Will self-developed CXL undermine their own DRAM business? Samsung, SK Hynix, and Micron simultaneously abandon self-development of CXL controllers.

CN
23 hours ago

Original author: Li Jia

Original source: Wall Street Journal

The three major global memory chip manufacturers have completely halted their independent research and development plans for CXL (Compute Express Link) controllers and have turned to procure products from specialized chip design companies. This collective shift reflects the core contradiction in the memory industry: the aggressive promotion of self-developed chips may erode its most important revenue source - the general-purpose DRAM module market.

On July 20, according to a report by Korean technology media ZDNet Korea, Samsung Electronics, SK Hynix, and Micron Technology have all scaled back or abandoned their commercialization plans for CXL expansion device controllers. Companies such as Montage Technology, Astera Labs, and PrimeMass are filling this gap.

This change indicates that the division of labor in the CXL ecosystem is being reshaped, with memory manufacturers focusing on manufacturing while design leadership shifts to independent chip design companies.

For the capital market, this is not only beneficial for related chip design companies, but also means that the three major memory manufacturers will not engage in new competition around CXL complete solutions in the short term, and their core profit model will still revolve around traditional DIMM memory products.

Three major manufacturers adjust strategies sequentially

The rhythm at which the three companies exited independent research and development of controllers is not the same.

Micron's actions are the most thorough. According to reports, the company has shut down its CXL controller R&D department and adopted PrimeMass's controller solution, with related products now included in Micron's product catalog.

SK Hynix has formally notified its partners to terminate the self-developed controller project and has redeployed the R&D team to PIM (Processing In Memory) business, indicating its intention to concentrate resources on more competitively viable next-generation memory architectures.

Samsung's adjustment is relatively cautious. Reports quoting semiconductor industry insiders say that Samsung has removed its self-developed CXL controller from its official commercialization roadmap, retaining only cutting-edge R&D projects, with the team currently focusing on directions such as the combination of LPDDR and CXL.

In the meantime, Samsung has turned to externally sourcing controller products. This means that the company's previous plan to launch an "integrated product of self-developed controller + CXL Memory Module (CMM)" has essentially been put on hold.

Why has self-developed controller lost its appeal?

The initial vision of the three memory manufacturers was highly consistent: to package controllers and DRAM on the same circuit board and sell it as a complete solution to obtain higher product added value.

However, the market ultimately chose a different path. Data center customers prefer to adopt a modular architecture, with controllers independently deployed on the motherboard, paired with standardized, low-cost DIMM memory modules. This solution reduces procurement costs while also preserving flexibility for future upgrades and supply chain choices.

The change in customer demand puts memory manufacturers in a dilemma. If they insist on promoting integrated products, they not only have to bear controller R&D costs but may also affect market acceptance due to higher prices; more importantly, if customers reduce purchases of standard DIMMs, it would impact the memory manufacturers' core revenue source.

According to reports quoting a semiconductor industry insider: if the CXL complete product must directly compete with the company's DIMM business, then the significance of promoting the commercialization of the controller will significantly decline. In the presence of risks that can erode core business, management would find it challenging to continue investing resources.

In other words, rather than cultivating a new market that could weaken the DRAM business, it is better to leave the controllers to specialized manufacturers and focus on memory manufacturing, where they have the most scale advantage.

The CXL ecosystem enters a specialized division of labor stage

Industry insiders believe that the three major manufacturers exiting self-developed controllers does not mean giving up on CXL.

On the contrary, this is more like a typical specialization in the process of industrial maturity. Controller design is entrusted to fabless chip design companies, while memory manufacturers continue to capitalize on their manufacturing and DRAM technology advantages, each creating value in the most competitive segments.

As this trend solidifies, Astera Labs, Montage Technology, and PrimeMass are expected to further consolidate their market position in CXL controllers, while Samsung, SK Hynix, and Micron can concentrate their R&D resources on advanced DRAM processes, HBM, and PIM, among other areas with more long-term value.

From the perspective of the industry chain, the development logic of CXL has not changed; what has changed is the roles of the participants—memory manufacturers are no longer trying to build a complete platform but are returning to supply chain division of labor. This also means that the CXL ecosystem is moving from the early exploratory phase to a more mature commercialization phase.

免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到support@aicoin.com,本平台相关工作人员将会进行核查。

Share To
APP

X

Telegram

Facebook

Reddit

CopyLink