A Japanese semiconductor startup founded four years ago says it will close the technology gap with the world's biggest chipmaker to six months. Backed by $19 billion in government funding and IBM's chip technology, Rapidus is Japan's most audacious bet on reclaiming its lost semiconductor glory.

Rapidus CTO's Bold Prediction: "Six Months Behind TSMC at 1nm"

Kazunari Ishimaru, CTO of Rapidus, recently made a striking claim in an interview with Nikkei xTECH: while the company currently trails TSMC by roughly two years in 2nm chip production, it plans to progressively close that gap, aiming for just a six-month lag at the 1nm generation.

This isn't mere bravado. Rapidus employs a distinctive "single-wafer processing" approach that handles silicon wafers one at a time rather than in large batches. This allows engineers to fine-tune each production step in real time, iterating faster than conventional fabs. While the Big Three chipmakers, TSMC, Samsung, and Intel, rely on high-volume batch processing optimized for massive output, Rapidus is betting that precision and speed of adjustment will let it leapfrog development cycles.

The Road to 2nm: A 2027 Deadline

Rapidus' immediate goal is to begin mass production of 2nm chips in the second half of 2027. Here's where things stand:

In April 2025, the company fired up its pilot line at the IIM-1 (Innovative Integration for Manufacturing) fab in Chitose, Hokkaido. By July, it had successfully fabricated prototype GAA (Gate-All-Around) transistors, the first time this next-generation architecture had been demonstrated in Japan. CEO Atsuyoshi Koike called it "an extraordinarily rare speed of achievement on a global scale."

From late 2026 into early 2027, Rapidus will run a shuttle service producing test chips designed by a handful of lead customers, who can then evaluate the resulting characteristics and start work on production designs. On the tooling side, a revised version of the current PDK0.5 (Process Design Kit) arrives around May to June 2026, and PDK1.0, which supports production chip design, is planned for late 2026 or early 2027. Mass production follows in the second half of 2027, at a target capacity of around 25,000 wafers per month.

For context: TSMC began 2nm mass production in late 2025, and Samsung and Intel entered around the same time. By the time Rapidus ships its first commercial 2nm chips, it will be roughly two years behind the leaders. Ishimaru's roadmap envisions closing this gap incrementally, at 1.4nm, then 1nm.

IBM Partnership: The Technology Backbone

Underpinning Rapidus' ambitions is its strategic partnership with IBM, which unveiled the world's first 2nm chip prototype back in 2021. Since formalizing their collaboration in December 2022, over 150 Rapidus engineers have trained at IBM's Albany NanoTech Complex in New York, mastering nanosheet transistor fabrication and multi-Vt (threshold voltage) technology. About 80 have since returned to Chitose to optimize the production process.

Mukesh Khare, General Manager of IBM's Semiconductor Division, has stationed approximately 10 IBM engineers at the Chitose fab full-time. He has also signaled IBM's commitment to collaboration beyond 2nm, eyeing sub-1nm technology that would represent the next frontier of chip miniaturization.

Rapidus has also forged partnerships with imec (a Belgian semiconductor research hub), Siemens, and California-based AI chip designer Tenstorrent, creating a multilayered technology ecosystem.

Hokkaido's Chitose: Japan's New Silicon Valley?

The IIM-1 fab in Chitose represents a remarkable transformation for this Hokkaido city best known for its proximity to New Chitose Airport. Groundbreaking took place in September 2023, and the facility now houses over 200 pieces of cutting-edge equipment, including ASML's EUV (Extreme Ultraviolet) lithography system, each costing upward of $300 million. This marked the first deployment of EUV technology in Japan.

Major equipment makers including ASML, Lam Research, and Applied Materials have established service centers nearby. ASML plans to quintuple its local maintenance staff. The vision is to create a semiconductor ecosystem around Chitose, attracting related industries, research institutions, and talent from around the world.

A second factory is planned for construction starting in 2027, targeting 1.4nm chip production by around 2029. Total investment for both facilities is expected to exceed several trillion yen.

$19 Billion and Counting: Japan's Unprecedented Government Backing

The scale of government support for Rapidus is without precedent in Japanese industrial policy. Japan's Ministry of Economy, Trade and Industry (METI) announced additional funding of roughly $6.5 billion for fiscal years 2026-2027, bringing cumulative government support to approximately ¥2.9 trillion (about $19 billion). This includes R&D subsidies, equity investment through the Information-technology Promotion Agency (IPA), and an innovative plan to convert government-funded factory equipment into Rapidus shares.

The support structure firmed up in February 2026, when the government took a ¥100 billion equity stake and 32 private companies invested a combined ¥167.6 billion ($1.1 billion). The investor roster reads like a who's who of Japanese industry: Toyota, Sony, NTT, SoftBank, Canon, Honda, Fujitsu, and Fujifilm, among others. Japan's three megabanks have expressed willingness to provide up to ¥2 trillion ($13 billion) in loans from 2027 onward, contingent on government-backed debt guarantees.

Rapidus' total investment plan, covering both 2nm and 1.4nm production, exceeds ¥7 trillion ($46 billion). The company aims to achieve operating profitability around 2030 and an IPO by fiscal year 2031.

How Rapidus Differs from TSMC

The fundamental difference lies in business model. TSMC dominates through massive scale, producing chips for Apple, NVIDIA, AMD, and Qualcomm with industry-leading yields and enormous volumes. It commands over 60% of the global foundry market.

Rapidus is pursuing a fundamentally different niche: small-volume, high-value custom chips with ultra-fast turnaround. Its RUMS (Rapid and Unified Manufacturing Service) model integrates design support, front-end manufacturing, and packaging into a single streamlined process. Where typical prototype cycles take two to three months, Rapidus targets just 12 days.

This approach appeals to AI startups, cloud computing companies, and defense-adjacent industries that need custom silicon quickly, not mass-market consumer chips.

On the technical side, early analysis suggests Rapidus' 2HP process achieves a transistor density of 237.31 million transistors per square millimeter, virtually matching TSMC's N2 (236.17 MTr/mm²) and significantly exceeding Intel's 18A (184.21 MTr/mm²).

Perhaps Rapidus' most compelling selling point is geopolitical diversification. With approximately 92% of the world's most advanced semiconductors currently manufactured in Taiwan, customers face significant supply chain risk. A cutting-edge fab in politically stable Japan offers an attractive alternative, especially as U.S.-China tensions continue to reshape the global chip landscape.

Challenges and the Ghost of Elpida

For all the optimism, significant hurdles remain. Rapidus announced at a July 2025 customer event that its 2nm transistor was working, but Ishimaru was blunt about what that meant: "The operating characteristics at that point were not satisfactory. It was a rush job, and neither the process technology nor the equipment was adequately prepared." That candor underscores how far the company still has to go on yield.

Talent recruitment is another concern. Advanced semiconductor engineers are in short supply worldwide, and Hokkaido's remote location adds an extra layer of difficulty. The company will need to attract top-tier global talent to compete with TSMC's deep bench of experienced engineers.

Japan also carries the painful memory of Elpida Memory, a government-backed semiconductor venture that collapsed in 2012 after receiving public funding. Critics worry that Rapidus could follow a similar path, leaving taxpayers on the hook for trillions of yen.

Yet the explosive growth of AI demand may be Rapidus' strongest tailwind. IBM's 2nm prototype demonstrated 45% better performance or 75% lower power consumption compared to 7nm chips, exactly the kind of gains that power-hungry AI data centers desperately need. Some analysts believe AI-driven demand will soon exceed existing foundry capacity worldwide, potentially creating the market opening Rapidus needs.

Japan held over 50% of the global semiconductor market in the 1980s before a dramatic decline. Rapidus represents the country's boldest attempt at a comeback. Whether Ishimaru's "six-month gap" becomes reality will be determined in the latter half of this decade, and the world will be watching.

In Japan, opinions on Rapidus range from patriotic pride to taxpayer anxiety. How does your country approach semiconductor self-sufficiency? Do you think government-backed mega-projects like this can succeed, or are they destined to fail? We'd love to hear your perspective.

References