The Tiny Memory Cell Holding Up Japan’s Big Chip Bet: Why Rapidus Is Running Late
Japan’s most ambitious industrial project of the decade has run into a problem that is measured in atoms. Rapidus, the government-backed foundry meant to put Japan back at the leading edge of chipmaking, is behind schedule on its promise to mass-produce 2-nanometre logic chips by March 2028 and may miss that deadline, according to people familiar with the matter cited by Bloomberg and widely reported across Asian media on Tuesday.
The sticking point is not a headline-grabbing breakdown but a basic building block. Rapidus has yet to produce working static random-access memory (SRAM) cells — the small, fast memory embedded by the hundreds of millions in every modern processor. That milestone needed to be reached by this past summer to keep the schedule intact. Missing it leaves the company at least months behind, the people said, with engineers under heavy pressure to catch up. Keeping dust and particles under control inside the cleanroom is another unresolved issue.
Why SRAM is a litmus test
SRAM is often the first thing a new manufacturing process has to get right, and it is a demanding examiner. Each cell uses several transistors packed as tightly as the process allows, and every one of those hundreds of millions of cells must behave identically. At the 2-nanometre scale, a microscopic deviation in a single layer can be enough to make a cell unreliable — and a chip full of unreliable memory is useless. Foundries use SRAM test arrays precisely because they reveal, quickly and unforgivingly, whether a process is mature enough for customers.
Seen that way, the reported delay is less a single setback than a signal that the process itself is still being tuned. It also explains why the company’s visible progress since July last year, when it announced functioning transistors from its pilot line in Hokkaido, has mostly come in the form of process design kits for chip designers rather than demonstrations of complex working circuits.
The company’s answer
Rapidus has not conceded anything. A spokesperson said the company “continues to devote its full efforts toward the start of mass production in the second half of fiscal 2027” — that is, between October 2027 and March 2028 — and declined to discuss technical details, calling them confidential. The company also pointed to a September review by outside experts, commissioned by a government agency, which found development to be progressing. Last month, chairman Tetsuro Higashi said the project was on track, without elaborating.
There is also a quieter debate about what “mass production” should mean. An independent director told Bloomberg in September that the goal is a yield of at least 80 percent — four good chips out of five. Other executives have said that definition is not shared by everyone at the top. Where the bar is set will matter a great deal when the deadline arrives.
A lot of public money on the line
The stakes go far beyond one start-up. Tokyo has officially earmarked more than 2 trillion yen for Rapidus, founded in 2022 with the explicit aim of reducing Japan’s dependence on Taiwan for the most advanced chips. The company has also floated a stock market listing by 2032. A serious delay could test political patience with a subsidy programme that critics already viewed as a long shot, and could push that listing back.
The competition is not waiting. TSMC is expanding overseas, including in the United States and in Japan itself, while Samsung and Intel are racing for the same customers: AI companies hungry for ever more powerful processors. Every quarter Rapidus spends debugging its process is a quarter in which potential clients sign long-term deals elsewhere.
Strong in materials, still climbing in logic
The news landed one day after a very different chip story. On Monday, Mitsui Chemicals agreed to buy JSR — one of the world’s leading makers of photoresists, the light-sensitive chemicals used with extreme-ultraviolet lithography — for about 1.84 trillion yen, with backing from the state-owned Development Bank of Japan. That deal is part of a government-encouraged consolidation of Japan’s chip materials industry, where the country already dominates several critical niches.
Put side by side, the two stories sum up Japan’s position. In the chemicals, wafers and equipment that every fab needs, Japan is a heavyweight. In actually manufacturing the most advanced logic chips, it is trying to rebuild a capability it lost more than a decade ago — and discovering how unforgiving that climb is.
What to watch next
People familiar with the company say Rapidus will keep pursuing 2-nanometre production and may eventually succeed, even if years later than planned. The real signals will be concrete: a public demonstration of working SRAM arrays, the first named customers committing to production volumes, and how the government frames its next round of funding. Until then, Japan’s chip comeback rests on whether a few hundred million identical memory cells can be made to behave.