache, ahead of tariffs, ahead of currency risk. Seventy-one percent reported delays specifically on China-bound licenses. Sixty-six percent had licenses pending past the 90-day statutory window that Commerce sets for itself. Thirty-one percent had applications sitting for one to two years, long enough for an entire product generation to turn over.
The costs show up in market share, not just paperwork. Seventy-three percent of respondents lost a deal outright to a Chinese competitor while their license sat in queue. Fifty-five percent lost orders to non-Chinese rivals able to ship faster from somewhere else. Sixty-four percent said they had lost standing inside the Chinese market specifically. More than a third put their losses at tens of millions of dollars or higher. Several told USCBC they are looking at hundreds of millions of dollars in losses. A few said billions.
The defenders of the licensing system argue that friction is the point. Slow the flow of sensitive goods, and China’s military and industrial base falls further behind. USCBC’s 82 percent figure guts that logic. USCBC represents the exporters as carrying the cost, so treat its numbers as a complaint, not a neutral audit. Sean Stein , the council’s president, put it more bluntly: export controls, he said, are “important, but if they are not calibrated, then they have the reverse effect,” undermining “U.S. competitiveness” and “U.S. technological leadership while doing nothing to protect national security.”
What is not in dispute is the paper trail behind it: if the customer can already buy the same part from a Chinese or third-country supplier the moment the American license gets denied or delayed, the review has not kept the technology out of China. It has only kept the American company out of the sale. What Washington calls a security tool is, for four out of five pending applications, a subsidy to Beijing’s own suppliers, paid for by the very firms the policy claims to be defending.
Last September, Commerce automatically extended Entity List restrictions to any company, anywhere, that is at least 50 percent owned by a blacklisted Chinese entity, with only a 60-day transitional license for allied-country dealings before full compliance landed. That catches joint ventures and subsidiaries across Asia as readily as it catches firms in Ohio. A month later, at the trade truce President Trump and Xi Jinping struck in South Korea on October 30, the rule went into a one-year suspension , off since November 10, 2025, one of the concessions that bought a year of calm in the tariff war.
It is due to snap back automatically on November 10 unless Commerce acts again. Compliance officers at exposed firms, in Seoul and Singapore as much as in Seattle, are right now building parallel screening protocols for a rule that may or may not exist in three months. That is not export control. That is hostage-taking with a scheduled release date.
Enforcement against real diversion, the part of the system meant to actually stop technology transfer, has been asleep. Commerce admitted on June 2 that it had failed for more than a year to enforce restrictions on high-end AI chips. Its own guidance now lets firms that bought controlled hardware without a license keep running it “until further notice.”
Nvidia’s H200, treated as too capable to authorize under the earlier framework, was cleared in principle in December, and by January Jensen Huang was reporting that Chinese demand was “very high.” That same month, Chinese regulators cleared ByteDance, Alibaba, and Tencent to buy more than 400,000 H200 units combined. American licenses started moving in February , in token amounts. By May, Washington had approved roughly 10 firms to receive the chip, and not a single one had shipped. The first deliveries did not land until August, when ByteDance and Tencent each took delivery of about 10,000 units apiece, a fraction of what both governments had already cleared seven months earlier.
The H200 is fabricated by TSMC in Taiwan and leans on HBM memory supplied by SK Hynix and Samsung. So, every month of enforcement drift or acceleration reshapes procurement decisions across three Asian supply chains at once, not just Nvidia’s balance sheet. Speed correlates with how useful a deal is to trade diplomacy, not with how sensitive the underlying technology is.
Beijing, for its part, is not waiting to see what happens with U.S. policy. It weaponized rare earth licensing in April 2025, approving barely a quarter of foreign applications by June before a London framework pried some of that back open. In June this year it added 10 more American firms to its own control list. China treats export licensing as an instrument it can tighten or loosen within weeks to extract concessions. The United States treats it as a queue that self-perpetuates regardless of the policy goal it was built to serve, and every extra week of drift is a week Seoul, Tokyo, and Taipei spend building the workaround.
November 10 is the date that will tell the region what Washington’s own system actually is. If the Commerce restrictions automatically renew because nobody got around to deciding otherwise, the USCBC diagnosis will be confirmed. The U.S. licensing regime is running on inertia, not judgment, costing American firms deals it cannot even claim credit for denying to China, while pushing allied chipmakers further down the road that Samsung and SK Hynix have already started walking.
If Commerce instead uses the next 11 weeks to separate the handful of licenses that matter for national security from the vast majority that do not, it would be the first sign someone sensible is in the driver’s seat. Every month it doesn’t, China’s major semiconductor manufacturers will get more customers they didn’t have to fight for.
The post America’s China Licensing Regime Now Helps Beijing appeared first on Foreign Policy In Focus .
America’s China Licensing Regime Now Helps Beijing
Aggregated summary from an independent source. Read the original at FPIP.