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

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

Search
LIVE
›

Faraday wins US cloud AI chip project for Samsung 2nm production

Faraday won on 27 October a US cloud AI chip project that Samsung is set to make on 2nm, Chosunbiz reported, worth tens of millions of dollars with first revenue expected in the first half of 2027.

TechnologyAnalysisGrace OkonkwoPublished: 6 November 20264 min readSources 31
Faraday wins US cloud AI chip project for Samsung 2nm production

Faraday, listed on Taiwan's stock exchange, won the design project and is registered as a design solution partner for Samsung Foundry. Samsung will handle wafer production, and Faraday is also working with partners including Taiwan's UMC, according to Chosunbiz.

The customer originally planned to use a 4nm node, then switched to Samsung's 2nm gate-all-around process after weighing performance, power efficiency and product lifespan, according to the briefing. The chip is expected to use 3D custom DRAM and hybrid bonding technology. Faraday did not disclose the size of the non-recurring engineering contract, the value or volume of wafers Samsung will produce for mass production, or the customer's name.

Faraday said the project is its largest ever by design complexity, NRE contract value and advanced packaging scale. Its third-quarter sales were NT$3.664 billion, up 13% from a year earlier.

Samsung's 2nm pitch

On its third-quarter earnings call on 29 October, Samsung said 2nm orders this year were more than double last year's level and included major cloud and AI customers. Yonhap News reported that Samsung secured additional 2nm projects from cloud service providers and AI and HPC customers in the third quarter, building on wins in the previous quarter.

Kang Suk-chae, the executive vice president in charge of the foundry business, said demand across processes of 8nm and below would far exceed production capacity in 2027, pushing utilisation above 100%.

Samsung expects HPC to account for more than half of foundry revenue in 2027. It plans to expand capacity across its 8nm, 4nm and 2nm processes to meet that demand.

The utilisation figure is a forecast, not an output number. Samsung did not explain the baseline used to calculate 100%, or how much planned capacity expansion is factored in. It also did not disclose actual project counts for 2025 or 2026, so the doubling is a rate without a base. A doubling in the number of projects does not mean wafer shipments or revenue grow by the same proportion.

Losses are narrowing. The combined operating loss of Samsung's foundry and System LSI businesses is projected to shrink from KRW 6.742 trillion in 2025 to KRW 3.921 trillion in 2026, according to Digital Today, citing Kiwoom Securities.

Quarterly losses are estimated to have eased to KRW 777 billion in the third quarter of 2026, after widening in the previous quarter.

Why a Taiwan design partner matters

DSPs sit between chip designers and foundries, handling design work and then coordinating wafers, packaging and test schedules. Chosunbiz reported, citing Samsung Foundry's official website, that the roster is centred on South Korean firms such as ADTechnology, GAONCHIPS, Semifive and CoAsia, with Faraday understood to be the only Taiwan-based firm on it.

According to Chosunbiz, Samsung is pursuing a strategy of growing its HPC customer base through DSPs after signing direct contracts with customers such as Tesla and Broadcom. Those direct deals are far larger: a $16.5 billion foundry contract with Tesla and a $200 billion memorandum of understanding with Broadcom covering memory and foundry cooperation.

Domestic DSP workloads have been driven largely by AI chip startups such as DeepX, and there is a view that relying on domestic partners alone made it difficult to land large US cloud customers.

A semiconductor industry official told Chosunbiz that tight capacity at TSMC for advanced nodes created an opening. The official called the Faraday deal "the most important example and supply reference" of a Taiwan design firm choosing Samsung as an alternative.

Production of advanced chips for AI accelerators and cloud infrastructure remains concentrated at TSMC, though Samsung's advanced nodes could become an option for cloud providers looking to diversify, XenoSpectrum wrote.

The gap is still wide. TrendForce reported that HPC contributed 68% of TSMC's total revenue in the third quarter of 2026, against Samsung's target of more than 50% of foundry revenue from HPC in 2027. Faraday is pursuing additional 4nm and 2nm projects and expects to finalise designs for four advanced-node ASIC projects in the first half of next year.

What happens next

Seoul Economic Daily reported that Samsung plans to begin mass production of its second-generation 2nm mobile chip, the Exynos 2700, in the fourth quarter of 2026, alongside a ramp-up of 4nm NPUs for a major customer. Samsung expects double-digit revenue growth in the fourth quarter. The company is also considering investments in new fabs in the US and South Korea to meet demand for 2nm AI and HPC chips, according to Yonhap News.

Comments 0

Sources

31
  1. 01Samsung Foundry wins 2nm cloud AI chip orders via Taiwan's FaradayEN
  2. 02Samsung Foundry Taps Taiwan Partner Faraday to Land 2nm Cloud AI Chip DealEN
  3. 03Samsung Foundry Expects Advanced-Node Utilization Above 100% in 2027 as 2nm Deals Set to More Than DoubleEN
  4. 04[News] Samsung Sees 8nm-and-Below Utilization Topping 100% in 2027; 2nm Projects to More Than DoubleEN
  5. 05Samsung secures Faraday 2nm ASIC order, broadening AI and HPC footprintEN
  6. 06Tesla and Samsung Strengthen Chip Alliance with AI4+ Production Plans - SammyGuruEN
  7. 07Samsung Reportedly to Exit 8-Inch Foundry by 2030 as TSMC Also Cuts CapacityEN
  8. 08 Nvidia tightens tracking of AI GPUs in Southeast Asia as Chinese chip smuggling draws scrutiny, report claims EN
  9. 09Chip Industry Week In ReviewEN
  10. 10[News] Samsung 2nm Taylor Fab 2 Reportedly Fully Booked Before 2030 Output, Led by AMD HBM Base Die OrdersEN
  11. 11Nvidia’s AI Chip Exports Emerge as Target in Democratic CongressEN
  12. 12Details zu OneUI 9.5: Diese 2 Apps soll Samsung verbessernEN
  13. 13 Intel’s next-gen Nova Lake needs new motherboards — what we expect from Nova Point and Z990, Z970, and B960 chipsets EN
  14. 14heise+ | Angriffe auf Prozessor-Hardware: Wie der TÜVIT Chips knacktEN
  15. 15China’s AI chipmakers report first profits, but will the trend continue?EN
  16. 16Cloudflare Rises After AI Demand Lifts Annual Profit ViewEN
  17. 17Künstliche Intelligenz: EU will Chips aus Europa mit halber Milliarde Euro fördernEN
  18. 18 Apple MacBook Pro (14-inch, M6) review: Speedy M6 chip slips into comfortable shoes EN
  19. 19 Grab a huge 44% saving on a 4K 43-inch Samsung smart monitor, now under $280 EN
  20. 20Inside the chip talent war: Trade secrets and rising Chinese rivalsEN
  21. 21Europe’s Chip Industry: Perfect Storm or Perfect Timing?EN
  22. 22The Two-Ledger Model: Transaction-Grain Cost Accounting for Multi-Agent AI Systems Across Clouds, Models, and Systems of RecordEN
  23. 23 Nintendo ROM chip supplier Macronix quits to focus on chips for AI data centers, reports indicate EN
  24. 24 Nvidia can only fulfill 1% of China’s massive $70 billion GPU order for AI chips, report claims EN
  25. 25 Cloudflare engineers optimize DNS build services for twice the speed and 85% lower RAM usage EN
  26. 26Accelerating Plug-And-Play Prototyping In Chiplet-Based DesignsEN
  27. 27Chinese chipmaker Hua Hong bets big on silicon photonics amid AI boomEN
  28. 28Samsung suffers its worst 5G quarter but knows it can outlast rivalsEN
  29. 29Li Auto's Mega faces Samsung OLED screen shortage as deliveries reboundEN
  30. 30Cloud Revenue Hits $168B as AI Capacity ExpandsEN
  31. 31DarkBlinders Compromises Israeli Target and Iraqi Kurdish Government Cloud EnvironmentEN

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

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

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