On May 29, 2025, at the European Technology Symposium held in Amsterdam, the Netherlands, TSMC once again made it clear that it will not use the next generation of high numerical aperture extreme ultraviolet lithography in its 1.6nm (A16) and 1.4nm (A14) process nodes. This statement continues TSMC's consistent prudent strategy, highlighting its technical path to break through the technical bottleneck by tapping the potential of existing low-NAU extreme ultraviolet lithography technology and continuously optimizing the manufacturing process.
The technical value of high numerical aperture extreme ultraviolet lithography and TSMC's considerations
High NAU ultraviolet lithography is a new generation of lithography technology that has attracted the most attention, with a resolution of about 8nm, which is significantly higher than the traditional low NAU ultraviolet lithography technology (13.5nm), and can theoretically support a single exposure in the process below 3nm, thereby reducing the complexity caused by multiple exposures. However, High-NA devices are not only expensive (more than $300 million per unit), but also extremely demanding in terms of process integration, mask control, flatness, etc., and only a few companies such as Intel plan to introduce them for the 14A process from 2027 to 2028.
Kaivin Chang, deputy co-COO of TSMC, said at the meeting: "We will only introduce high NA extreme ultraviolet lithography when it really brings clear and quantifiable advantages. At present, the A14 node has been able to achieve the target performance by optimizing the existing technology.”
He pointed out that TSMC will continue to rely on low numerical aperture extreme ultraviolet lithography technology, and continue to promote process scaling through multiple exposures, pattern self-alignment (and design rule innovation, etc.) without the immediate introduction of High-NA.
A14: "Full node" upgrade can be achieved without relying on new lithography
A14 is the next-generation advanced process node that TSMC plans to mass-produce in 2028, and its technical advantages are not only reflected in the process scale scaling, but also in the in-depth innovation of device structure and standard cell architecture:
The second-generation full-surround gate transistor (GAA): Through the optimization of the nanosheet structure, the current driving ability and gate control accuracy are further improved, and the power consumption can be reduced by 25~30% under the same performance, or the performance can be improved by 15% under the same power consumption compared with the previous generation process.
Next-generation standard cell design: Structural reconfiguration for different circuit types (logic, memory, analog) to increase overall transistor density without sacrificing yield. Up to 23% increase in logic circuit density.
Deeply explore the potential of low numerical aperture extreme ultraviolet lithography: Combine multiple exposures and computational lithography to break through the traditional resolution limit. Although Zhang did not disclose the specific technical implementation, he said that the team has found a way to complete the A14 node process without relying on High-NA.
It is worth noting that the subsequent version of A14 (which is expected to introduce the backside power supply network SPR in 2029) will also not use high numerical aperture extreme ultraviolet lithography, which further proves TSMC's confidence in the existing lithography system.
Figure: TSMC: The 1.4nm process does not require high numerical aperture extreme ultraviolet lithography (Source: ASML)
Different routes: TSMC and Intel's technology strategy diverges
The difference between Intel and TSMC in advanced process strategies is becoming more and more obvious. Intel prefers to simplify the process with high numerical aperture extreme ultraviolet lithography and reduce overlay errors caused by multiple exposures, and plans to introduce this technology at the 14A node. TSMC, on the other hand, attaches more importance to the maturity of existing equipment and the stability of the overall manufacturing ecology, and chooses to resolve the pressure of lithography through collaborative innovation of design and process.
This difference stems from the two companies' different business models: Intel, as an IDM vendor, prefers to streamline internal processes through the introduction of new devices; TSMC, as a pure foundry, needs to achieve the optimal balance between meeting customer needs, controlling costs and reducing technical risks. After all, high NAE extreme ultraviolet lithography not only means a huge investment in equipment, but also comes with systemic costs such as plant upgrades and supply chain restructuring.
The "best time" for High-NA has yet to come
Although there is no rush to introduce high numerical aperture extreme ultraviolet lithography at present, TSMC has not denied its long-term value. Zhang Kaiwen pointed out that High-NA will eventually become a part of the advanced manufacturing process, but "we must wait until the critical moment when technology, cost and customer needs are aligned."
The future of high NAV lithography will depend on a number of factors:
Approaching the physical limits of the manufacturing process: If the complexity and cost of multiple exposures continue to rise below 3nm, and the single exposure of High-NA can bring better solutions, the technology switch will be a matter of course.
Increased equipment maturity and reduced costs: ASML is expected to gradually reduce manufacturing and maintenance costs over the next few years as it plans to deliver the first High-NA devices in 2025.
Customer demand: Major customers (e.g., Apple, Qualcomm) require higher graphics accuracy for advanced packaging or heterogeneous integration, which may also push the technology adoption time forward.
Lithography Route Bifurcation: Progressive vs Subversive
TSMC's latest statement reflects its "gradualism" route in the context of the slowdown of Moore's Law: to achieve continuous improvement in performance and yield through in-depth optimization of the existing technology system, rather than rushing to embrace immature disruptive technologies. While this strategy may face yield challenges from multiple exposures, it avoids high investment and uncertain risks.
From the perspective of the industry pattern, Intel and TSMC are walking on two completely different paths: on the one hand, there are "subversives" who hope to quickly break through the limitations of the process through high numerical aperture extreme ultraviolet lithography, and on the other hand, there are "modest" people who rely on the gradual optimization of the existing system. In the future, who can achieve the best balance between performance, cost and large-scale mass production remains to be verified by the market.
