[News] Samsung Accelerates U.S. Expansion as Taylor Fab Targets 2027 Start, HQ Reportedly Moves to Texas in 2026
As Samsung pushes ahead with HBM4E sampling to stay ahead in the memory race, it is also accelerating its foundry expansion, with a strong focus on the U.S. market. According to Aju News and Tech Times, Margaret Han, VP of Samsung Electronics Foundry’s North American operations, confirmed that the Taylor, Texas fab will begin production in 2027, with further details on its advanced process roadmap to be unveiled at the Korea SAFE Forum on July 1.
Aju News adds that with Executive Chairman Jay Y. Lee personally leading efforts to secure overseas customers and taking a hands-on role in steering the foundry business, expectations are rising for a turnaround in Samsung’s loss-making non-memory division. According to the report, its Foundry and System LSI divisions are estimated to have posted an operating loss of around KRW 6.8 trillion in 2025.
Meanwhile, according to SeDaily, Samsung is also moving its U.S. headquarters from New Jersey to Texas, in a push to strengthen synergies with its semiconductor operations in the region. The report adds the relocation is expected to be completed within this year, with the company having recently informed employees at its U.S. subsidiary.
Big Techs’ Interest Fuel Turnaround for Samsung Foundry
As highlighted by Aju News, Samsung has already attracted strong interest from major tech players. Alongside longtime customer Qualcomm, Tesla has reportedly secured around US$16.5 billion in capacity for its next-generation AI and autonomous driving chip, AI6, set for production in 2027, while Apple executives are also said to have recently visited the site to explore potential chipset partnerships.
On the other hand, The Elec reports that Samsung’s participation in Anthropic’s US$65 billion Series H funding round has fueled speculation that its partnership could extend beyond memory supply into AI chip development.
Among the three memory suppliers cited by Anthropic, Samsung is the only one with a logic semiconductor foundry business, raising expectations it could potentially manufacture AI chips for Claude models, the report notes, adding that industry observers suggest the option remains open as hyperscalers and AI companies increasingly shift toward in-house chip design.
SF2 Yields in Focus as Samsung Advances 2nm Roadmap
Samsung announced the Taylor facility in November 2021 with an initial US$17 billion investment, originally targeting production in the second half of 2024. However, as noted by Tech Times, the timeline has since been pushed back multiple times—to 2025, then to the second half of 2026, and most recently to early 2027.
Notably, according to Tech Times, the latest delay is partly linked to Samsung’s decision to drop the originally planned 4nm node and jump directly to 2nm, based on its Gate-All-Around (GAA) transistor architecture that has been under refinement since the 3nm ramp.
While yield remains a key challenge, the report adds that by early 2026, performance had reportedly improved to around 55–60% on the first-generation SF2 node, with even higher levels achieved on the enhanced SF2P variant.
As outlined in the report, Margaret Han also elaborated at Samsung’s SAFE Forum 2026 in the U.S. on May 28, noting that the company continues to expand its 2nm roadmap. The SF2 family spans SF2 (first generation), SF2P (second), SF2P+ (third), and SF2X, a later-stage node targeting AI and high-performance computing workloads.
Read more
- [News] Samsung’s Taylor Fab Reportedly Nears 90% Readiness as Tesla Chip Push Advances; LPDDR6 Ties Emerge
- [News] Delay at Samsung’s Taylor Reportedly Slips Mass Production to 2027, Raising Concerns for Tesla
(Photo credit: Taylor Government)
