Soaring inference AI demand triggers severe nearline HDD shortages

TrendForce’s latest investigations reveal that the massive data volumes generated by AI are straining the global infrastructure of data center storage. Nearline HDDs, traditionally the backbone of large-scale data storage, are now facing severe supply shortages, pushing high-performance yet higher-cost SSDs into the market spotlight. In particular, shipments of high-capacity QLC SSDs could see explosive growth in 2026.

In the traditional tiered storage architecture of data centers, HDDs have long dominated cold data storage thanks to their exceptionally low cost per gigabyte. Cold data typically includes backup files and historical records—datasets that are rarely accessed but must be archived long-term. With the rapid expansion of inference AI applications, demand for cold data storage is surging.

SSDs, by contrast, are valued for their superior read/write speeds and are primarily used for hot and warm data that require frequent access. Compared with nearline HDDs, QLC SSDs not only have higher performance, but also deliver about 30% lower power consumption.

TrendForce notes that because major HDD manufacturers have not expanded production capacity in recent years, they are unable to keep up with the sudden, massive surge in storage demand driven by AI. As a result, lead times for nearline HDDs have ballooned from just a few weeks to over 52 weeks, further widening storage gaps for CSPs.

North American CSPs had already been planning to increase SSD adoption for warm data workloads. However, the severity of the current HDD shortfall has prompted some providers to consider deploying SSDs for cold data. Yet, large-scale adoption still faces two key hurdles: cost and supply chain constraints.

TrendForce emphasizes that if CSPs plan to use QLC SSDs for cold data storage, they will need to account for adjustments to data management algorithms, compatibility across software stacks, and precise calculations of total cost ownership. Maintaining a firm pricing threshold will be critical to achieving cost balance. 

For SSD suppliers, this wave of redirected demand presents an excellent opportunity to improve profitability. But, with limited capacity for high-capacity products, suppliers are unlikely to make significant price cuts. Consequently, a round of price negotiations between buyers and sellers is expected, with enterprise SSD contract prices projected to rise 5-10% QoQ in 4Q25.

Source: TrendForce, Taiwan.