Intel and SK Hynix have opened exploratory talks on US memory chip manufacturing, a signal that the gap between AI demand and high-bandwidth memory supply has become the kind of problem that pushes competitors into alliance.
The discussions cover US production of HBM or advanced packaging, areas where SK Hynix holds a commanding position and where Intel has been trying to re-establish a credible presence. The strategic logic is straightforward: the US government wants critical memory nodes on domestic soil, SK Hynix wants to reduce single-country risk, and Intel wants access to HBM integration capacity it has been chasing for years.
Market context sharpens the stakes. ASML's EUV machines are sold out through 2027, CXMT is mass-producing 5th-gen DRAM in China, and memory price pressure is already showing up in component and hardware margins. An Intel–SK Hynix US program, if it materializes, would be a statement that the memory tier of the AI stack is moving into geopolitical production planning, not just commercial negotiation.
For investors watching the chip complex, the announcement adds a second front of US industrial policy to the existing push on advanced logic and lithography. The risk is execution: HBM requires a tight coupling between logic, packaging, and memory design, and a new US plant years from now is not a hedge against the current shortage. It is a bet on a future in which US-based HBM is a policy requirement, not just an economic preference.
Image source: i.ibb.co