Samsung's second Texas plant reportedly pre-sold to AMD and Nvidia for 2030 output

A Texas fab pre-sold to AMD and Nvidia

The report landed on 3 November. Maeil Business Newspaper, citing industry sources, said the second Taylor plant will build HBM base dies for both companies.

According to the newspaper, Samsung has secured orders for the plant's entire planned output, and most of that volume is for AMD. Samsung declined to comment, so none of the reported customer commitments are officially confirmed. The site will use 2-nanometer technology and cover about 2.7 million square feet, roughly 1.5 times the size of the first Taylor plant. That first plant is anchored by a Tesla supply agreement valued at approximately $16.5 billion. Crypto Briefing described the new facility as Taylor Fab 2, a sibling to the Tesla-linked plant.

Why the deal matters beyond Samsung

If the report holds, AMD becomes the main customer of a plant that has not yet started construction. For AMD and Nvidia, the commitments lock in specialized base die capacity years before it runs. They add no chips before 2030.

The commitments do not immediately boost Samsung's foundry or HBM market share, GuruFocus noted. TSMC currently dominates the foundry market, and landing orders from AMD and Nvidia would position Samsung as a more credible alternative. Samsung's foundry business has struggled to attract major customers away from TSMC, and a large AMD order could give its 2-nanometer process a needed boost. The arrangements also suggest that leading AI chipmakers are locking up long-term access to specialized manufacturing capacity. Samsung has not said what the contracts are worth, and the reported volumes have not been disclosed.

Buyers of AI accelerators get a second source of base dies in the long run. They get nothing extra in the near term.

What the plant will actually make

The base die is the logic layer at the bottom of an HBM stack, and it manages how the memory above it talks to the AI processor. HBM stacks combine multiple DRAM dies to deliver the bandwidth AI accelerators need. Samsung intends to make base dies in Texas and pair them with DRAM core dies produced in South Korea, so the finished stacks are a transpacific product. Current HBM4 technology uses base dies built on a 4nm process, while the Taylor plant aims for 2-nanometer or more advanced nodes.

The split matters for how American the final stack really is.

Building base dies in Texas adds domestic US capacity for a critical AI component, though the DRAM core dies still come from South Korea, which limits how much of the stack is made in America.

The timeline and the spending

Construction is due to start formally this month, with equipment installation expected from February 2029 and mass production in 2030. That leaves roughly four years between groundbreaking and the first revenue from the site, a gap investors will have to carry.

Samsung has increased capital spending by over 60% to boost semiconductor competitiveness, and the company is betting that AI memory demand holds through the gap. The heavy spending, combined with the wait, means profit margins and cash flow could face pressure if AI chip demand disappoints before the plant runs. The industry is cyclical, so the long lead time amounts to a bet on sustained demand.

What happens next

The next catalyst will be confirmation of the customer agreements and construction progress at the Texas facility. The nearer test is the first Taylor plant, expected to reach full operational capacity by the end of 2026 and then ramp toward production in 2027.

Equipment installation at the second plant is scheduled from February 2029, ahead of mass production in 2030. Until then, the reported bookings are a public signal of demand for the site.