TSMC roadmap; Intel scores Tesla
- TSMC published a process roadmap through 2029 adding A12, A13 and N2U while delaying A16 to 2027. - Intel reportedly landed Tesla as a customer for its 14A process at Tesla's Terafab AI hub in Austin. - TSMC is extending current toolsets instead of rushing High‑NA EUV, while Intel seeks credibility via marquee external customers. (tomshardware.com) (cnbctv18.com)
TSMC and Intel used the same week to make opposite pitches to chip designers: one stretched its roadmap, the other flashed a new customer. (pr.tsmc.com) (finance.yahoo.com) TSMC said on April 22 at its North America Technology Symposium in Santa Clara that A13 will enter production in 2029, A12 will also enter production in 2029, and N2U will start in 2028. A13 is a direct shrink of A14, which TSMC unveiled in April 2025 with production planned for 2028. (pr.tsmc.com 1) (pr.tsmc.com 2) A chip “process” is the recipe a foundry uses to print transistors on silicon, and each new recipe usually trades money, power use, and speed against one another. TSMC said A13 cuts area by 6% versus A14, while N2U offers 3% to 4% higher speed or 8% to 10% lower power than N2P with a 1.02x to 1.03x logic-density gain. (pr.tsmc.com) (tsmc.com) The roadmap also moved A16, a more aggressive node for artificial intelligence and high-performance computing chips, to 2027 volume production from an earlier 2026 target. Tom’s Hardware reported the slip on April 22, and TSMC’s current public roadmap now places A16 after N2P rather than beside it. (tomshardware.com) (tsmc.com) TSMC’s new nodes point to a split strategy. A13 and N2U are built to let customers reuse more of today’s design work, while A12 and A16 add backside power delivery, which feeds electricity from the back of the chip to free wiring space on top for signals. (pr.tsmc.com) (tomshardware.com) TSMC executives also said they do not need ASML’s pricier High-NA extreme ultraviolet tools for these nodes through 2029. Electronics Weekly reported deputy co-chief operating officer Kevin Zhang said TSMC can keep “harvest[ing] the benefit” from current extreme ultraviolet tools instead. (electronicsweekly.com) (trendforce.com) Intel’s news came from the customer side. Reuters reported on April 22 that Elon Musk said Tesla plans to use Intel’s 14A process for chips at the Terafab project in Austin, a move that would make Tesla the first major outside customer disclosed for that node. Intel declined to comment on Musk’s remarks, Reuters said. (finance.yahoo.com) Intel had signaled the opening months earlier. At its Direct Connect event in April 2025, the company said it had given early 14A design kits to lead customers and that multiple customers intended to build test chips on the process, which will use PowerDirect backside power delivery. (intc.com) Musk said in March that Tesla and SpaceX would build two chip factories at Terafab, and on April 22 he said the site could eventually produce 1 terawatt of computing capacity a year. Reuters said Bernstein estimated enough chip capacity for that level of compute would cost $5 trillion to $13 trillion in capital spending. (finance.yahoo.com) The contrast is now plain: TSMC is extending proven toolsets and adding more node variants, while Intel is still trying to prove its foundry can win and keep large outside accounts. The next test is whether those roadmaps turn into high-volume wafers on the dates both companies just put in front of customers. (pr.tsmc.com) (intc.com)