Herald Business ran a [단독] — an exclusive — on 31 August at 10:01 KST (01:01 UTC) reporting that SK hynix is considering giving Intel the job of making base dies for HBM4E, the logic layer that sits under the memory stack and acts as its controller. It is a genuinely large claim: that work currently goes to TSMC, and moving it would be one of the biggest foundry reallocations in the memory business.
The denial is in the article
Further down the same piece, SK hynix is quoted directly. On its roadmap generally: "당사 기술 로드맵 관련 내용은 확인이 어렵다" — it is difficult to confirm anything about our technology roadmap. Then, specifically, on the Intel question: "HBM4E 베이스다이 생산을 인텔에 맡기는 부분에 대해서는 검토하고 있는 바 없다" — there is no such review under way regarding entrusting HBM4E base-die production to Intel. That is not a no-comment. It is a flat denial of the headline claim, printed inside the article making it.
What is actually established
Strip the denied part away and a smaller, sourced fact remains: SK hynix has said publicly that it is comparing advanced packaging approaches, weighing TSMC's CoWoS-S and CoWoS-L against Intel's EMIB. Comparing packaging methods is not the same as awarding base-die production, and the two got compressed into one story. A companion piece 28 minutes later, at 10:29 KST, framed the same material as a multi-vendor strategy for price leverage — which is what a comparison exercise looks like.
What the common framing gets wrong
English syndication of this story softened the denial to something like "some of the reports are not true," which reads as a partial quibble. The Korean is not partial. A second distortion travelled with it: a claimed cost gap between Intel and TSMC base dies compares a logic die on a leading node with a DRAM core die — different products on different processes, so the ratio is not a like-for-like price.
Why the shape recurs
Korean tech exclusives routinely run with the company's denial included, because the outlet has sourcing it trusts and the company will not confirm a roadmap. That is a defensible way to publish. What breaks is the next hop: aggregators lift the headline, drop the eighth paragraph, and a denied plan becomes a decided one. The primary document says both things — and only one of them is on the record from the company.