IBM Plants Flag in Hokkaido: Rapidus 2nm Mass Production Enters Countdown
In a landmark development for the global semiconductor supply chain, IBM has officially announced the establishment of a new office in Chitose, Hokkaido, dedicated to supporting Japanese chipmaker Rapidus in its ambitious quest to achieve 2-nanometer (2nm) mass production by 2027. This strategic move signals a deepening of US-Japan collaboration in advanced semiconductor manufacturing. IBM will provide core 2nm-class manufacturing expertise and oversee the architecture of semiconductor production systems at the Chitose plant, where prototype chip fabrication began in April 2025. The partnership represents far more than a simple technology transfer — it is a calculated bid to establish a new node of advanced chipmaking capability outside the traditional strongholds of Taiwan and South Korea.
IBM’s technological credentials in this arena are unimpeachable. The company unveiled the world’s first 2nm chip technology back in 2021, and its deep expertise in advanced node processes makes it an ideal partner for Rapidus. According to Hokkaido Shimbun, Rapidus CEO Atsuyoshi Koike has identified IBM and AI chip design startup Tenstorrent as the company’s primary prospective customers, indicating that the 2nm production line is being purpose-built for AI computing and high-performance computing (HPC) — the two fastest-growing segments in the semiconductor market today.
Beyond 2nm, IBM is already pushing the boundaries further. In collaboration with Synopsys (which acquired Ansys), the company has developed a groundbreaking 1.4nm node thermal modeling technology under DARPA funding. The machine learning tool, dubbed Thermonat, can predict chip thermal behavior at the atomic level with accuracy within 1°C, operating tens of thousands of times faster than conventional simulation tools. As transistor nodes shrink and AI-driven power densities soar, thermal management has become one of the most critical bottlenecks in chip design, making this breakthrough highly significant for the entire industry.
¥160 Billion and Counting: Japan Inc. Goes All-In on Semiconductor Revival
The funding story behind Rapidus is equally compelling. According to Japan’s Sankei Shimbun, total private investment in Rapidus has surpassed ¥160 billion (approximately $1.07 billion) as of early February 2026, significantly exceeding the company’s FY2025 target of ¥130 billion. This figure reflects a collective bet by Japan’s industrial establishment on semiconductor revival — Canon, Honda, Fujitsu, and Fujifilm Holdings have all increased their stakes, demonstrating unwavering confidence in Rapidus’s mission. For a startup foundry that is barely a few years old, this pace and scale of fundraising is virtually unprecedented in the global semiconductor industry.
The Japanese government’s role in the Rapidus project is equally significant. As a state-backed enterprise, Rapidus has already secured thousands of billions of yen in government subsidies. This dual injection of public and private capital reflects Japan’s determination to reclaim its voice in the global chip competition. Historically, Japan dominated global semiconductor manufacturing in the 1980s and 1990s but gradually ceded ground to Taiwan and South Korea. Now, with geopolitical tensions escalating and AI-driven demand for computing power exploding, Japan is seizing this strategic window to attempt a comeback in advanced process nodes.
On the capacity front, Rapidus has laid out an extraordinarily ambitious expansion roadmap: scaling monthly production from 6,000 wafer starts to 25,000 within just one year — a fourfold capacity leap. This ramp-up speed is virtually unheard of in the semiconductor industry, where new fabs typically take 2-3 years to go from pilot production to full capacity. Rapidus’s confidence stems largely from IBM’s mature manufacturing process technology and Japan’s traditional strengths in precision manufacturing equipment and materials.
TSMC Kumamoto 3nm Upgrade: Automotive and AI as Dual Growth Engines
Running parallel to Rapidus’s 2nm push, global foundry leader TSMC continues to escalate its strategic presence in Japan. According to DIGITIMES analysis, TSMC has decided to upgrade its second Kumamoto fab to 3nm technology, driven by long-term customer demand for automotive and AI chips rather than competitive pressure from Rapidus. This distinction reveals an important reality: different players in advanced semiconductor manufacturing are targeting distinctly different market segments. TSMC’s 3nm line will primarily serve automotive electronics and edge AI applications that demand the highest levels of performance and reliability.
TSMC’s decision to build 3nm capacity in Japan carries deep strategic significance. Japan hosts the world’s most mature automotive supply chain, with Toyota, Honda, and Nissan ranking among the largest consumers of automotive chips globally. As the automotive industry accelerates its transition toward electrification and autonomous driving, semiconductor content per vehicle is surging from approximately $500 in traditional ICE vehicles to over $1,500 in EVs. The Kumamoto upgrade will enable TSMC to supply cutting-edge automotive-grade chips to these Japanese automakers with dramatically reduced supply chain lead times.
From a global supply chain security perspective, the parallel advancement of TSMC and Rapidus in Japan carries profound implications. Japan is effectively building a complete advanced-node manufacturing ecosystem spanning 3nm to 2nm and beyond. If this strategy succeeds, Japan could become the third economy — after Taiwan and South Korea — to possess comprehensive advanced semiconductor manufacturing capabilities, fundamentally altering the currently concentrated global chip supply landscape.
SEMICON Korea 2026 Sets Records: Yield Becomes the Make-or-Break Factor
Shifting focus to the Korean Peninsula, SEMICON Korea 2026, held February 11-13 at Seoul’s COEX, drew equally intense attention from the global semiconductor industry. The exhibition set new records with 550 exhibiting companies, over 2,400 booths, and 75,000 pre-registered visitors — the largest edition in the event’s history. This scale reflects South Korea’s position as the world’s leading memory chip manufacturing nation and the semiconductor industry’s explosive growth momentum in the AI era.
The central theme running through this year’s exhibition was yield. As HBM4 (sixth-generation High Bandwidth Memory) and 2nm processes approach mass production, making chips properly has superseded simply making them as the decisive competitive factor. An equipment industry executive at the venue explained: “HBM requires stacking chips to 12 or even 16 layers, but the tiny voids and bonding defects that occur between layers cannot be detected by existing optical inspection methods. What was sufficient with sampling inspection in the past now requires full inspection.” This shift signals an explosive growth phase for the metrology and inspection (MI) equipment market.
According to Research Nester, the semiconductor inspection system market is projected to grow at over 8% CAGR from 2026 to 2035, reaching over $15.78 billion by 2035. The World Semiconductor Trade Statistics (WSTS) organization forecasts the global semiconductor market will grow over 25% year-over-year in 2026, reaching approximately $975 billion. In a market of this magnitude, every percentage point improvement in yield can translate to profit differences worth hundreds of billions of Korean won.
SK Hynix and Samsung’s AI R&D Revolution: From Man-Month to AI-First
Another major highlight of SEMICON Korea 2026 was the technology innovation roadmaps unveiled by Korea’s two memory giants, SK Hynix and Samsung Electronics. SK Hynix Senior Vice President of R&D Process Lee Sunghoon delivered a keynote sending a powerful signal: “The memory industry has reached a technology inflection point.” He noted that DRAM has entered sub-10nm nodes while NAND has transitioned from 2D to 3D architectures, with the rate of technical difficulty increase now outpacing historical growth curves. More critically, he predicted another technology inflection point beyond 2027, at which point traditional R&D methods will no longer suffice.
To address this challenge, Lee proposed AI-driven R&D innovation as the solution — shifting from the traditional “man-month” model that relies on adding headcount to an AI-first paradigm that maximizes time efficiency. “If we use AI models, we can evaluate a far broader range of materials in a much shorter time than conventional methods, while identifying optimal process conditions with fewer experiments,” he explained. SK Hynix also disclosed a concrete case study: collaboration with NVIDIA to optimize process simulation using AI, achieving significant improvements in R&D efficiency.
Samsung Electronics CTO Song Jaehyuk displayed equally strong technological confidence, announcing that HBM4 quality has reached the highest level and revealing that Samsung is preparing to manufacture next-generation form factors including Z-HBM — a novel 3D structure that goes beyond conventional vertical stacking. Once Z-HBM achieves mass production, it will further enhance the memory bandwidth and capacity required for AI training and inference workloads. In the HBM market, SK Hynix currently commands approximately 50% market share with Samsung close behind — together, the two Korean companies control over 90% of global HBM capacity, constituting one of the most critical supply chain nodes of the AI era.
The New Japan-Korea Semiconductor Landscape: Co-opetition Shaping the Global Supply Chain
Taking a comprehensive view, the Japan-Korea semiconductor supply chain in 2026 presents a picture brimming with transformation and opportunity. On the Japanese side, Rapidus is accelerating toward 2nm mass production with IBM’s full backing, TSMC’s Kumamoto fab is upgrading to 3nm for automotive and AI applications, and over ¥160 billion in private funding demonstrates the scale of Japan’s national commitment. On the Korean side, SK Hynix and Samsung continue to lead globally in HBM and advanced DRAM, SEMICON Korea’s record-breaking scale showcases the vitality of Korea’s semiconductor ecosystem, and the AI-driven R&D paradigm shift points toward the industry’s future direction.
From a supply chain perspective, Japan and Korea have formed a distinctive co-opetition dynamic in the semiconductor space. Japan dominates globally in advanced manufacturing equipment and critical materials such as photoresists, while Korea commands the high ground in memory chips and advanced packaging. The technological complementarity between the two nations far outweighs their direct competition. As the global semiconductor supply chain transitions from an efficiency-first to a security-and-resilience-first paradigm, the strategic value of the Japan-Korea semiconductor cluster will only increase.
Looking ahead, 2027 will be the critical year that tests the success of Japan-Korea semiconductor strategies. Whether Rapidus can deliver 2nm mass production on schedule, whether Korean companies can maintain yield leadership in HBM4 and 2nm processes, and whether AI-driven R&D models can truly be implemented at scale — the answers to these questions will profoundly shape the evolution of the global semiconductor supply chain. What is certain is that Japan-Korea semiconductor industries are entering an unprecedented period of rapid development, with implications radiating across automotive, AI, telecommunications, and virtually every modern industry sector.
Source: TrendForce | SK Hynix Newsroom | Digital Today
