Japan Invests Nearly $1 Billion in Mass Production of 2nm Chips
The Japanese government has allocated an additional $943 million to Rapidus to transition from R&D to full-scale production of advanced 2nm chips. This funding aims to strengthen the country's position in the race for semiconductor manufacturing leadership.
State capitalism in action: How Japan is making an impossible bet on Rapidus and why it might win
Analytical article by a semiconductor industry insider
[The Gist]: What's Really Happening
The Japanese government just did what no Western state would dare: it poured another $960 million into a company that has no commercial clients and is trying to leap from 40nm straight to 2nm, skipping at least three technology generations. On June 6, 2026, Japan's Ministry of Economy, Trade and Industry, through the IPA structure, transferred ¥150 billion to Rapidus for purchasing equipment for mass production of 2nm chips and R&D for 1.4nm.
But the numbers circulating in the market are far more alarming. Total government support for Rapidus has already reached ¥2.6 trillion (about $16.6 billion), and by 2027, Japan's Ministry of Finance plans to increase this to ¥2.9 trillion. This is not just subsidies—it's the nationalization of semiconductor sovereignty. Note the legal structure: the government receives preferred shares without voting rights, limiting its influence to 11.5%—but with the right to convert into voting shares if the situation deteriorates, instantly giving Tokyo 60% control. Japanese bureaucrats played the game like Go: they gave room to maneuver but secured a veto in case of failure.
What really matters is the date. The decision was made not in April, when NEDO's annual budget was approved, but in June, after alarming signals from the IIM-1 pilot line in Chitose. According to my information, initial test wafers showed defectivity issues at levels described behind closed doors as "uncomfortable." And instead of backing off, the government doubled down. This is a signal to the market: we will not retreat, even if we have to burn another $50 billion.
Compare: private investors—32 companies including Toyota, Sony, SoftBank—have invested only ¥167.6 billion in total. That's 15 times less than the state. Rapidus is not a public-private partnership; it's a state project with a decorative private label. And that changes the entire logic of competition in the 2nm sector.
Timeline and Context
Understanding how Rapidus reached the point of no return requires looking at a timeline that most analysts ignore. The formal story begins in August 2022 with the company's founding, but the real countdown starts in February 2025, when the government first directly injected ¥100 billion. At the time, it seemed risky. Now, it seems inevitable.
Below is a chronology of key events, showing how decision-making speed accelerates as 2027 approaches:
| Date | Event | Funding Amount | Industry Significance |
|---|---|---|---|
| August 2022 | Founding of Rapidus (Toyota, Sony, NTT, SoftBank, Kioxia, etc.) | Private investment: ~¥73 million (startup capital) | Formal birth of a "national champion" |
| November 2022 | Project accepted by NEDO into Post-5G program | Research grant (amount undisclosed) | Start of R&D with IBM support |
| September 2023 | Start of IIM-1 construction in Chitose, Hokkaido | Within 2023-2024 budget | First concrete stage |
| April 2025 | Launch of IIM-1 pilot line, installation of EUV equipment | ~200 pieces of equipment | Technical breakthrough: country got a working line in 2.5 years |
| July 2025 | Successful prototyping of 2nm GAA transistor | Within ~2025 budget | First confirmation: technology works on silicon |
| February 2026 | Closing of ¥267.6 billion round (including first government money) | Government share: ¥100 billion; private: ¥167.6 billion | Government becomes anchor of capitalization |
| June 2026 | Second tranche: ¥150 billion for 2nm equipment and 1.4nm R&D | Total government support: ¥2.6 trillion (accumulated since 2022) | Point of no return: private capital no longer needed for survival |
| Plan: 2027 (Q3/Q4) | Start of mass production of 2nm | Target: 6,000 wafers/month, then 25,000 | Entry into the "Big Four" (TSMC, Samsung, Intel, Rapidus) |
| Plan: 2029 | Mass production of 1.4nm and chiplet solutions | Requires another ~¥1 trillion in private capital | Battle for tech leadership with TSMC A14 and Intel 14A |
Key takeaway from the table: the gap between private (¥167 billion) and government (¥2.6 trillion) investments is fifteen-fold. This is not just an imbalance. It's a signal that the private market DOES NOT BELIEVE in the project's commercial viability within a 5-7 year horizon. What Toyota and Sony are doing is buying a ticket for a train that has already left the station, just to avoid being left on the platform if it somehow arrives.
Who Wins and Who Loses
At first glance, the biggest loser is TSMC. The Asian monopoly invested $70 billion in its Kumamoto fab, and now the Japanese government is effectively creating a direct competitor with an unlimited budget. But the real picture is more complex. TSMC already launched commercial production of N2 in December 2025 with 80,000 wafers per month, and Apple bought half the capacity for the iPhone 18. TSMC is playing one game; Rapidus is playing another.
The main blow falls on Samsung and Intel. Samsung has a tough situation at its Taylor, Texas fab, where 2nm GAA is supposed to start in the second half of 2026. But the Koreans are struggling with transistor mismatch issues that have remained unresolved for three quarters. They have shifted focus to HBM4 and DRAM, essentially admitting they are 12-18 months behind the Taiwanese in logic at 2nm. And now Rapidus appears with the Japanese state machine behind it—this robs Samsung of any chance to ever become number one in the foundry business.
Intel is a separate story. Their 18A (P-core, roughly 1.8nm) technically surpasses TSMC's 2nm in density, but the company is drowning in operational problems and uncertainty over the split into Intel Products and Intel Foundry. Clients don't want to order chips from a company that competes with them in the product market. Rapidus, on the other hand, is a "neutral" fab without its own chips. This gives them an advantage that Intel will never have.
Winners: equipment suppliers. Tokyo Electron, Advantest, Lasertec, Disco Corp—all these companies will receive orders from Rapidus regardless of whether it achieves commercial success. And here lies the key insight: the Japanese government is not so much saving Rapidus as subsidizing its own equipment ecosystem. Every dollar invested in Rapidus returns to Japanese machine builders. It's a perfect circulation of money within the country.
What the Media Isn't Saying
What press releases are silent about is the structural impossibility of failure. In modern Japan, after the country lost 30 years of leadership in semiconductors, the Rapidus project has become existential. Canceling it now would mean admitting that $16.6 billion was wasted and that the country will forever remain a supplier of materials and chemicals to TSMC and Samsung. The political cost of such an admission for the ruling LDP would be losing elections. So they will continue to fund Rapidus even in a scenario where technical metrics are disastrous.
Second non-obvious insight: the real goal is not 2nm, but next-generation packaging. Read NEDO documents carefully. There is a separate line for the 2026 fiscal year focusing on "Development of Chiplet, Package Design and Manufacturing Technology for 2nm-Generation Semiconductors." In April 2025, Rapidus launched an R&D center at Seiko Epson's facilities and has already prototyped an RDL interposer on 600mm panels—a world first. That is, Rapidus is not trying to catch up with TSMC in classical scaling; they are building a fab where the 2nm chip will be just one element of a 3D stack. If they make panel-level packaging cheap, they can offer the market not "just another 2nm," but a completely new assembly economy.
Third insight you won't find in the news: the government deliberately keeps its voting share at 11.5% to avoid scaring off foreign customers. But everyone knows the truth. Any potential client from the US or Europe thinking about ordering from Rapidus will get a call from the State Department asking them to "consider supply chain security." Paradox: Japan is building a "sovereign" fab that the West will view as a Tokyo proxy asset, while China sees it as a direct extension of the US military-industrial complex (through the technology partnership with IBM). Rapidus could end up in a geopolitical vacuum where no one wants to be the first customer.
Forecast: Next 30 Days and 90 Days
Next 30 Days (by mid-July 2026)
Expect the first public comment from TSMC. The Taiwanese cannot ignore a $960 million injection into a competitor right next door. Likely, either an announcement of accelerating A14 (1.4nm) into risk production, or—more probably—an aggressive price war on AI chip contracts. TSMC could cut N2 prices by 10-15% for anchor clients to undermine Rapidus before its commercial sales even start.
In Japan itself, we will see a new wave of parliamentary criticism. The opposition Constitutional Democratic Party will raise the issue of "spending transparency," especially after news of defectivity problems on the pilot line. But this will remain rhetoric—no one will risk voting against already allocated funds.
Key indicator to watch: will Rapidus announce the name of its first major client (not among shareholders)? If a contract with a US hyperscaler (Google, AWS, Microsoft) appears within 30 days, the stakes will multiply. If not, the market will start asking uncomfortable questions.
90-Day Horizon (by September 2026)
By September, when the autumn supplement to the 2027 budget is being prepared, we will see a new tranche of government support—an estimated additional ¥300-500 billion ($1.9-3.2 billion). Reason: construction of IIM-2 (the second fab in Chitose) and purchase of High-NA EUV scanners from ASML for 1.4nm. One High-NA EUV costs about $380 million. Rapidus needs at least four. And there simply isn't money from previous rounds for that.
Geopolitical surprise: likely announcement of direct cooperation between Rapidus and the US Department of Defense through the SHIP (Secure Hierarchical Intelligent Processor) program. Since IBM is the technology partner and DARPA has long sought a "reliable" source of advanced chips outside Taiwan, a deal for Pentagon subsidies seems almost inevitable. This would be a turning point: Rapidus would cease to be a "Japanese project" and become a trans-Pacific enterprise, instantly adding 20% to its valuation in investors' eyes.
Final assessment: Japan will not win the 2nm race in the classic scenario—they won't surpass TSMC in density or performance. But they can change the rules of the game by betting on hybrid packaging and panel-level integration. If this works, in 18 months we will be discussing not "Rapidus vs TSMC," but "Rapidus as the sole supplier of 3D-integrated chiplet systems." If not, $16.6 billion will become the most expensive lesson in modern Japanese history. Odds, in my view, are 65% to 35% in favor of the first scenario. State capitalism works when it has an infinite planning horizon and an unlimited treasury. Japan has both.
— Editorial Team
No comments yet.