AMD has reportedly canceled its 4nm chip orders with Samsung Foundry and shifted production to TSMC's U.S. facilities in Arizona, dealing another blow to Samsung's efforts to gain ground in the advanced foundry market.

The decision, first reported by @Jukanlosreve and echoed by Wccftech, appears to reflect growing concerns over the stability of Samsung's 4nm SF4X process. AMD had been collaborating closely with Samsung on the node for a range of products—including EPYC server CPUs, Ryzen APUs, and Radeon GPUs—as part of a dual-sourcing strategy. That partnership now seems to be unraveling.

While it's not yet clear whether AMD's move affects only server CPUs or broader product lines, the company's confidence in Samsung Foundry appears to be fading. AMD has reportedly opted not to proceed with mass production on Samsung's 4nm process for GPUs, citing process maturity issues.

The shift underscores AMD's deepening reliance on TSMC, particularly its U.S.-based operations. TSMC's Arizona fab has already validated AMD's 5th-gen EPYC “Venice” chips, which will use the company's next-generation 2nm “N2” node. Slated for launch in 2026, the chip will be AMD's first high-performance computing (HPC) product built on 2nm.

Industry analysts point out that geopolitical risks and growing concerns over supply chain concentration may eventually push companies like AMD to diversify their foundry partners. For now, however, TSMC maintains a clear edge in yield, reliability, and production scale.

Meanwhile, Samsung is pressing ahead with its own 2nm development. Korean media report that the company has achieved an initial 30%+ yield on its SF2 test runs—surpassing expectations—and is targeting stable mass production in the second half of 2025. Samsung is also reportedly close to securing a 2nm deal with Qualcomm for the next-generation Snapdragon 8 Elite 2.

Although Samsung's recent progress is promising, losing AMD as a major 4nm customer is a significant setback, highlighting the Korean firm's ongoing struggle to win and retain high-end foundry clients.