China Has Put Its Solar And AI Know-How Behind A Licence, And The Clock Started On Day One

China's Ministry of Commerce and Ministry of Science and Technology jointly issued Announcement No. 48 on 16 October, proposing revisions to the country's Catalogue of Technologies Prohibited or Restricted from Export.

The draft covers 40 technologies across artificial intelligence, photovoltaics and biomedicine. Fourteen are proposed for complete export prohibition and 26 for restricted export requiring state licensing. Twenty-three relate to AI and twelve to photovoltaics. Public consultation runs until 15 November.

The prohibited list includes production technologies for perovskite and tandem solar cell modules, AI-driven technology capable of generating and mutating malicious code, core large-model training technology involving sensitive national data, and proprietary platforms for cell and gene therapies. Thresholds apply to large-area modules, covering single-junction perovskite modules of at least one square metre with conversion efficiency of 18 per cent or higher, and perovskite tandem modules of the same area at 28 per cent or above.

The restricted category includes heterojunction solar cell technology, polysilicon plant engineering for production lines of at least 30,000 tonnes a year, and advanced crystal-pulling and wafer-cutting processes.

Separately, an interim pre-export registration requirement took effect on the day of publication. Until the revised catalogue is formally issued, exporters must register listed technologies through the Technology Trade Management Information System on the commerce ministry's platform before proceeding. The requirement covers intellectual property transfers, overseas licensing, changes in ownership or control involving foreign entities, and transactions already underway but not completed.

The measures target technology transfer rather than finished goods. Chinese-made solar panels, AI chips and biomedical products remain exportable. The commerce ministry cited national security, prevention of misuse and the need to keep AI under human control, and said China remained committed to an open-source AI ecosystem, with open-source projects and overseas chip fabrication eligible for facilitation and publicly available technologies generally exempt.

Two design choices in this draft do more work than the headline count of 40 technologies, and they point in opposite directions.

The first is the gap between the consultation and the registration. Comments are open until 15 November, which suggests a proposal still being shaped. But the interim registration requirement took effect on 16 October, immediately, and explicitly covers transactions that have started and not finished. Anyone mid-deal on polysilicon plant engineering or a wafer-cutting licence is already inside the regime, whatever the final catalogue says. The consultation is real for what gets listed permanently; it is irrelevant to anyone trying to close something this quarter.

The second is the open-source carve-out, which reads like a concession and functions as strategy. China is prohibiting proprietary transfer of core model training technology while stating that open-source projects get facilitation and publicly published work is generally exempt. The effect is to make publishing the path of least resistance and licensing the path of most. DeepSeek released a new model under an MIT licence the day after the announcement.

That combination spreads Chinese AI faster rather than slower, but on terms where nobody acquires transferable proprietary capability from a Chinese counterparty. Adoption is encouraged; dependence on Chinese open weights is encouraged; building an independent equivalent using Chinese process knowledge is not.

It is also a close structural mirror of what Washington has done to Beijing, applied to the layer where China actually leads. The United States restricted chips, the physical artefact China could not make. China is restricting process know-how, the thing others cannot replicate. Both are export controls. Only one of them is easy to enforce at a border.

For India this lands on a door that was opened from this side earlier in the year.

Delhi eased Press Note 3 restrictions in early 2026 specifically to fast-track Chinese investment in polysilicon, wafers and capital equipment, on the reasoning that domestic solar manufacturing cannot move upstream without Chinese process technology and plant engineering. That liberalisation only works if Beijing issues licences. Announcement No. 48 puts polysilicon plant engineering for lines above 30,000 tonnes a year, crystal-pulling and wafer-cutting into the restricted category, which is precisely the equipment and know-how that easing was meant to attract.

The result is a one-sided opening. India removed its barrier and China is raising one in the same lane.

The timing compounds a problem already running. Roughly 97 per cent of India's solar module exports go to the United States, where a 126 per cent countervailing duty has been under threat, and the official answer has been to move up the value chain and build domestic depth. Moving up the chain requires upstream technology that has just been put behind a licensing counter.

The near-term effect is on schedule rather than capability. Indian manufacturers with Chinese technology agreements in progress now sit inside the interim registration regime, which applies to incomplete transactions. Projects planned around a commissioning date assumed a transfer that no longer has a certain timeline.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.