I.
As for the cooperation between SK hynix and Intel, it is still in the discussion stage.
On September 16, Micron?—no, Hynix—sorry: Hynix already officially clarified that it is reviewing various proposals. But regarding Intel, the Ohio facility, and producing memory chips in the U.S., there is currently no final decision.
Reuters also made it clear that even now, it is impossible to confirm what the future Ohio factory will produce—DRAM, NAND, or HBM.
However, if you string together what happened over the past few months, the direction of cooperation between the two sides has gradually become clear.
In June this year, Intel hired Lee Si-hee, the former CEO of SK hynix, to join its foundry business, responsible for advanced packaging and back-end processes.
Lee Si-hee previously led DRAM development at SK hynix for a long time, and also served as COO and CEO. He is one of the key figures in the development of the HBM industry.
Therefore, back then, industry in South Korea began discussing whether Intel would use this opportunity to win SK hynix’s next-generation HBM base die orders—that is, Base Die.
By August 31, Korean media further reported that SK hynix is considering, starting with HBM4E, shifting some of the base-die orders that have previously been mainly handled by TSMC to Intel wafer foundry, thereby establishing a dual-supplier system.
Hynix later said that “some parts of the coverage are not true,” but it did not specify which parts are being denied. That is also quite interesting.
On September 16, Reuters pushed the story to another level: SK hynix and Intel are discussing having SK hynix produce memory chips in the United States for the first time.
This also brings up a key question: Why does SK hynix need Intel?
II.
I think this is the most important point for understanding the whole matter.
The real problem SK hynix is facing is gradually shifting from “whether HBM can be made” to “how to expand HBM capacity with sufficiently low costs and sufficiently stable supply.”
Because starting with HBM4, the base die is no longer just a simple substrate board. It is getting closer and closer to a real logic chip—and it is highly customized.
HBM is evolving from being a simple memory product into a system-level product of “memory + logic + advanced packaging.”
At present, SK hynix’s basic HBM4 die supply is highly dependent on TSMC.
TrendForce News summarizes reports from South Korea’s industry chain: the cost of HBM base dies produced by TSMC may already be 3 to 4 times that of the DRAM base dies SK hynix makes in-house. After entering HBM4E, cost pressure may rise even further.
Meanwhile, SK hynix has already signed long-term supply agreements with a large number of customers. This means that after wafer-foundry costs rise, the company likely cannot pass all of the increased costs to customers such as Nvidia, Broadcom, and Meta without limit.
So the first value of Intel is right here.
Even if in the end Intel can only secure 10% to 20% of the HBM base-die orders, it is still meaningful for SK hynix.
Because as long as a second supplier is formed, SK hynix’s single-source reliance on TSMC will decrease. That would also give it greater bargaining power in terms of pricing, capacity, and supply assurance.
III.
The second value is to build a domestically based HBM supply chain in the United States.
SK hynix is already building an HBM advanced packaging plant worth about $4 billion in Indiana, and plans to start mass production after 2029.
However, the Indiana facility mainly handles back-end packaging. The DRAM wafers themselves are still produced primarily in Asia.
In other words, the supply chain today is roughly: DRAM wafers are first manufactured in South Korea, then shipped to Indiana, the United States, for HBM packaging.
If Intel’s Ohio facility also joins in the future, this chain could become: front-end manufacturing of DRAM or other memory chips in Ohio, U.S., with Intel or TSMC providing HBM base dies, then advanced packaging in Indiana, and finally supplying Nvidia, Broadcom, Meta, and other major cloud computing companies.
As a result, SK hynix would, for the first time, have an opportunity to form a relatively complete industry chain in the U.S.: “memory wafers—logic chips—advanced packaging—AI data centers.”
Reuters also specifically noted that the Indiana project currently only belongs to back-end packaging. If in the future SK hynix truly produces memory wafers in the U.S., it would mean the company would, for the first time, have both front-end manufacturing and back-end packaging capabilities in the U.S.
This is also attractive to large U.S. cloud companies.
Another detail worth paying attention to is why Reuters specifically mentioned the possibility that “large cloud computing companies participate in the joint venture company.”
Cloud companies may not necessarily want to produce DRAM themselves, but they are very willing to ensure they can obtain HBM and DRAM reliably in the coming years through long-term contracts, advance payments, or even direct investment.
Reuters mentioned that one core purpose of large cloud companies participating in a joint venture is to ensure the supply of memory chips.
SK hynix has also disclosed this year that it is moving forward with around ten key customers on multi-year long-term supply agreements. This means the HBM market is gradually shifting from the traditional one-year spot-and-contract model to multi-year strategic capacity agreements.
Once this model keeps expanding, it could also reduce one of the most severe problems the storage industry has faced in the past: cyclicality.
IV.
From the perspective of institutional investors, the market’s understanding of this issue is broadly centered on three directions.
TrendForce News and related industry-chain reports believe that if Intel can become the second supplier of SK hynix’s HBM4E base dies, on the one hand it can reduce the company’s reliance on TSMC as a single source, and on the other it can help improve cost control.
If the cooperation succeeds, it could even attract other memory makers such as Samsung and Micron to reevaluate Intel’s foundry offering. For SK hynix, this is cost control and supply-chain insurance; for Intel, it means its foundry capacity has been validated by external customers.
In its earlier Intel research report, the U.S. bank also said it considers negotiations with external foundry customers to be more concrete, U.S. advanced manufacturing capability, and Lee Si-hee joining Intel as important signals that the foundry business is improving.
Therefore, SK hynix’s news aligns with the earlier assessment by the U.S. bank regarding the repair of Intel’s foundry business.
For SK hynix itself, UBS previously expected the company’s HBM bit shipment share this year to be about 48%, but as Samsung expands HBM4 capacity, next year’s market share may face competition.
Therefore, in the future SK hynix will need to maintain its profit margins through cost control, the supply chain, and customized HBM—not rely solely on its current leading market share.
LS Securities’ view is even more direct. The firm has lowered its expectation for SK hynix’s HBM operating margin for the next fiscal year from around 80% to around 60%. The main reason is that as Samsung enters the market and customers adopt multi-supplier systems, the extremely high excess margins for HBM in the past may gradually normalize.
In this situation, if Intel can lower the cost of HBM base dies, it could potentially become an important tool for SK hynix to maintain its HBM profit margins.
Wolfe Research believes that currently there are still clear constraints on storage and wafer-founding capacity. AI chip supply may remain constrained by capacity bottlenecks at least until 2028.
Therefore, it is necessary to increase capacity over the long term. However, the Ohio facility is expected to ramp up around 2030 to 2031, which will not help much for the current round of HBM tight supply.
This also means the timing of Intel’s stock price and earnings impact from this matter is different.
V.
For INTC, the time when the stock price truly trades this theme is likely to concentrate between Q4 2026 and 2027, because the market is pricing expectations, not waiting until the factories are actually in production before it starts setting prices.
Intel currently does not lack factories, nor does it lack technology storylines such as 18A, 14A, and advanced packaging. What the market truly lacks is an external customer with enough weight.
Intel has also already admitted in regulatory filings that the number of external foundry customers is still limited. Whether 14A can gain a sufficiently large base of external customers will directly affect whether the company continues to invest heavily in the next generation of advanced process nodes.
Therefore, one of the biggest valuation variables for INTC right now is actually very simple:
In the end, are there truly large external customers who are willing to have advanced chips produced by Intel?
If SK hynix ultimately officially confirms that Intel becomes the second supplier of HBM4E base dies, the amount of information would be enormous.
Because it proves not just that Intel gets an order, but that a globally leading semiconductor company is willing to entrust key products to Intel’s foundry production.
The stage where revenue is truly validated is more likely to appear in the next 6 to 18 months—that is, in 2027.
HBM4E itself is expected to enter the mass production cycle in 2027. If Intel truly becomes the second supplier of base dies, by then you should gradually see more tangible changes, including increases in external foundry revenue, higher wafer shipment volumes, growth in advanced packaging revenue, improved capacity utilization, and foundry losses narrowing.
At this stage, Intel’s story would truly shift from “expectations and narrative” to “revenue and profit.”
Therefore, for INTC, the stock price could potentially trade the benefits of Intel’s collaboration with SK hynix ahead of time in the fourth quarter of 2026;
The upside that truly shows up in the wafer-founding business performance is more likely to be gradually validated starting in 2027. As for the large revenue brought by the Ohio facility itself, it likely won’t be seen until after 2030.
I will keep tracking this news as well; you can give me a follow.