NVIDIA's Vera Rubin AI systems projected to consume 9.3% of global NAND supply by 2027

NVIDIA Vera Rubin Systems Set to Drive Massive NAND Demand

NVIDIA's next-generation Vera Rubin AI systems are poised to consume staggering amounts of storage capacity, potentially triggering a NAND supply shortage that could reshape the memory market. According to a Citi analysis, NVIDIA's Vera Rubin AI systems could require approximately 1,152TB of additional SSD NAND per server system to support the company's Inference Memory Context Storage (ICMS) operations1

1

. With projected shipments of 30,000 units in 2026 and 100,000 units in 2027, NAND demand driven by ICMS is expected to reach 34.6 million TB in 2026 and 115.2 million TB in 20272

2

. This translates to 2.8% of expected global NAND demand in 2026 and 9.3% of global NAND supply in 2027, a scale that Citi warns "is likely to create significant upside potential for demand"1

1

.

Understanding the Storage Architecture Behind the Demand

The surge in storage requirements stems from a fundamental shift in how agentic AI environments process data. Query processing in these systems generates massive temporary memory logs for context building called the KV Cache, which currently relies on HBM modules2

2

. As data requirements within AI clusters scale rapidly, HBM can no longer hold the necessary capacity onboard. At CES 2026, NVIDIA announced that Bluefield-4 DPUs would connect to a new storage solution called Inference Memory Context Storage (ICMS), designed to address this bottleneck2

2

. Each Vera Rubin system could utilize approximately 16TB of NAND per GPU in a rack, totaling 1,152TB in a single NVL72 configuration1

1

. While this significantly boosts data processing capabilities, it creates unprecedented demand for NAND SSDs that the industry hasn't fully anticipated.

Storage Shortages and Market Impact on the Horizon

The NAND supply shortage could intensify dramatically as this structural demand increase takes hold. Industry experts warn that the situation mirrors the ongoing DRAM crisis, with storage shortages potentially becoming the next major supply chain disruption1

1

. The NAND industry already faces pressure from ongoing data center buildout and the inference craze, and NVIDIA's acquisition of a significant portion of total global NAND output could create a supply shock that manufacturers haven't factored into their production forecasts2

2

. For average consumers, accessing general-purpose SSDs and storage devices could become increasingly difficult and expensive. Industry contacts predict that after the memory crises involving DRAM, SSD price increases are inevitable as M.2 SSDs and other consumer storage products skyrocket in cost1

1

. The projections don't even account for Rubin Ultra or NVIDIA's next-generation Feynman GPU architecture, suggesting that demand pressures will only intensify beyond 2027. Given that NVIDIA has positioned agentic AI as the primary focus for future applications, the need for sufficient KV Cache pools in server systems means ICMS demand will continue climbing, potentially creating long-term structural changes in the storage market2

2

.