[Insights] Exploring Rapidus’ Path to 2nm Mass Production: 3 Key Challenges Ahead

---

---

Japan’s Rapidus has gained attention for its progress on 2nm chips, a market supposedly dominated by foundry giants like TSMC, Intel, and Samsung. According to Nikkei, the company plans trial production in April and aims to provide 2nm samples by June.

To fully understand Rapidus’ potential and clarify uncertainties, TechNews columnist Liang-Yang Lin, associate professor of National Kaohsiung Normal University, discusses the three key challenges the company must overcome for 2nm mass production.

Tough Early Challenge: Converting Prototypes into Products

According to a previous eeNews Europe report, Rapidus joined imec’s Core Partner Programme in 2023, following a Belgian economic delegation to Japan and a Memorandum of Collaboration in late 2022. The new foundry also announced plans to use IBM’s semiconductor expertise and 2nm process.

However, Rapidus should be mindful of the limitations of lab results, as highlighted in Lin’s column. Though IBM’s Thomas J. Watson Research Center is a renowned institution, it now focuses more on research, and the latest “Gate-All-Around” (GAA) process is still in the lab phase.

Similarly, imec, while skilled in semiconductors, is also a research organization with limited mass production experience. Lin points out that whether their technology can be commercialized will be a major challenge for Rapidus.

Challenges in 2nm Mass Production Undermined?

For now, FinFET (Fin Field-Effect Transistor) remains the mainstream, with major players like TSMC, Samsung, Intel, UMC, GlobalFoundries, and SMIC having extensive experience in it.

Then why does Rapidus bet on GAA? In Lin’s view, Rapidus sees GAA as the next-gen tech offering better performance and energy efficiency, particularly for AI and HPC. Additionally, with TSMC’s Japanese fab hitting the road, which features 22/28 nm and 16nm, investing in GAA may boost Rapidus’ chances of securing government subsidies.

However, as Japan’s semiconductor technology currently lags at 40nm, whether Rapidus, a company which has been around for just 2.5 years, could leap straight to 2nm process presents a significant challenge.

Notably, the new GAA process is more complex than the FinFET architecture, as it faces challenges like precision control, material defects, heat dissipation, leakage current, and material-interface issues, as highlighted in Lin’s article.

Unique Production Model: A Cure or Poison?

According to Henri Richard, Rapidus Design Solution’s first president, the company initially targeted AI chip startups as its main market, believing it could succeed without directly competing with TSMC.

In Lin’s article, he notes that due to limited EUV machines and scale, Rapidus’ so-called mass production differs from the mainstream, focusing on diversified, small-scale production for niche markets, setting it apart from other major players.

The rumor that U.S. semiconductor giant Broadcom might become a customer of Rapidus’ 2nm chips marks a promising start. Indeed, Rapidus’ success hinges on its leadership ensuring that the Rapidus-IBM-imec collaboration delivers high-quality, high-performance chips. Its small-scale, diversified production model could be key—if it works, it may create a miracle, as noted by Lin’s article.

Read more

• [News] Broadcom Reportedly Taps Rapidus for 2nm Chips, Samples Due by June
• [News] Foundry Giants’ Advanced Node Expansion Efforts Beyond 2025: TSMC, Intel, Rapidus and More

(Photo credit: Rapidus)