Beijing proposes export prohibitions and licensing on 40 technologies spanning artificial intelligence, photovoltaics and biomedicine, and imposes an immediate registration requirement. The photovoltaic list reaches directly into the process knowledge behind the American solar manufacturing buildout.
WASHINGTON, October 16, 2026 – China’s Ministry of Commerce and Ministry of Science and Technology published a draft revision to the country’s Catalogue of Technologies Prohibited or Restricted from Export on Friday, proposing to control forty technologies across artificial intelligence, photovoltaic manufacturing and biomedicine, and simultaneously imposing an interim registration requirement that took effect the same day.
The measure, issued as Joint Announcement No. 48 of 2026, covers twenty-three artificial intelligence technologies, twelve photovoltaic technologies and five biomedicine technologies. A ministry spokesperson said fourteen would be prohibited outright and twenty-six would be restricted, meaning they could be exported only with government approval and a license.
Public comments are due by November 15, 2026. The spokesperson called the move “necessary” and cited national security, adding that the measures do not target any particular country, region or company.
For American businesses, the significance is not in what the catalogue restricts from leaving China as merchandise. Finished polysilicon, wafers, cells and modules are untouched. The significance is in what it restricts from leaving China as knowledge, at precisely the moment when American industrial policy is attempting to rebuild a domestic photovoltaic supply chain that runs on Chinese process technology.
What takes effect immediately
The catalogue revision itself is a draft. The registration requirement is not.
From October 16, exporters must register planned transfers of listed technologies, including associated key equipment, through the Ministry of Commerce’s Technology Trade Management Information System before the transfer takes place. The requirement applies to exports that began before that date but are not yet complete, and it remains in force until the revised catalogue is formally issued.
Registration covers four transaction types: transfer of technology ownership overseas, licensing of domestically held technology for overseas use, a change in corporate structure that makes an overseas entity the owner or controller of the technology, and other cases the ministry designates. Filers must supply details on the exporter and the recipient, the destination, the technology and its intended use, and the estimated transaction value. Multiple technologies moving to a single recipient in one transaction may be filed together.
The legal basis cited is Articles 18 and 20 of China’s Foreign Trade Law, together with the Regulations on the Administration of Technology Import and Export.
Two carve-outs matter. Technologies already in the public domain through open-source releases or published literature are excluded. And for artificial intelligence specifically, the spokesperson indicated that application programming interface access, software-as-a-service delivery and ordinary application or web use fall outside the controls, while overseas local deployment of a restricted model does not.
The photovoltaic list, read closely
The twelve photovoltaic entries are drafted against technical thresholds rather than product categories, which is the characteristic signature of an export control regime designed to be durable.
On the prohibited side, the draft would bar export of production processes and equipment know-how for perovskite-silicon tandem cells using wafers of at least 182 millimeters at efficiencies of 30 percent or higher. It would bar slot-die coating technology for perovskites at coating widths of at least 1.1 meters. It would bar manufacturing technology for single-junction perovskite modules of at least one square meter at 18 percent efficiency or higher, and for tandem modules of at least one square meter at 28 percent or higher.
It would also bar two polysilicon production routes at scale: cold-hydrogenation fluidized-bed trichlorosilane processes at lines of at least 250,000 metric tons per year, and silane fluidized-bed processes at at least 20,000 metric tons per year. Wafer manufacturing processes producing wafers 70 micrometers or thinner would be prohibited, as would certain edge-passivation technologies for thin crystalline silicon cells.
On the restricted side, requiring license approval rather than facing an outright ban, the draft lists polysilicon plant engineering for lines of at least 30,000 metric tons per year, advanced crystal-pulling and wafer-cutting processes including slicing at 130 micrometers or thinner, and high-performance tunnel oxide passivated contact, back-contact and heterojunction cell processes, including certain p-type heterojunction technologies evaluated under AM0 illumination.
The AM0 reference is worth noting. AM0 is the air mass zero spectrum, which is to say the solar spectrum outside the atmosphere. Evaluating a terrestrial cell technology under AM0 conditions points at space applications, and the ministry’s stated rationale explicitly references photovoltaic technologies increasingly reaching aerospace and defense use.
The artificial intelligence list
The twenty-three artificial intelligence entries range from frontier model technology to agent architectures.
The draft would prohibit export of core technologies for large models trained on sensitive national data, and for artificial intelligence agents designed for sensitive settings including national defense and critical information infrastructure. Models with two trillion or more parameters, or with certain world-leading capabilities, could be restricted. The draft does not name specific models, though reporting has noted that Moonshot’s Kimi K3 and Alibaba’s Qwen 3.8-Max exceed the two trillion parameter threshold.
Certain chip design technologies, including for artificial intelligence processing and memory chips, are also covered.
The symmetry with American policy is deliberate and unmistakable. The United States and its partners have for several years restricted China’s access to the most advanced artificial intelligence accelerators and to the lithography equipment used to manufacture them. The draft catalogue applies the same instrument in the opposite direction, at the layer where China has an advantage.
Analysts quoted in coverage of the announcement read it as a structural rather than tactical move. Lian Jye Su of Omdia characterized it as a response to Western sovereignty requirements on artificial intelligence, robotics, telecommunications and other key technologies, aimed at regulating the terms on which foreign parties access Chinese technology rather than at cutting access off. Gary Ng of Natixis framed it in broader terms, observing that “The world is getting more fragmented in trade and investment.”
Why the solar provisions hit American policy hardest
The United States has spent four years building a domestic photovoltaic manufacturing base through a combination of tax credits, trade remedies and, most recently, Section 232 tariffs and minimum import prices on polysilicon and its derivatives.
That buildout has a dependency that policy has not yet addressed. The equipment that populates a modern cell line, the process recipes that make it yield, and the engineering services that commission it and debug it are overwhelmingly Chinese in origin. A tunnel oxide passivated contact line installed in Ohio or Texas typically runs on tooling sourced from Chinese suppliers, configured by Chinese process engineers, and improved over its first two years through continuing technical support from the same sources.
The draft catalogue reaches exactly that dependency. Finished cells and modules can still be exported. What becomes licensable, or in some cases prohibited, is the transfer of the process technology, the export of key equipment that embodies controlled technology, the establishment of overseas production lines and the provision of on-site technical support.
An analysis published by Shanghai Metals Market identified the ramp-up period as the central exposure. Plants outside China that own their technology independently and have established stable processes face limited near-term risk. Plants that depend on continuing Chinese technical support to raise yield and lower cost may see those improvements delayed. The analysis also made a point that is easy to miss: relocating a factory to another country does not by itself resolve a technology export requirement, because the control attaches to the technology transfer rather than to the geography of the plant.
That observation has direct consequences for the third-country assembly model that has absorbed American solar demand over the past two years. Cell capacity in Kenya, Nigeria, Ethiopia, the Philippines and South Korea was stood up quickly, and much of it was stood up with Chinese tooling and Chinese engineering. If licensing becomes a gating factor for continued technical support to those lines, the supply that American developers have shifted toward becomes less reliable at the same time that American trade remedies are closing off the alternatives.
The collision with December 4
The timing creates a specific compression.
On December 4, 2026, the American Section 232 regime on polysilicon and its derivatives takes effect, imposing an additional 15 percent ad valorem duty and minimum import prices of 21 dollars per kilogram for polysilicon, 100 dollars per kilogram for ingots and wafers, 22 cents per watt for cells and 38 cents per watt for modules. The policy logic is to make domestic and allied production competitive by putting a floor under import prices.
That logic assumes domestic and allied production can scale. The Chinese draft catalogue is, among other things, a statement that the inputs to that scaling are now subject to Beijing’s approval.
The two measures are not formally linked, and the Chinese ministry has said its action does not target any country. But the sequencing is legible. China’s photovoltaic industry has watched a succession of American trade actions close export markets for its product, most recently this week’s International Trade Commission determination on cells from India, Indonesia and Laos, which penalizes capacity widely understood to be Chinese-affiliated. The technology catalogue addresses the one asset that trade remedies cannot reach.
Precedent from rare earths and gallium
The most useful guide to how this regime will behave in practice is the record of China’s previous technology and materials control actions.
China has operated export licensing on gallium, germanium, antimony and graphite since 2023, and on a set of rare earth processing technologies since the end of 2023. Those measures followed a recognizable pattern. The initial announcement was broad and the implementing detail arrived later. Licensing throughput in the first quarter was low, with applications taking considerably longer than the published statutory window. Approvals, when granted, were end-user specific and conditioned on documentation about ultimate use. And the practical effect on trade volumes was front-loaded: shipments fell sharply in the first months as the system absorbed the backlog, then partially recovered as applicants learned what the ministry wanted to see.
Applying that pattern here suggests three expectations. First, the gap between the November 15 comment deadline and a final catalogue will not be short, and the interim registration requirement will govern in the meantime. Second, the first wave of license applications under the finalized catalogue will take longer than the regulations suggest. Third, the firms that fare best will be those that filed early, documented end use thoroughly and maintained a direct relationship with their Chinese counterparty’s compliance function rather than working through intermediaries.
There is one structural difference worth flagging. The materials controls regulate physical shipments, which are observable at a border. Technology controls regulate transfers of knowledge, which are not. Enforcement therefore falls more heavily on the Chinese exporter’s own compliance obligations and on the registration record, which means the practical burden lands on the Chinese counterparty even though the commercial consequence lands on the foreign recipient. Foreign firms accustomed to managing their own export compliance will find that they have limited visibility into, and no control over, the process that determines whether their project proceeds.
What is not covered
It is as important to be precise about the boundaries of this measure as about its reach, because overbroad readings produce unnecessary disruption.
The draft catalogue does not restrict exports of polysilicon, ingots, wafers, cells or modules as goods. Chinese photovoltaic product can continue to be shipped under the same terms as before, subject to whatever importing-country duties apply. An American buyer purchasing Chinese modules faces no new Chinese-side restriction, though it continues to face substantial American-side ones.
The catalogue does not restrict technologies that are genuinely in the public domain. Published academic literature and open-source releases are excluded by the terms of the announcement. A process described in a journal article is not controlled merely because a Chinese firm also practices it.
The catalogue does not reach technologies below the stated thresholds. A perovskite tandem process at 27 percent efficiency, to take the obvious example, sits outside the prohibition drafted at 30 percent. Whether that gap persists in the final version is a reasonable thing to comment on, but as drafted the thresholds are the control.
And for artificial intelligence, consumption through hosted interfaces is explicitly outside scope. An American company using a Chinese model through an API is not conducting a technology export.
What American businesses should do now
The immediate obligations fall on firms with existing or planned technology transactions involving Chinese counterparties.
Any American company licensing Chinese process technology, purchasing equipment that embodies controlled technology, retaining Chinese engineering services for line commissioning, or restructuring a joint venture in a way that shifts ownership or control of Chinese-developed technology to a foreign entity should assume the registration requirement applies from October 16 and should confirm with the Chinese counterparty that filing has been made. The requirement reaches transactions already in progress.
Companies with artificial intelligence deployments should distinguish carefully between access modes. Consuming a Chinese model through an interface or a hosted service is outside the controls as the ministry has described them. Deploying model weights locally on overseas infrastructure is not.
Technical thresholds should be checked against actual specifications rather than against general product categories. The controls are drafted to catch particular capability levels, and a line running below the stated thresholds may fall outside them entirely.
The comment period closing November 15 is a genuine opportunity. Foreign firms and industry associations can file, and the gap between a draft catalogue and a final one has historically been where scope gets narrowed.
Finally, procurement and project teams should build schedule contingency into any photovoltaic manufacturing project that depends on Chinese technology transfer or on-site support during 2027. The licensing process is new, the approval timelines are unknown, and the practical experience of other Chinese export control regimes, notably the rare earth and gallium and germanium licensing systems, suggests that initial processing is slow and that approvals are granted case by case.
Reading the policy signal
Beyond the compliance mechanics, the announcement carries information about how Beijing is positioning for the next phase of the trade relationship.
The United States and China agreed in late September to reduce tariffs on roughly 60 billion dollars of two-way trade, covering goods ranging from agricultural commodities to household items, and the two governments have since been working through the implementing lists. That truce addressed tariffs. It did not address technology, and the catalogue revision makes clear that Beijing regards the two tracks as separable.
That separation is itself a message. It says that tariff de-escalation is available and technology de-escalation is not, and it establishes a Chinese capability in the technology domain that can be adjusted up or down in future negotiations in the same way that American entity listings and license policies have been.
The choice of sectors reinforces the point. Artificial intelligence is the domain where American controls have been most aggressive. Photovoltaics is the domain where Chinese advantage is most complete. Biomedicine, the smallest of the three lists at five technologies covering cell and gene therapy, signals that the instrument is not confined to the two headline industries.
For American trade policy, the question this raises is whether the industrial strategy of the past four years accounted for the input side. Tariffs, price floors and tax credits can make domestic production commercially viable. They cannot, by themselves, supply the process engineering that makes a line yield. That gap was always visible to practitioners. It is now a matter of another government’s licensing discretion.
The wider pattern
This announcement belongs to a category that has grown steadily over the past three years: measures that are not tariffs, that are not captured in effective tariff rate calculations, and that nonetheless determine whether cross-border commerce in a sector happens at all.
Export controls, licensing regimes, origin substantiation requirements, forced labor import bans and investment screening now sit alongside duties as instruments of trade policy, and they interact in ways that single-instrument analysis misses. An American solar manufacturer in 2027 will be navigating Section 232 price floors on its inputs, antidumping and countervailing duty orders on competing imports, domestic content requirements attached to tax credits, forced labor documentation obligations on its polysilicon, and now Chinese licensing requirements on the technology that makes its line run.
None of those instruments is individually novel. The combination is, and the compliance function that can hold all of them simultaneously is still rare.
For now, the practical guidance is narrow and clear. Identify every point in the business where Chinese technology or Chinese technical personnel are load-bearing. Determine whether those points sit above or below the published thresholds. File comments if they sit above. And plan on the assumption that the final catalogue, when it issues, will look broadly like the draft.