"Turning the tide" or "money-burning pit"? Intel Corporation (INTC.US) aims to re-enter the storage market, betting on a new architecture to stand strong in the AI era.

Intel Corporation (INTC.US) CEO Chen Liwu recently signaled that the company may be returning to the storage chip market. If true, this move would mark a strategic shift for Intel Corporation, potentially expanding its product portfolio and influencing its competitive stance within the semiconductor industry and its partnerships with other companies. Reportedly, on August 11 local time, Chen Liwu revealed during a podcast that a new type of storage architecture, once considered a commodity business, has now emerged as a strategically significant field that he is focusing on. He also stated that the storage industry is at a critical point for innovation and hinted that Intel Corporation is exploring options to stack storage on top of CPUs. After years of competitive commoditization, storage chips have regained their status as a strategic asset in recent quarters, a trend that may continue for some time. This has led to substantial profits for major global storage chip manufacturers, which Chen Liwu is evidently keenly aware of. Chen Liwu remarked, "In the past, my idea was 'don't invest in storage chips because it is a commodity business,' but the situation has changed now." He added, "A lot of new technologies are coming into play. Therefore, we are looking at some new storage architectures, which is one of my key areas of focus." He also mentioned the former CEO of SK Hynix, Lee Seok-hee, who joined Intel Corporation in June this year. Lee is currently serving as the executive vice president of Intel Corporation's foundry business, responsible for advanced packaging, system integration, and backend technology development and manufacturing. When discussing this arrangement, Chen Liwu indicated that outsiders could "probably guess" what direction he is considering, although he is not yet ready to disclose more details. In fact, Intel Corporation originally started out as a storage chip company. After its founding in 1968, Intel Corporation achieved significant success in the storage chip sector. However, by the 1980s, Japanese companies gradually took the lead, causing Intel Corporation to incur severe losses and ultimately forcing it to exit the storage chip market entirely. Since then, Intel Corporation has made several attempts to re-enter the storage field. The company has ventured into NAND and Optane storage businesses or attempted to identify new growth opportunities through new storage technologies like RDRAM. However, in all three attempts, Intel Corporation ultimately abandoned the related storage businesses without incurring particularly severe financial losses. Since the beginning of this year, Intel Corporation has been actively working in the storage and advanced packaging sectors. In February, Intel Corporation announced a partnership with SoftBank's SAIMEMORY to co-develop ZAM (Z-Angle Memory). This technology, aimed at AI and high-performance computing, focuses on balancing memory capacity, bandwidth, and power consumption. Prototypes are scheduled for development completion in fiscal year 2027, ending March 2028, and commercialization will advance into fiscal year 2029. Some technologies for ZAM stem from Intel Corporation's previous involvement in the NGDB (Next Generation DRAM Bonding) project, which has completed 8-layer DRAM vertical stacking tests to research ways to reduce high bandwidth's sacrifices on memory capacity through improved stacking and connection methods. In July, a patent application by Intel Corporation regarding Cross-Batch Memory (XBM) was disclosed by the media. XBM still uses DRAM but replaces the wide parallel interfaces of HBM with once used back-end process (BEOL) transistors and serial UCIe links. This design can save on the silicon interposer that traditional HBM relies on while occupying a package area similar to HBM4, thus shrinking the overall package and reducing assembly complexity and costs. Additionally, Intel Corporation continues to advance its advanced packaging and 3D integration technologies that support these storage architectures, including EMIB, Foveros, and 18A-PT processes geared toward 3D integration. When looking at these moves together, it becomes clear that Intel Corporation is not simply betting on a return to traditional DRAM or NAND production, but rather seeking new entry points in storage architecture. Previously, CPUs, GPUs, and storage chips typically existed as relatively independent chips, but as the requirements for data throughput in large model training and inference continue to rise, data movement has become a new performance and power bottleneck, prompting the industry to further explore closer chip stacking and system integration beyond HBM. Chen Liwu also discussed the stacking of CPUs and storage chips together, believing there are still new combinations possible between the two, and that altering the memory architecture can further shorten the distance between CPUs and storage. For investors, Intel Corporation's signal regarding a possible return to the storage market is directly related to the company's renewed focus on its business mix and rebuilding trust in its role within the AI (artificial intelligence) infrastructure sector. Reintegrating the storage business into the product portfolio may help Intel Corporation achieve its goal of providing a more comprehensive platform for emerging AI workloads, rather than just offering CPU and wafer foundry capacity. Considering the current profitability of 3D NAND and DRAM manufacturers, producing storage chips has indeed become a lucrative business again, and may remain highly profitable for the foreseeable future. However, this also raises a risk in the existing investment logic. Intel Corporations complex organizational structure, along with high operating and capital expenditures, has long been a central concern for the market. To re-enter the storage market, substantial capital investments will be required, including the construction of at least one new wafer fab, along with R&D funding to develop competitive manufacturing processes, necessitating a significant amount of time to achieve scale. Re-entering such a capital-intensive area of storage chips will also test Intel Corporations ability to simplify its operations. An immediate metric for investors to watch is how Intel Corporation will position its storage business within its broader capital planning and AI product roadmap in upcoming financial earnings calls and industry events. Whether the company can clearly disclose how much funding it plans to allocate toward the storage business, whether target customers will focus on data centers and AI sectors, and whether Intel Corporation will utilize existing fabs or require new capacitythis information will indicate whether this strategic adjustment supports the beneficial factor of "refocusing," or may increase the risks of execution difficulty and business complexity. However, several media outlets have pointed out that the prospects of Intel Corporation returning to the storage market remain unclear. The well-known tech hardware media platform Tom's Hardware noted that if Intel Corporation truly resumes large-scale production of storage chips, it would not only need to reinvest in fabs but also establish competitive manufacturing processes while facing lengthy construction and validation cycles. Given that Intel Corporation still needs to continue investing in CPU products and foundry business, it remains questionable whether the company is willing to invest substantial capital in traditional storage chips again. Tech news media ZDNet also believes that Chen Liwu's statement should not be interpreted as Intel Corporation preparing to resume general DRAM business, but more likely indicates that the company is increasing its focus on next-generation storage technologies. Even if the new storage architecture can achieve commercialization, entering the AI market is equally challenging. Some media point out that NVIDIA Corporation currently occupies over 80% of the AI accelerator market, and its hardware and software ecosystem has already been built around HBM, making it difficult for new storage architectures to completely replace HBM in the short term. Even if ZAM later achieves mass production, it is more likely to appear first in custom AI chips or products primarily focused on inference, serving as a complement to HBM. Despite Chen Liwu's acknowledgment that plans to return to the storage chip market have not been finalized, and no timeline, product roadmap, or capital commitment has been disclosed, his comments have opened up new possibilities for Intel Corporation's transformation narrativeafter missing multiple tech waves, this once-giant from Silicon Valley is attempting a crucial return in the AI era through a new architecture that merges computing and storage.