The Zhitong Finance App learned that some media quoted information revealed by people familiar with the matter as reporting that Solidigm, a subsidiary of SK Hynix Inc. (SK Hynix Inc.), one of the world's largest memory chip giants, is considering an independent US initial public offering (i.e. US stock IPO) in the US stock market as early as next year, and plans to go public independently in the US.
The expansion of intelligent applications represented by Muse and Astra is increasing the demand for AI infrastructure, extending from model computation to task execution, efficient contextual management, and massive data access. As a result, enterprise-grade solid-state drives (or enterprise-grade SSDs) have become an important observation direction for storage investment. In this context, Solidigm, a subsidiary of SK Hynix, is considering an initial public offering in the US as early as 2027 and discussing listing matters with potential advisors.
People familiar with the matter said that the world's second-largest DRAM/NAND memory chip manufacturer is discussing the US stock listing of its NAND flash memory business subsidiary with potential advisors. Since the relevant information has not been made public, these individuals requested anonymity.
Some people familiar with the matter said that the listing could raise Solidigm's valuation to an astonishing $100 billion. They said discussions are still ongoing, and details such as listing time and valuation may change.
Some US media were the first to disclose the details of the bank's contacts and listing schedule, citing people familiar with the matter who did not immediately disclose their identity. A Solidigm spokesperson declined to comment.
Solidigm's industrial location is in line with the trend of AI investment spreading from accelerators to complete data center systems. Its core business involves NAND flash memory and enterprise-grade SSDs, and its products enter scenarios such as cloud computing, servers, and data centers. Earlier, on August 5, CoreWeave announced the signing of a multi-year strategic agreement with SoliDIGM to obtain priority supply arrangements for enterprise-grade SSD capacity, and clearly stated that storage has become a key constraint in AI platform capacity planning. This also means that large AI cloud service providers are incorporating storage supply into long-term infrastructure construction plans to ensure simultaneous expansion of computing, network, and storage. For Solidigm, this type of cooperation helped increase demand visibility and also provided an actual customer basis for an independent valuation of its enterprise-level storage business; the announcement did not disclose the contract amount or specific procurement capacity.
In the age of AI reasoning, the core value chain of enterprise-grade SSDs is to provide customers with stable data access capabilities through comprehensive collaboration of NAND, controller, firmware, and system verification: when expensive AI accelerators need to continuously obtain data, the value of storage is also reflected in capacity supply, computational utilization, and operating costs of the entire system.
In 2021, SK Hynix acquired Intel's flash memory chip business and renamed it “Solidigm,” and the company was introduced. According to its website, the company produces massive data NAND storage products for data centers. Some of these devices are only about the size of a deck of playing cards, yet the capacity is as high as 122 TB.
In August of this year, the company announced that it had reached an agreement with CoreWeave to sell enterprise-grade solid-state drive storage capacity to this so-called “new cloud” AI cloud computing enterprise to support the AI cloud platform of CoreWeave, the leading “AI New Cloud” force. According to its website, other customers include VAST Data, Dell Technologies, and Chinese internet giant Tencent Holdings.
According to information, this leading memory chip company headquartered in Rancho Cordova, California, has 13 business locations around the world, including Mexico, Canada, and China, and employs more than 2,000 people.
The memory chip components of AI data center server clusters are still the clearest supply bottleneck at the AI computing power industry chain level. Market research agency TrendForce predicts that in 2026, server DRAM contract prices will increase by about 270%, and enterprise-grade SSD prices will increase by about 235%; HBM contract prices may still rise 70% to 140% in 2027. These data show the combined effects of AI computing power expansion and storage price increases. According to TrendForce's latest estimates, the combined share of DRAM and NAND in capital expenditure of major cloud service providers will rise from 47% in 2026 to 68% in 2027, behind which there is a simultaneous increase in procurement volume and price increases.
In terms of stock prices, as of the closing price of US stocks on September 25, 2026, calculated at the closing price at the end of 2025, and not including dividends, the US memory chip leader Micron (MU.US) has accumulated a crazy increase of about 279.2% this year; SK Hynix's listed stocks in Korea have accumulated a cumulative increase of about 186.0%.
How sacred is Solidigm owned by SK Hynix?
Solidigm is an enterprise-level data storage company owned by SK Hynix and headquartered in the US. Its core business is solid-state drives (SSD) and supporting storage technology based on NAND flash memory, focusing on data centers, cloud computing, and edge AI. It operates in the form of an independent operating subsidiary and is headquartered in Rancho Cordova, California. According to its official website, it has 13 business locations and more than 2,000 employees worldwide.
Its business base comes from Intel's original NAND flash memory and SSD business. SK Hynix announced in 2020 that it would acquire related businesses at the initial agreed total consideration of about 9 billion US dollars, completed the first phase of settlement in December 2021, and established SoliDigm to undertake product development, manufacturing and sales of the original Intel SSD business; the second phase of delivery of the remaining NAND technology and manufacturing business was completed on March 27, 2025. As a result, Solidigm inherits Intel's long-standing enterprise storage technology, engineering team, and customer relationships.
Specifically, it delivers a complete enterprise-grade SSD product to customers, and the value of a complete SSD comes from the collaboration of flash media, controller, firmware, and system design. NAND is responsible for storing data even after power failure; controllers and firmware are responsible for arranging data reading and writing, error correction, wear management, and performance scheduling; enterprise-grade products also need to meet requirements such as continuous operation, data integrity, writing durability, and stable response time. As a result, Solidigm's business capabilities cover storage hardware, firmware, and ancillary software, and optimize products around the customer's actual workload.
Solidigm's main business needs to be clearly distinguished from SK Hynix's current main business, the HBM business. HBM is a high-bandwidth DRAM, which mainly provides high-speed data access during operation for accelerators such as GPUs; SoliDigm's core product is a NAND flash memory storage system, which is used to store large amounts of data and undertakes partially reusable inference caches under an appropriate software architecture. SK Hynix Group also covers DRAM, HBM, NAND, and SSD businesses. Solidigm represents an important enterprise-grade flash storage platform, and the entire group's NAND memory chip business cannot be classified as SoliDigm
The more advanced the performance of AI agents, the more they can work, and the more large-scale expansion of memory chips
The core change brought about by Muse and Astra is that a user command can start a multi-stage, sustainable workflow. Meta revealed that Muse runs on a dedicated secure virtual machine and can perform tasks across applications; Astra enhances computer operation, programming, and complex professional work capabilities. A research or development task may continuously trigger data retrieval, file reading, code execution, model inference, and result verification, and produce intermediate results that need to be preserved.
Derived from the engineering architecture, GPUs and dedicated AI accelerators undertake model computation, and high-performance CPUs undertake execution and scheduling of browsers, virtual machines, and tools; enterprise knowledge bases, data and indexes, working files, and audit records required for enhanced retrieval (RAG) generation (RAG) expand memory and persistent storage requirements. The penetration rate of smart devices is increasing, so it is expected to boost both “computing power” and “data processing ability.”
The second increase in storage requirements comes from the need for cache management due to longer contexts and more concurrent tasks. In mainstream Transformer inference architectures, processing input is pre-filled, output is gradually generated by decoding, and a key-value cache (KV Cache) saves an intermediate calculation state that can be reused. The high-frequency data required for active generation is carried by HBM, and the system DRAM is responsible for buffering. Caches suitable for reuse can be layered into SSD and shared flash memory according to access frequency and latency requirements, and then loaded back into memory in advance. The Nvidia CMX architecture has been clearly proposed to add a context-oriented inference layer of flash memory between GPU memory and traditional shared storage. Its economic significance is to expand the amount of contextual capacity that can be retained and reused, and reduce repeated computation and data waiting, thereby supporting more concurrent tasks. As a result, enterprise-grade SSDs have gained additional application space to participate in the inference operation process.
Solidigm's high-density products are specifically linked to these needs. The D5-P5336 has a maximum capacity of 122.88TB. Using QLC technology, it is mainly aimed at high-capacity, read-intensive workloads such as data lakes and object storage. For data center operators, higher single disk capacity helps reduce the number of devices required to achieve the same capacity and optimizes rack space, power supply, and cooling expenses; loads requiring higher write performance or more stringent response latency are handled by other matching SSD products and software configurations.
From an investment perspective, SoliDigm's strong growth opportunities come from the expansion of AI data scale, enterprise storage configuration upgrades, and customers' continued pursuit of cost per capacity and system efficiency. Improving model efficiency can also reduce the cost of completing tasks and attract more work into large-scale AI inference systems; when the scale of users and tasks exceeds the resource savings per task, the demand for computing, storage, network, and electricity can continue to grow simultaneously.