Taiwan Semiconductor Manufacturing Co revealed its latest chip production advances on Wednesday, demonstrating how the company can develop smaller and faster processors while avoiding the purchase of expensive new equipment from Dutch supplier ASML.
The global manufacturing leader, which produces semiconductors for major tech companies including Nvidia, Apple and Google, unveiled two enhanced production methods during its presentation in Santa Clara, California. The first technique, designated A13, is scheduled for commercial production in 2029 and will primarily serve artificial intelligence applications. The second method, called N2U, offers a more cost-effective approach for manufacturing chips used in smartphones, laptops, and AI systems.
Rather than investing in ASML’s newest “high NA” extreme-ultraviolet lithography machines, which cost approximately $400 million each—double the price of current models—TSMC plans to maximize the capabilities of its existing EUV equipment for all the technologies demonstrated Wednesday.
“This is where I think our R&D has done exceptionally well in terms of leveraging existing EUV technology while setting an aggressive technology scaling roadmap,” stated Kevin Zhang, deputy co-chief operations officer and senior vice president, in an interview with Reuters. “This is definitely a strength.”
While the improvements in chip size and speed are incremental, TSMC also presented ambitious plans for connecting complex AI processors, an area where industry experts anticipate companies like Nvidia will achieve the greatest performance improvements in upcoming years. Current AI systems such as Nvidia’s Vera Rubin, launching this year and manufactured by TSMC, incorporate two large processing units and eight high-bandwidth memory modules. By 2028, TSMC projects it will possess the capability to combine 10 large chips with 20 memory modules.
The famous Moore’s Law, named for Intel CEO Gordon Moore, forecasted that computing capability would approximately double every two years while becoming more affordable. Recently, some industry leaders including Nvidia CEO Jensen Huang have questioned whether this principle remains valid.
According to Dan Hutcheson, vice chair of TechInsights, TSMC is essentially extending Moore’s Law through its chip-connecting technology.
“Moore’s law is morphing from a monolithic, single die in a package to multi-die in a package,” Hutcheson explained in an interview. “And that allows the power and performance gains.”
However, combining multiple chips creates unique engineering obstacles. The processors generate heat during operation, and the various materials used in packaging expand at different rates, presenting new challenges for chip designers.
According to Ian Cutress, chief analyst at consultancy More Than Moore, large chip assemblies can warp and fracture, problems that affected Nvidia’s Rubin AI processor.
“(TSMC) aren’t addressing directly how they are solving those challenges,” Cutress noted.
