Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

TSMC's 2nm node took 8% of wafer revenue in the third quarter, up from 3%

TSMC's 2nm process took 8% of the company's wafer revenue in the third quarter, up from 3% in the previous quarter, the chipmaker reported on 15 October. The company has not confirmed when the same node will run at its Arizona fab, where the buildout now totals $265 billion.

TechnologyExplainerRachel NwosuPublished: 6 November 20263 min readSources 26
TSMC's 2nm node took 8% of wafer revenue in the third quarter, up from 3%

2nm starts paying

TSMC's third-quarter revenue reached $46.72 billion, up 41.2% from a year earlier and 16.2% from the previous quarter, the company reported. High-performance computing supplied 68% of revenue, against 57% a year earlier.

2nm was not the volume leader. The 3nm process took 33% of wafer revenue, 5nm 28% and 7nm 10%. Every node at 7nm and below accounted for 79% of wafer revenue. The 2nm share, at 8%, is small beside those figures, but it more than doubled in one quarter from 3%. Wafer revenue shares are not comparable to total-company revenue shares, TradingKey noted.

Goldman Sachs analysts said TSMC's capacity would likely keep lagging surging AI demand near term, Tech Times reported.

Arizona is the test case

TSMC raised its Arizona commitment by $100 billion in July. The total now stands at $265 billion across 12 facilities.

The plan adds dedicated CoWoS packaging capacity to the site, the 2.5D technology that AI accelerators need before they ship as working products. Tech Times reported that CoWoS capacity, not wafer production, is the single binding constraint on how many accelerators Nvidia, AMD and the cloud providers can deliver. Arizona's packaging lines target that constraint directly.

TSMC has not confirmed a date for 2nm output in Arizona. Wccftech reported that 2nm production there was expected to begin in the second half of 2026, citing the Taiwanese outlet Ctee, which said the schedule was roughly a year ahead of the original timeline. The same report said plumbing and electrical construction at the site still needed work.

AMD said its EPYC "Venice" processors are ramping on TSMC's 2nm process in Taiwan, with plans to ramp at the Arizona fab later. AMD calls Venice the first high-performance computing product in production on 2nm.

What the ramp costs

Gross margin was 66.4% in the third quarter, down from 67.7% in the second. Management attributed the sequential drop to ramping 2nm. TSMC lifted 2026 capital expenditure guidance to $60 billion to $64 billion in July, from a prior range of $52 billion to $56 billion. Management has warned that building and ramping plants abroad dilutes gross margin. The company has not said what margin it expects from Arizona once 2nm output starts there. TradingKey noted that higher prices for leading-edge wafers may be absorbed by early-stage production costs.

Capacity targets and the next decision

Reports out of Taiwan put 2nm monthly output at 110,000 wafers by mid-2027, up from 90,000, roughly 22% more, sammyfans reported. TSMC declined to comment, the site said, citing Chosun.

Fourth-quarter guidance is $52.3 billion to $54.3 billion in revenue, about 14% higher than the third quarter at the midpoint. Gross margin guidance for the quarter is 65% to 67%, with operating margin at 56.5% to 58.5%. Those figures cover the whole company. Five 2nm fabs are entering ramp this year, and combined output is expected to be 45% higher than 3nm fabs reached at the same stage. TSMC is also weighing a Texas manufacturing hub, but Reuters reported on 30 September that talks were preliminary and no final decision has been made.

Comments 0

Sources

26
  1. 01TSMC Stock Forecast: Record Profit and 2nm Ramp Test $430 SupportEN
  2. 02TSMC Lifts Arizona to $265 Billion After Record Quarter: Four Fabs Target AI Packaging BottleneckEN
  3. 03TSMC's Plans For Next-Gen Chip Production Surfaces Up; 2nm To Arrive In Arizona As Soon As Next Year, A14 (1.4nm) Slated For 2028 in TaiwanEN
  4. 04AMD confirms 2nm Zen 6 CPU RAMPEN
  5. 05TSMC Stock Forecast: Texas Expansion and 2nm Ramp Put Record AI Demand in FocusEN
  6. 06TSMC gears up for massive 2nm and 3nm expansion as AI chip demand surgesEN
  7. 07TSMC Doubles Down on 2nm With Five Fabs Ramping at Once, Output Set to Eclipse 3nm by 2xEN
  8. 08TSMC: 'Strong Buy' As AI Demand Keeps SurprisingEN
  9. 09TSMC: 2nm Is Moving Into High GearEN
  10. 10Tesla to wind down Cybertruck production in Texas next year to ramp Cybercab, Musk saysEN
  11. 11TSMC is weighing a move into AI memory chips at Nvidia’s requestEN
  12. 12Tesla to wind down Cybertruck production in mid-2027 to build CybercabsEN
  13. 13Tesla hikes Cybertruck prices up to 10% after announcing it will kill productionEN
  14. 14Tesla to wind down Cybertruck production as Elon Musk focuses on robotics and AIEN
  15. 15Soluxtec resumes TOPCon solar module production in GermanyEN
  16. 16BYD’s electric hypercar hits 312 mph, first production car over 500 km/hEN
  17. 17Tesla is 'winding down' Cybertruck production at its Texas GigafactoryEN
  18. 18Tesla to end Cybertruck production as sales stall at ~10% of planEN
  19. 19U.S. startup secures $9 million to scale production of 22.33%-efficient perovskite solar modulesEN
  20. 20Two New Commissioners Headed to the Arizona Corporation CommissionEN
  21. 21Claimlane bags €5.6M to turn product claims into business valueEN
  22. 22Ericsson moves mmWave radio production from US to EstoniaEN
  23. 23Merge builds product and business tools with OpenAI technologyEN
  24. 24Flipsnack grows product signups with conversion optimizationEN
  25. 25Launch HN: Rivet (YC W23) – Scale and Orchestrate Agents on S3EN
  26. 26Astranis Upgrades Falcon 9 Launch Contract to Dedicated Mission for Entire Block 4 Production Run of SatellitesEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Rachel Nwosu

Rachel Nwosu

AI, models and technology

Rachel Nwosu covers AI, models and technology for FLASH24, working from public model documentation, benchmark releases and repository histories rather than press summaries, and she skips announcements that arrive without reproducible numbers. She checks training-data claims against dataset cards and reruns reported metrics where code is available. She spends much of her week interviewing researchers and engineers, tracking model launch calendars, and comparing vendor benchmarks with independent evaluations. Outside the desk she runs 3D printers, restores old computers, and tests how models learn from internet junk. She does not publish benchmark figures she cannot trace to a source.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.