TradingKey - AI infrastructure buildout in the U.S. is entering an unprecedented phase of expansion. As tech companies continue to expand compute capacity to satisfy growing demands for artificial intelligence training, inference, and agentic workloads, both the number of large-scale data centers and planned investment amounts are surging in tandem.

According to a report by The Kobeissi Letter, the U.S. currently has approximately 90 operational data centers with a power capacity exceeding 100 megawatts, reaching a record high; if all currently planned projects materialize, this number could grow to around 280 by 2030.

Market data shows that tech companies are expected to spend approximately $7 trillion on data centers by 2030; Amazon, Microsoft, Alphabet, and Meta have signed over $1.5 trillion in data center lease agreements since the start of the AI boom, including commitments that have not yet taken effect. The scope of construction has long extended beyond hyperscale cloud vendors themselves, with semiconductors, networking equipment, servers, power, cooling systems, and data center operators all expanding capacity for higher compute demands.

In the AI data center supply chain, storage is an irreplaceable link. AI accelerators require high-bandwidth memory for data transfer, while training and inference systems need high-capacity DRAM and enterprise SSDs to store models, data, and intermediate results. Among them, Micron (MU) is positioned across DRAM, HBM, and NAND; SanDisk (SNDK) places its main bet on NAND flash memory and enterprise data center storage. Both companies benefit from AI storage demand, but what investors truly need to evaluate is: is this a diversified platform better suited for allocation, or a pure-play NAND cyclical stock with higher elasticity?

What Roles Do Micron and SanDisk Play in the AI Storage Chain?

Micron is one of the world's major memory chip manufacturers, adopting a vertically integrated IDM model with a product portfolio spanning DRAM, HBM, and NAND. DRAM and HBM primarily serve servers, AI accelerators, and high-performance computing, with HBM directly impacting GPU data transmission efficiency; NAND, meanwhile, is more commonly used in enterprise SSDs, client devices, and other storage equipment.

SanDisk, on the other hand, is a pure-play NAND manufacturer that re-listed independently after spinning off from Western Digital, focusing on enterprise SSDs, data center storage, and edge storage. The company secures wafer supply through a joint-venture manufacturing system with Kioxia, combining it with in-house packaging and testing, controllers, and product solutions to form a manufacturing model of "joint-venture wafers + proprietary products." Compared to Micron's full-stack IDM approach, SanDisk carries a relatively lighter capital expenditure burden, but its business concentration is higher: NAND pricing, enterprise customer demand, and data center orders will more directly determine its earnings performance.

Can the AI Memory Supercycle Last? Key Lies in Supply, Not Just Demand

The market has now widely accepted that AI will drive greater memory and storage demand, but what truly determines the earnings sustainability of the two companies is not the number of data centers itself, but whether storage supply can keep pace in a timely manner, and whether new demand can translate into favorable pricing and contract terms.

Micron's supply logic is more geared toward high-end products. HBM, AI server DRAM, and high-capacity enterprise SSDs impose higher requirements on process technology, packaging, validation, and customer collaboration, leading to longer capacity expansion cycles. The company disclosed that it has signed multiple strategic customer agreements covering approximately 20% of DRAM shipments and about one-third of NAND shipments, with some agreements including financial commitments from customers. If these agreements are executed as planned, Micron's future revenue visibility could be higher than in traditional memory cycles, though at the cost of requiring significant upfront capital expenditure.

SanDisk's core changes lie in its customer structure and contract models. The proportion of revenue from data centers has grown rapidly, and eight NBM agreements provide baseline support for future shipments. The company also emphasized that AI inference workloads feature higher storage density, which could drive sustained increases in demand for enterprise SSDs. However, SanDisk lacks HBM and DRAM businesses; when NAND enters a period of loose supply, its earnings downside risk will be more concentrated than that of Micron, which has a more diversified product portfolio.

Therefore, the impact of this round of AI buildout on the two companies is not entirely identical: Micron profits more from structural upgrades to high-end memory and product mix improvements, whereas SanDisk profits more directly from tight NAND supply and demand, rising prices, and the expansion of data center orders.

Which Is the Better Buy? Micron Wins on Certainty, SanDisk Wins on Flexibility

Both Micron and SanDisk benefit from the memory market upcycle and the expansion of AI infrastructure, but their investment logics differ. Below is the financial performance comparison table for Q3 FY2026:

Micron's competitive advantage lies in its 'full-stack memory' offering and deeper data center exposure. Covering HBM, server DRAM, and enterprise SSDs simultaneously, it benefits from both the increase in AI computing bandwidth and the expansion of data storage. Strategic customer agreements also offer an opportunity to dampen the traditional memory industry's cyclical volatility of 'booming during upswings and plunging during downswings.' The core investment logic for Micron hinges on determining whether growing AI demand can offset high capital expenditure and allow high-end products to serve as a new profit center.

SanDisk's advantage lies in higher operational leverage during NAND market upcycles. Its data center business remains in a rapid ramp-up phase, with NBM agreements and enterprise SSD demand helping enhance revenue visibility. However, its pure-play NAND model makes it more sensitive to changes in prices, inventories, and supply. The core investment logic for SanDisk hinges on determining whether the current NAND tightness can transition from short-term price hikes into a multi-year structural supply-demand deficit.

In simple terms, Micron is more like a 'full-stack AI memory platform,' with its edge in business portfolio and certainty, while SanDisk acts like a 'pure-play NAND cyclical leverage tool,' featuring strong earnings upside during price rallies. For investors who can choose only one stock, those who prioritize medium-term stability and diversification should focus on Micron first; those who seek upside potential and can tolerate higher risks of a market turn should consider SanDisk.

Find out more