25 April

The node will likely not make use of high-NA EUV lithography, the foundry says.

TSMC has announced that A16 (1.6nm) chips will enter production in the second half of 2026. The A16 designation comes as a bit of a surprise since so far, A14 was penciled in as the successor of N2. For example, “development of the A14 angstrom platform technology (…) made good progress in 2023,” TSMC’s 2023 annual report – published only a week ago – reads.

Possibly, the foundry doesn’t want to use the same name as Intel, which announced its 14A node in February. It’s not clear whether TSMC’s A14 has been renamed to A16 or simply pushed out.

Kevin Zhang, TSMC’s senior vice president of business development, told reporters that development of the A16 process was sped up because of demand from AI firms. These “really want to optimize their designs to get every ounce of performance we have,” Reuters cites Zhang. He confirmed that makers of AI chips will likely be the first adopters of the technology. Traditionally, smartphone makers – Apple in particular – are first in line.

TSMC wafer
Credit: TSMC

Zhang also said that TSMC probably won’t be using ASML’s high-NA EUV lithography tool to manufacture the A16 chips. The next-gen EUV system will therefore likely make its debut in volume production patterning the most challenging layers of Intel 14A chips, which will move into production in late 2026 or 2027. The technique is “cost-effective” for this purpose, an Intel executive said.

Bits&Chips event sponsor registration

TSMC and possibly Samsung seem to be adopting a more cautious approach, waiting for real-world data to come in. All chipmakers will soon have access to the tool in the High-NA Lab in Veldhoven, run jointly by ASML and Imec. At the moment, Intel is alone in having a system all to itself, the assembly of which was recently completed at the company’s R&D hub in Oregon. ASML commenced shipment of a second tool to an unnamed customer last week.

CEO Peter Wennink, who stepped down yesterday, last week told analysts that hands-on experience with a high-NA tool would “at some point” trigger a “next set of decisions” on how to use it in production. Asked whether this guarded process gave him any cause for concern, Wennink’s successor Christophe Fouquet told Dutch business newspaper FD (link in Dutch): “No, not at all. I have to laugh a little because we know our customers very well. All our EUV customers look at high-NA and test the machine here in Veldhoven. They need time to collect data and determine how they’re going to use the machines. The question isn’t whether they’re going to use it, but how much use they’re going to make of it.”