Samsung Electronics plans a second foundry plant in Taylor, Texas, covering about 2.7 million square feet — roughly 1.5 times the size of its first facility — to make memory chip components for AMD and Nvidia, Reuters reported on 3 November, citing South Korea’s Maeil Business Newspaper. The planned plant would expand Samsung’s US contract-manufacturing operation, with construction due to start formally in November.
Key points
- The reported orders cover the second Taylor plant’s entire output, with most of the volume allocated to AMD.
- Samsung plans to make high-bandwidth memory base dies there using 2-nanometre technology.
- Equipment installation is expected from February 2029, followed by mass production in 2030.
- Samsung’s first Taylor facility has a contract with Tesla.
AMD takes most of Taylor Fab 2’s reported orders
The second facility, called Taylor Fab 2, is intended to produce base dies for high-bandwidth memory, or HBM, used with AI processors. The base die sits beneath the memory chips in an HBM stack and handles communication with the processor. Its manufacture is a separate step from producing the memory dies above it.
Orders have been secured for the plant’s entire planned output, most of it for AMD, the Reuters report said. Nvidia would take part of the remaining volume. The reported allocation gives the two chip designers places in Samsung’s production plans before construction formally begins, although the first base dies are not expected to enter mass production until 2030.
Samsung declined to comment on the matter, Reuters reported. The orders and production schedule therefore rest on the account of Maeil’s industry sources rather than a public announcement from the manufacturer. The reported volume allocation also makes AMD the larger of the two named customers for the new plant.
Samsung plans 2-nanometre base dies in Texas
The plant is planned around a 2-nanometre manufacturing process. Crypto Briefing reported that current HBM4 base dies use a 4-nanometre process and that Samsung intends to pair base dies made in Texas with DRAM core dies produced in South Korea. Those components would be assembled into complete HBM stacks.
That production plan assigns the logic layer of the memory stack to the Texas foundry while retaining production of the memory dies in South Korea. Crypto Briefing described Taylor Fab 2’s intended customers as exclusively AMD and Nvidia. Its account said the facility was scheduled for an early-November groundbreaking, ahead of the production timetable reported for the plant.
Samsung reportedly planned to use TSMC’s process for Nvidia memory chips. AI Affairs previously reported a plan to use TSMC’s 3-nanometre process for Nvidia memory base dies. Making dies at Taylor would place that part of the work inside Samsung’s own US foundry operation.
Tesla holds the Taylor Fab 1 contract
Samsung’s first Taylor facility, Taylor Fab 1, is under construction and has a contract with Tesla. Crypto Briefing put the value of that agreement for next-generation AI chips at approximately $16.5 billion. The Tesla arrangement provides a separate named customer for Samsung’s first plant, while the reported AMD and Nvidia orders cover the planned output of the second.
Crypto Briefing said Fab 1 was expected to reach full operational capacity by the end of 2026 and move towards production in 2027. For Fab 2, the schedule extends further: Samsung is expected to begin installing equipment in February 2029 and start mass production in 2030.
.jpg){kind=link}