SK Hynix denies report that Intel foundry will use HBM4E memory

SK Hynix has denied a market report that it has entered a foundry partnership with Intel for HBM4E, according to an update published by Yahoo Finance on August 31, 2026. The denial was highlighted in the opening paragraphs of the report, reflecting how quickly semiconductor supply-chain rumors can move when investors look for indicates on next-generation AI memory and advanced manufacturing.

HBM4E is the next generation of high-bandwidth memory, a type of memory designed to feed data-intensive AI and accelerator chips at very high speeds. It is typically packaged closely with compute logic, because the performance advantage comes not only from the memory itself but also from how the memory is integrated into the final processor or accelerator module.

The report framed the question as whether Intel’s foundry efforts, which aim to manufacture chips for a range of customers, would include SK Hynix’s HBM4E. Intel is not the only potential beneficiary of a memory upgrade, but HBM4E is widely watched because leading-edge AI systems often require the latest memory generation to keep throughput high and bottlenecks low.

Beyond the denial itself, the Yahoo update did not provide details in the text made available here on what, if anything, Intel and SK Hynix may be working toward, or whether any agreement exists that was being characterized as a foundry partnership. The distinction matters because “partnership” can refer to many things in semiconductors, from long-term supply commitments to packaging and qualification timelines for specific chip designs.

Intel, through its foundry strategy, has positioned its manufacturing services as a way to support clients that need leading-edge process technology. In that context, memory procurement and qualification are part of the same ecosystem decision as wafer starts, packaging readiness, and the ability to ramp production for high-volume AI accelerators.

The denial from SK Hynix suggests the market report may have overstated the state of commercial arrangements, or that the companies have not publicly confirmed the specific linkage being discussed. In semiconductors, even when components are planned for future products, confirmation can arrive later, after design wins, qualification milestones, and supply chain contracts are finalized.

Still, the episode does not fully resolve the broader timing question around HBM4E and advanced AI systems. While SK Hynix rejected the specific framing that it has a foundry partnership with Intel for HBM4E, it did not, in the material available here, lay out an alternative public roadmap for how or when HBM4E would be used across customers.

What to watch next is any clarification from either company that narrows the gap between rumor and reality, such as public statements about customer qualification progress for HBM4E, packaging and integration commitments for high-volume accelerator platforms, or confirmed supply arrangements tied to Intel’s manufacturing engagements.