China boxed out of high-NA lithography race to 1nm chips

Taiwan’s TSMC is likely to be the second chip-making company after Intel to receive the semiconductor industry’s most advanced lithography tool as the race to 1-nanometer chips gathers pace.

According to a Nikkei Asia report, the world’s leading integrated circuit (IC) foundry will install ASML’s new high-NA extreme ultra-violet (EUV) lithography system at its R&D center in Hsinchu, Taiwan, by the end of this year.

That’s about three months later than industry sources had initially expected but not significantly later than America’s Intel, which installed its first high-NA lithography system at its R&D center in Oregon last April and its second in August.

Samsung Electronics will reportedly acquire its first high-NA system in early 2025. The Netherlands’ ASML has a monopoly on the EUV lithography systems that TSMC uses to make ICs at the 7nm and smaller process nodes.

High-NA lithography systems, also known as EXE systems, likewise use EUV light but employ a new optical system that increases the numerical aperture (NA) from 0.33 to 0.55. That reduces the critical dimension, or the smallest feature the system can print, by 1.7 times and increases the transistor density on a chip by 2.9 times.

Numerical aperture is “a measure of the ability of an optical system to collect and focus light,” in the words of ASML: “The higher NA is what gives the systems their better performance.”

According to reports, TSMC aims to start using high-NA lithography at the 1.4nm node (14A, A for angstrom) in 2028 or later, or at 1nm (10A) in 2030 or later. The 2nm process it plans to introduce by the end of 2025 will use currently available EUV lithography systems.

That is probably just as well since the development and evaluation work required to use high-NA EUV efficiently in commercial production will take a long time.

The technology is also very expensive, reportedly costing at least $350 million per system or roughly twice the price of standard EUV equipment. Presumably that figure will come down as unit demand rises.

While TSMC integrates high-NA EUV into its technology upgrade roadmap, Intel is using it as part of CEO Pat Gelsinger’s aggressive catch-up strategy, which to date has been dogged by disappointing yields, restructuring charges, lots of red ink and outsourcing to TSMC down to 3nm.

Intel’s share price has dropped more than 50% so far this year, while TSMC’s is up more than 70%.

Intel expects that “High NA EUV tools will play a critical role in advanced chip development and the production of next-generation processors. Intel Foundry — the industry’s first mover on High NA EUV — will be able to deliver never-before-seen precision and scalability in chip manufacturing, enabling the company to develop chips with the most innovative features and capabilities that are essential for driving advancements in AI and other emerging technologies.”

That is, if everything goes according to plan. Specifically, “Intel expects to use both 0.33NA EUV and 0.55NA EUV alongside other lithography processes in developing and manufacturing advanced chips, starting with product proof points on Intel 18A in 2025 and continuing into production of Intel 14A. Intel’s approach will optimize advanced process technology for cost and performance.”

It may, and Intel’s restructuring program is, by most accounts, making progress, but the words “or later” should probably be appended to its timeline.

As for China, at least for now, this is not a contest in which it can compete. US sanctions prevent it from buying EUV systems, and the previous generation ArF (Argon Flouride) immersion Deep Ultra-Violet (DUV) systems it has been able to acquire mean it is capped at 5nm.

Recent reports indicate that Chinese lithography equipment made by Shanghai Micro Electronics Equipment Co (SMEE) is commercially viable down to 65nm but that its attempt to develop an ArF immersion system capable of producing ICs at the 28nm process node is taking longer than originally planned.

In comparison, it took ASML six years to advance from the shipment of its first EUV system to shipping batches of high-volume production machines and 10 years of R&D before it was able to ship its first high-NA EUV lithography system.

(For a compact history of IC lithography, see “China’s lithography gains a glass half full, not half empty.)

It is possible that China might not have developed its own lithography equipment without the push of US sanctions. South Korea, home to Samsung Electronics and SK Hynix, two of the world’s leading producers of memory ICs, hasn’t developed its own lithography equipment. Japan’s Nikon, unable to compete with ASML, does not make EUV systems. Canon never even tried.

Sanctions have also given China an incentive to develop silicon photonics, a technology that combines silicon-based integrated circuits and optical components to process and transmit massive amounts of data without using EUV lithography.

Designers and manufacturers of ICs, AI systems and telecommunications equipment, including Nvidia, AMD, Intel, TSMC, IBM, Cisco Systems, NTT, Huawei and other Chinese companies and laboratories, have been working on the technology for many years.

This year, silicon photonics finally caught the attention of the US House Select Committee on the Strategic Competition Between the United States and the Chinese Communist Party (CCP).

On October 27, committee chairman John Moolenaar (Republican-Missouri) and Ranking Member Raja Krishnamoorthi (Democrat-Illinois) wrote to Secretary of Commerce Gina Raimondo to warn that photonics technology could allow China to overtake the US in semiconductors and asked her to take action to prevent it.

Quoting Matthew Reynolds, author of “Controlling Light: Is Silicon Photonics an Emerging Front in US-China Tech Competition?”, the congressmen wrote:

When combined with electronics in semiconductors, silicon photonic technology can ‘create large-scale computing systems with higher bandwidth and improved energy efficiency that go beyond the physical limitations of traditional electronic chips.’

Some experts believe photonic chips can offer a 1,000-fold improvement in computational speed compared to existing electronic chip designs.

Silicon photonics has the potential to upend the semiconductor industry and redefine battlelines in the United States’ technological competition with the PRC, rendering moot the October 7, 2022, export control rules and creating a critical chokepoint for future semiconductor supply chains.

The US government should examine the tools at its disposal—both preventing US investment and know-how from supporting our adversaries and bolstering domestic innovation—to ensure continued American leadership in critical and emerging technologies like silicon photonics.

The committee has done its homework, noting that photonics was listed in China’s 14th Five-Year Plan (2021–2025) as a technology for which national laboratories should be built and that CCP General Secretary Xi Jinping had called it “a high-tech industry in which our country has the conditions to achieve breakthroughs ahead of others.”

Chinese companies and research institutes are investing billions of dollars in silicon photonics. Furthermore, in defiance of US sanctions, “…researchers of the Nanjing Electronic Devices Institute determined that photonics is a disruptive technology with immense military potential. Huawei and Nanjing Electronic Devices Institute… are both listed on the Entity List with a presumption of denial license policy for unauthorized military end-use.”

Silicon photonic breakthroughs thus might give China a way to stay competitive in AI without the access to TSMC’s high-NA EUV lithography services enjoyed by Nvidia, AMD and other Western competitors. In any case, the US government once again seems to be behind the technological curve.

Follow this writer on X: @ScottFo83517667