Data tracking prices across three major storage categories—memory, HDDs, and SSDs—indicates that the current supply crunch is far from over. In mid-June, HDD prices saw a significant increase, with memory and SSD prices also rising.

Latest figures show memory prices remain at extreme highs, with a slight increase from the previous month. Using September 2025 as a baseline, the average price for DDR5 memory has risen by 288%, up from 282% a month ago.

The increase for HDDs is even more striking. Last month, the average HDD price was 88% above the baseline, but this month it has broken through the 100% mark to reach 107%.

SSDs are the only category among the three to remain relatively stable this month, yet the absolute increase is still substantial. The average price for a 1TB NVMe SSD is 112% above the baseline. For example, the Kingston NV3 1TB, which cost around 50 euros in September 2025, now costs close to 140 euros.

Increases for higher-end 2TB and 4TB models are more moderate but remain well above pre-crisis levels.

Analysis suggests the core drivers of this HDD price surge are twofold. First, AI data centers have dramatically increased procurement of high-capacity drives, with massive demand for cold data storage shifting from SSDs to more cost-effective HDDs. Second, major manufacturers are actively reducing production capacity, widening the supply-demand gap and squeezing availability for consumer-grade products.

Financial firm Aletheia Capital has significantly raised its price forecasts for DRAM and HBM memory in a recent report. The firm believes memory is becoming the most critical component in AI hardware systems, with its overall value share expected to rise from around 40% in 2025 to over 70% by 2027.

Aletheia expects the average selling price for DRAM to increase by 30% in Q3 2026, far above its previous estimate of 10-15%. It forecasts a further 10-15% increase in Q4 2026, consistent with prior predictions.

The firm also projects that HBM average selling prices could double annually through 2027, primarily due to exponential AI-driven demand growth that is outpacing the expansion capabilities of the three major suppliers relative to orders from companies like NVIDIA.

Data shows that global annual demand growth for DRAM and NAND over the past two years has been as high as 45% and 38%, respectively. However, overall industry capacity supply growth has remained at just 16-17%, leading to a continuously widening supply-demand gap.

Simultaneously, major cloud service providers, including Microsoft, Google, Meta, and Alibaba Cloud, are securing stable supply for their computing power businesses by signing 3-5 year long-term supply contracts. These agreements lock in capacity and pricing through advance deposits.

This large-scale, long-term contracting has significantly reduced the volume of goods available on the spot market, further intensifying supply tightness for end-market customers.

Looking at new capacity additions, key new production lines, such as Micron's ID1 facility, are expected to start production at the earliest by mid-2027. Furthermore, semiconductor production lines typically require 12-18 months for ramp-up and stable mass production, meaning full-scale output and volume supply may not be achieved until 2028.

For the memory industry, a period of highly certain super-cycle prosperity has arrived. Goldman Sachs points out that the global DRAM supply-demand gap for 2026-2028 is projected at 5.0%, 5.9%, and 3.9%, respectively.

For NAND flash, the gaps are forecast at 4.4% and 4.6% for 2026 and 2027, with a slight narrowing to 3.0% in 2028. HBM memory faces the most pronounced shortage, with projected gaps of 5.4%, 6.0%, and 4.3% for those years.

Overall, 2027 is likely to be the year of the most acute supply-demand imbalance in this memory cycle.

Morgan Stanley recently published a report stating that, driven by sustained growth in AI data center demand, supply tightness across the entire hard drive and storage industry chain is exceeding expectations. The shortage situation may persist until at least 2028, breaking previous expectations of relief by 2027.

On the demand side, an AI server uses 8-10 times more DRAM than a standard server, and over 3 times more NAND and enterprise-grade storage. The scaling of large model inference and AI agents is expected to drive explosive growth in storage demand.

Key Stocks to Watch

SMIC (HKG: 00981): The company's core products span multiple areas, including logic chips, memory chips, and analog chips. First-quarter sales revenue was $2.5055 billion, a year-on-year increase of 11.5%; net profit was approximately $197 million, up 5% year-on-year. For the second quarter, the company provided revenue guidance of 14% to 16% sequential growth, with a gross margin guidance of 20%-22%.

HUA HONG GRACE (HKG: 01347): Its memory business is divided into two main lines: embedded non-volatile memory (eNVM), which is the foundational business and contributes the highest revenue share, and standalone NOR Flash, which represents a high-growth second curve. First-quarter sales revenue reached $660.9 million, a year-on-year increase of 22.2%; net profit was $20.9 million, up 458.1% year-over-year and 19.9% sequentially.

GIGADEVICE (HKG: 03986): A leading domestic NOR Flash design company with the second-largest global market share. It is actively building a diversified product portfolio, with businesses covering NOR Flash, SLC NAND Flash, niche DRAM, MCUs, analog chips, and sensor chips. Central China Securities believes the company is a leader in domestic niche memory and MCUs. With the memory cycle continuing its upward trend, the company's niche DRAM, SLC NAND, and NOR Flash products are experiencing both volume and price increases. The company is actively developing customized storage solutions and has made smooth progress in expanding its customer base across multiple sectors, with revenue contributions expected in 2026.

MONTAGE TECH (HKG: 06809): As a leader in the memory interconnect chip market, the company is transitioning from a memory interconnect specialist to a platform-based interconnect chip company. In the AI era, leveraging its long-term focus on key high-speed interconnect chips between computing power and storage, it has developed a rich product portfolio and holds a leading position in the global market. Its PCIe Retimer products have entered the volume shipment stage. The company is extending along a similar path into the higher-value PCIe Switch domain, potentially opening up new growth avenues. As AI shifts from model training to inference, particularly with the development of AI agents, demand for server CPUs is rising steadily, positioning the company as a primary beneficiary of strong CPU demand.