At CES 2026, digital aficionados' conversations about artificial intelligence and memory chips have heated up, largely driven by the escalating prices of DRAM and NAND. NVIDIA's upcoming AI servers, built on the Vera Rubin architecture, are projected to substantially ramp up demand for SSDs and NAND flash memory in the years ahead, thereby tightening supply constraints even further. Based on Citi analysts' computations, if we take the configuration requirements of the Vera Rubin NVL72 AI server as a reference point and assume shipments of 30,000 units in 2026, escalating to 100,000 units in 2027, the NAND demand spurred by ICMS will hit 34.6 million TB in 2026 and skyrocket to 115.2 million TB in 2027. This would represent 2.8% and 9.3% of the total global NAND market demand, respectively. Such a scenario could open up space for market demand growth and worsen supply shortages. It's worth noting that this analysis doesn't yet factor in the extra demand from higher-specification systems and future architectures. Hence, the scaling up of shipments for the Vera Rubin platform could have a far-reaching influence on the future supply-demand balance of memory chips.