Storage Trends 2026 Storage Bottlenecks: WD Responds With HDDs, NVMe-oF And Decoupled Storage

From Manuel Christa Christa Manuel| Translated by AI 2 min Reading Time

Related Vendor

AI is eating up data, and WD expects new storage architectures. The manufacturer is focusing on HDD capacity, NVMe-over-fabrics via Ethernet and decoupled storage in 2026. For edge teams, this will depend on ports, protocols and power requirements.

Solutions for storage bottlenecks: Connect SSDs via NVMe-oF (left) or use inexpensive HDDs (right).(Image: Nano Banana / AI-generated)
Solutions for storage bottlenecks: Connect SSDs via NVMe-oF (left) or use inexpensive HDDs (right).
(Image: Nano Banana / AI-generated)

In an outlook, Western Digital (WD) names three building blocks for AI-capable storage: plenty of capacity on hard disks, storage via Ethernet fabrics and greater decoupling of computing power and storage.

WD refers to an IDC forecast to justify this: The volume of data generated worldwide is expected to grow from 173.4 zettabytes (2024) to 527.5 zettabytes (2029). For hyperscale and cloud data centers, WD also expects HDDs to account for "almost 80 percent" of the installed base by 2028. If the price of NAND flash for SSDs rises, good old magnetic hard disks will benefit—and vice versa.

NVMe-oF Tunnels the Memory Via the Network

WD also wants to move away from classic SAS paths towards NVMe-over-Fabrics (NVMe-oF). The approach not only transports NVMe via PCIe, but also tunnels the protocol over the network, such as TCP or RDMA. This is intended to better combine bandwidth and scaling when AI applications move large data streams.

It gets more specific where WD names platforms: The OpenFlex Data24 4000 series bundles NVMe SSDs in a 2U housing with 24 bays and can be connected via RoCE or NVMe/TCP, according to the data sheet. In practice, the benefits depend on the transport: RDMA reduces latency, but requires discipline in the network design, NVMe/TCP remains closer to standard Ethernet, but often costs CPU resources.

SMR And Disaggregation to Smooth Costs And Operation

For the hard disk part, WD continues to focus on high densities and mentions SMR as an option for large data pools. To ensure that SMR does not fail due to the "random write" problem, intelligent shelves and software are to bundle write accesses and translate them into sequential streams in the background.

The third topic mentioned by WD is disaggregated storage in general. Compute and storage should be able to grow separately instead of together in hyper-converged nodes. NVMe-oF is one option that makes this a practical reality. From the manufacturer's point of view, this reduces the risk of having to maintain expensive capacity or computing power simply because the other resource becomes scarce.

(mc)

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta)

Unfold for details of your consent