Samsung will use TSMC technology to produce next-generation memory chips for Nvidia, according to a report from a Seoul newspaper carried by CNA. The claim, unconfirmed by any of the three companies, signals a possible break in one of the semiconductor industry's most watched rivalries — the memory maker turning to its foundry competitor to secure a seat in Nvidia's roadmap.

That's the headline fact, and it arrives with caveats attached. The report originates from a single Seoul-based newspaper, citing the arrangement as a decision rather than a proposal. Samsung, TSMC, and Nvidia have not issued confirmations, and no deal value, capacity target, or production timeline appears in the account. Treat the specifics, in other words, as a report — not a signed supply agreement.

What is actually being claimed?

According to the Seoul newspaper's account, the arrangement centers on memory chips destined for Nvidia's next-generation products. The technical logic runs through how advanced memory now gets made.

Modern HBM-class memory stacks depend on a base die — the logic layer at the bottom of the stack — manufactured on a foundry process rather than on the memory maker's own lines. If Samsung builds those base dies at TSMC, it would effectively mean Samsung-designed memory sitting on TSMC-fabricated logic, assembled into packages that feed Nvidia's accelerators.

That structure has precedent. TSMC already fabricates logic dies used beneath high-bandwidth memory in leading AI systems, and foundry choice at the base-die layer has become a competitive lever as much as the DRAM itself. Nvidia, for its part, qualifies multiple memory suppliers for its flagship parts, and suppliers that can pair their stacks with the customer's preferred logic platform often move up the queue.

Why the foundry choice matters

The commercial stakes are straightforward. Nvidia's AI accelerator business has made next-generation memory a scarce, contested input, and memory vendors win sockets by aligning with Nvidia's packaging and logic requirements — not just by hitting density targets. A Samsung decision to route part of that stack through TSMC technology would acknowledge where Nvidia's platform preferences sit, even at the cost of handing business to the rival foundry.

For TSMC, the reported arrangement would extend its reach deeper into the memory supply chain, a segment where its share has historically been modest. For Samsung, it would mark a pragmatic turn: the company has invested heavily in its own foundry arm and its own advanced packaging, yet the reported deal suggests it would rather secure the Nvidia socket than insist on vertical purity.

What remains unconfirmed

Nothing in the report specifies volumes, pricing, node assignments, or a start date for production. The three companies have publicly said nothing, and past Samsung-Nvidia memory negotiations have stretched across qualification cycles before yielding confirmed shipments. Whether the reported technology arrangement covers one product family or a broader next-generation portfolio is also unstated.

The competitive dynamic, though, is already shifting. If the Seoul report holds, Samsung gains a path into Nvidia's next-generation memory supply with TSMC's process as the enabler — and the memory market's next round of socket competition gets decided as much in foundry agreements as in DRAM specs.