Last week a Shanghai startup called Yuanjiwei switched on what it says is the world’s first 8-inch pilot production line for 2D semiconductors, announcing that it would use the line to build 5nm equivalent chips by 2029 without going anywhere near an EUV machine.
This is a bold promise, and whether it can actually accomplish this is a question nobody outside Yuanjiwei is currently able to answer.
Where Did Yuanjiwei Come From?
Yuanjiwei is not a startup that appeared from nowhere. The company was founded in February 2025 by Bao Wenzhong, a Fudan University researcher who ranks among the most cited names in 2D materials and whose lab built WUJI, the most complex 2D microprocessor ever published. That chip, a RISC-V processor made from atom-thick molybdenum disulphide, appeared in Nature last year, and Yuanjiwei exists to commercialise the work behind it.
Bao is refreshingly candid about how hard that work is, describing circuits built from atom-thick sheets as tofu work rather than jade work and telling the South China Morning Post that “carving the same object out of tofu is far harder” than cutting the same shape from stone.
The appeal of persisting anyway is obvious, because materials only a few atoms thick could in theory keep chips shrinking after silicon hits its physical floor, and they could do it without ever touching the EUV machines that only ASML makes and that Washington is working to keep out of China.
What Has Yuanjiwei Actually Built?
According to the company, a pilot line covering everything from 2D material preparation to chip integration, including tape-out capability, which would move the technology from lab demo to engineering validation and count as a legitimate milestone in its own right.
Then there is the gap. WUJI, the best chip this team has published, holds 5,931 transistors, while a commercial 5nm processor holds tens of billions, a distance of roughly six orders of magnitude that the phrase “5nm equivalent” is being asked to cover without any published benchmark for what equivalence means. Nothing else has been disclosed either, not yields, not funding, not customers, not independent verification. The announcement came via the company’s WeChat account, with a director from the Shanghai Municipal Commission of Science and Technology on hand to provide the supporting quote.
The Case for Taking It Seriously
Dismissing all of this as vapourware would be too easy, because the lab work underneath it is peer reviewed and genuinely impressive, including a reported 99.77% yield on WUJI’s inverter circuits achieved on university equipment.
The direction of travel is legitimate as well, since the entire industry rather than just China is looking at 2D materials as silicon approaches its limits. Bao is open in acknowledging that silicon chips carry millions of times more transistors while arguing that serious industrial investment would close the gap faster than anyone expects.
Why the Timing Matters More Than the Technology
Yuanjiwei is not announcing into a vacuum. Huawei unveiled its LogicFolding architecture in May with a claim of 1.4nm class chips by 2031 that analysts promptly picked apart, and SMIC’s actual 5nm class process reportedly sits in pilot with yields below 20%. China’s EUV workaround announcements are arriving on a political schedule rather than an engineering one, and each lands as evidence for the self-sufficiency drive whether or not a product ever ships.
Yuanjiwei is a credible lab with an uncredible deadline, and the thing to watch is not the next WeChat statement but the first named customer, the first disclosed yield figure or the first independent teardown.
See Also:
US Claims ASML EUV Machine in China: ASML Says Impossible
Why Is TSMC Refusing to Buy the World’s Most Expensive Chipmaking Machine?
