The AI boom has sent shockwaves through the storage market, with hard drive prices surging. Western Digital revealed it has no HDDs left for 2026, with contracts already signed for 2027 and 2028. Seagate reported similar difficulties, making it clear that storage options for consumers are shrinking. The impact of this shift is evident not only in the rising cost of consumer drives but also in the depletion of long-term inventory. Both companies have acknowledged the strain on supply, which signals a prolonged period of unavailability for standard HDDs. The situation is particularly dire for consumers planning to build or expand their storage solutions. While the AI-driven demand has shifted focus toward DRAM and other components, the ripple effects are felt across the entire storage ecosystem.
Shifting storage strategies
Building a NAS has become more expensive, but not impossible. The price-per-terabyte for smaller drives like 1 TB, 2 TB, and 4 TB has risen sharply. Larger drives now offer a more cost-effective alternative, though they sacrifice some redundancy and recovery flexibility. Historically, NAS builders often preferred smaller drives to spread risk and redundancy more evenly. However, with the current market dynamics, buying a few high-capacity drives is often more economical than purchasing multiple smaller drives. The trade-off is the potential for more significant data loss in the event of a drive failure and longer recovery times during resilvering. For instance, a 20 TB drive failure can disrupt a larger portion of the storage array than a 4 TB drive. This change in cost structure forces users to adjust their strategies to minimize the financial and functional impact of a single point of failure.
Redundancy in a NAS setup is now more costly. In a two-way mirror setup, redundancy accounts for half the cost. For example, expanding a mirrored pool requires buying drives in pairs, and each pair only adds one drive’s worth of usable space. This method has become the most expensive for adding a terabyte when using ZFS. With drives doubling in price, the cost of redundancy has also doubled, making mirrored setups significantly more costly. If a user expands their setup by adding two drives, they effectively only gain the storage capacity of a single drive due to the redundancy requirement. This is particularly impactful in environments where efficiency and cost are critical, especially in scenarios involving ZFS, where mirrored pools are the norm. The result is a storage configuration that, while safe, comes at a higher financial cost.
Enterprise drives and flexible RAIDZ
Used enterprise drives offer a better alternative during the storage crunch. Recertified Ultrastar and Exos drives, for instance, maintain value despite the price surge. These drives, often tested and re-warranted by sellers, are more affordable and reliable than retail options. The disparity between consumer and enterprise drives has never been more pronounced. While consumer drives are priced under pressure from AI-related demand and limited supply, enterprise drives—especially used or recertified ones—have managed to remain relatively stable. This is partly because enterprise storage markets have a different demand structure, with companies prioritizing long-term reliability over immediate price points. For example, drives like the Dell-badged Seagate Exos 7E8 SAS offer both affordability and performance when purchased in the used market. Combined with the right hardware, like an HBA card, users can achieve a cost-effective and reliable storage setup that avoids the volatility of the retail market.
RAIDZ expansion in OpenZFS 2.3 now allows a single drive to be added to an existing RAIDZ vdev. This approach is particularly useful given the current storage market conditions. While it doesn’t provide immediate full capacity, it supports flexibility, and over time the usable space will converge as data gets rewritten. This feature, which became available in Proxmox 9, allows users to expand their storage incrementally. However, the expansion process is not straightforward. The blocks of data written before the expansion retain the old data-to-parity ratio, so the initial usable space gained is less than expected. As data is rewritten, the new parity layout spreads across the expanded array, gradually increasing available space. This method offers a pragmatic solution in an otherwise difficult market. For users considering future expansions, starting with RAIDZ setups makes more sense, as traditional plain ZFS configurations cannot be converted later.
The storage crunch has also highlighted the importance of adaptability in storage planning. New retail drives are heavily priced due to their competition with hyperscaler orders, which prioritize bulk purchases at the expense of retail availability. In contrast, the used enterprise drive market has proven more resilient, offering better value and performance for individual users. This has led to a shift in how people approach storage solutions, with more attention being paid to cost efficiency and long-term reliability. As the market evolves, users must remain flexible in their choices, balancing the need for data safety with the realities of supply and demand. In a world where storage options are limited and expensive, the right strategy can make all the difference.