In July 2025, at the Rapidus facility in Chitose (Hokkaido), the first test wafers were manufactured using 2 nm technology. Mass production has not yet begun — it is scheduled for 2027. However, the Japanese government has already invested the equivalent of $16.3 billion in the project and shows no signs of stopping.
What Japan is buying with this money
Rapidus is not an ordinary startup. It was founded in 2022 by eight corporations: Toyota, SoftBank, Sony, NEC, Kioxia, Mitsubishi UFJ Bank and others, with initial capital of ¥7.3 billion. The state entered as the primary investor — and retained the right to convert non-preferred shares into ordinary shares, effectively gaining a mechanism for forced control over the company if something goes wrong.
"Since the beginning of the year, customer demand for advanced chips has increased sharply"
— CEO Rapidus Atsuyoshi Koike, February 2026
New subsidies of ¥631.5 billion ($4 billion) are earmarked specifically for Fujitsu — the first confirmed customer. Fujitsu, in turn, is considering an investment of up to ¥20 billion in Rapidus itself. In other words, Japan is financing a company that manufactures chips for a customer who is itself its shareholder.
The technology gap that must be bridged
Today, Japan's most advanced domestic capacity is 40 nm. TSMC already produces 2 nm at industrial scale and is developing 1.6–1.4 nm. Rapidus uses High-NA EUV equipment from ASML and technical support from IBM — approximately 10 IBM engineers work constantly at the Chitose facility. After 2 nm, the goal is 1.4 nm, and then 1 nm. The total cost of this path is estimated to exceed ¥7 trillion.
- The state has already allocated: ~¥2.9 trillion
- Needs to attract from banks and private investors: ~¥4 trillion
- Private sector in the latest round: ¥130 billion — from 32 companies, including Sony, Toyota, Canon, and Seiko Epson
Why private business remains cautious
Despite a growing list of shareholders, private investors remain significantly more cautious than the state — due to technological uncertainty and the absence of a confirmed order portfolio. According to Digitimes, regional banks are only "coordinating possible investments," and the final figures of participants may still change. Prime Minister Shigeru Ishiba publicly acknowledged that taxpayer spending requires a clear strategy — but emphasized that falling behind in the global semiconductor race would be irreversible.
In parallel, Rapidus is in negotiations with over 60 potential customers in AI, robotics, and edge computing, including American startup Tenstorrent, which develops RISC-V processors as an alternative to Nvidia.
The bet is made, but not confirmed
Japan lost its leadership in chips in the 1990s, when Samsung and TSMC surpassed NEC, Hitachi, and Mitsubishi Electric. Rapidus is an attempt at a comeback not through catch-up, but through a leap directly to the cutting edge. There is a precedent: TSMC's involvement in Kumamoto showed that state subsidies can change industrial geography.
But TSMC came with established production. Rapidus starts from scratch — with first test wafers, an ambitious 2027 deadline, and a state that legally can take control of the company if it fails to deliver.
If by the end of 2026 Rapidus does not sign contracts with several customers independent of the state — the investment risks becoming the most expensive industrial experiment in Japanese history without commercial results.
