Western Digital blows hard disk drive future wide open

Western Digital is qualifying 40 TB hard disk drives, has outlined a roadmap to 100 TB, and is announcing two major performance-boosting technologies alongside reduced power consumption and a software abstraction layer aimed at hyperscalers operating at 200-plus petabyte scale.

The company unveiled the developments at its Innovation Day 2026 event, saying it is reinventing the HDD for the AI era with a new generation of storage technologies spanning scalable capacity, performance optimization, power efficiency, and an intelligent platform API designed to improve economics at scale. It is also emphasizing the WD brand.

CEO Irving Tan said: “Today, we are showcasing innovation that reflects our deep connection to our customers and how we are meeting demand for capacity, scale, quality, enhanced performance, and ease of technology adoption.”

It’s qualifying the world’s highest capacity drive, a 40 TB UltraSMR drive using ePMR technology, at two hyperscalers, with volume production planned in the second half of this year.

UltraSMR is a shingled magnetic recording technology in which disk drive tracks partially overlap to increase the number of tracks and hence capacity, whilst lengthening data rewrite operations as entire zones of tracks have to be rewritten. Perpendicular Magnetic Recording (PMR) is the conventional way of recording data on disk drives, with WD’s ePMR technology strengthening the write signal effectiveness. WD is also developing Heat-Assisted Magnetic Recording (HAMR) technology, as Seagate has done, to overcome PMR recording limitations as bit sizes shrink on disk to increase capacity.

The company has also built a 30 TB conventional, not shingled, ePMR drive, with Shihab saying: “We are extending that roadmap a little further. So we think, we know, that we can get to 50 and little beyond that on ePMR.”

WD is also qualifying HAMR drives at two hyperscalers with volume production starting, it says, in 2027. The company says it can extend its ePMR capacity out to 60 TB “by leveraging HAMR innovations without increasing power consumption.” There are no details as to how it will achieve this. Its HAMR and ePMR drives are built on a common architecture, providing manufacturing efficiencies, and can be mixed and matched in the same chassis as their interfaces are the same.

HAMR drives have the longer capacity roadmap, out to 100 TB vs ePMR’s 60 TB, and customers can transition from ePMR to HAMR at a relaxed pace. Chief product officer Ahmed Shihab said: “What we’ve done is make the performance and capabilities the same between ePMR and HAMR so that customers can put them side by side and use them at the same time. They get very comfortable with this plan. They’re very, very comfortable, and that parallel roadmap is really appealing to them.”

Quadrupling HDD bandwidth

WD has so-called High Bandwidth Drive (HBD) technology, taking advantage of having read/write heads for each platter surface whereby there is simultaneous reading and writing from many heads on many tracks. Initially this will provide an up to 2x increase in bandwidth, with a roadmap to 4x, 6x, and 8x. This means the I/O density of a drive, the ratio of its capacity and bandwidth will improve, escaping the traditional HDD I/O density dead-end, and without any increase in electricity draw. Customers are qualifying this technology.

Most nearline drives have a 250-275 MB/s I/O rate. A non-shingled Western Digital 26 TB Ultrastar drive with 11 platters has an up to 302 MB/s I/O rate. Shihab tells us customers with QLC flash SSDs can get “500 and something megabytes out of those drives.” With the High Bandwidth Drive tech, “you can get 500-plus MB/s.”

He said: “There are modifications inside the drive, but we’re not redoing anything to the outside. So same interface, same API, same set of commands.” It’s like adding a turbocharger to the I/O channel. We can envisage future WD drives with theoretical 4x, 6x, and 8X bandwidth increases having – the numbers seem far-fetched, ridiculous even – 1,000, 1,500 and 2,000 MB/s, if the technology delivers on its promise.

The company is going further by adding a second and independent set of read:write heads, actuators, on a separate pivot inside a 3.5-inch drive’s enclosure, and delivering another up to 2x increase in bandwidth. Its dual pivot engineering design actually reduces the space between the drive’s platters, enabling additional platters to be added to increase capacity. WD currently uses 11 platters, versus Seagate’s 10, and Toshiba has demonstrated 12-platter technology.

HDDs with dual pivot technology are currently in the lab and will become available in 2028. There have been dual actuator experimental developments in the past. Shihab told us: “It’s not the same way it was built before, where you had effectively two drives with a SAS interface or two SATA interfaces that was really hard for customers to use. This is a drive you can plug into your existing infrastructure, same software, and just double the I/O.”

Combining dual-pivot actuators with the initial 2x High Bandwidth Drive technology should provide an up to 4x increase in drive bandwidth “delivering 100 TB HDDs while maintaining the relative I/O per TB rate customers enjoy today.

The dual-pivot and High Bandwidth Drive tech applies to HAMR drives as well as ePMR disks. And, we understand, if we get 4X High Bandwidth Drives (1,000 MB/s) then dual-pivot technology would take that to 2,000 MB/s. You can work out the implications of dual-pivot technology on 6x and 8x High Bandwidth Drives yourself. SATA link saturation beckons.

Comment

This breaching of the HDD I/O density wall by WD completely resets the relationship between disk drives and SSDs, removing SSD’s bandwidth superiority, though not their latency advantage. If WD delivers 2x High Bandwidth Drives while SSD supplies are in shortage and priced high then it should find a ready market with SSD capacity-starved hyperscalers and enterprises needing SSD-class bandwidth nearline storage capacity.

Also, if WD can do this, then competitors Seagate and Toshiba will fall in line behind it as fast as they can, and disk drive tech gets much more interesting than bland, few-MB-at-a-time, incremental developments.

Why hasn’t it been done before? Shihab tells us: “We keep asking ourselves the same question, but the thing is, nobody knew that it was possible when customers didn’t know to ask for it. But when we sat down and talked to a bunch of my friends and various hyperscalers, they were talking about QLC and what you get, and they were frustrated by the SATA (interface). I said, “Wait, if I can double the performance, is that interesting?” They said, “Yeah.” When I talked to the engineers, they went, “Oh yeah, we can do it this way.”

It’s as if HAMR technology development has had such a grip on HDD technology developers that they were blind to other possibilities.

One more point. The drive’s internal controller now has to do more work. It has to have more processing power because it’s is getting more bits thrown at it. The error correction, the signal detection, amplification, decoding, head positioning, etc., everything’s got to work twice as fast one way (High Bandwidth Drive), and then twice as fast again another way (dual-pivot actuators).

Power-optimized drives

Another technology development are power-optimized drives which “reduces power consumption, and therefore customer operating costs, while maintaining a sub-second access storage tier.”

It says: “These drives will trade minimal random I/O for higher capacity and substantially lower power… shrinking the gap between warm and cold storage tiers and enabling economically sustainable AI data storage at scale.”

If we envisage tape as cold storage and current nearline disk drives as warm storage them power-reduced HDDs will be room temperature storage, an intermediate tier, offering nearer tape power consumption and a little slower than current disk latency and bandwidth.

WD is not releasing details of its power-optimized drives but we think it could involve slower spin speeds.

Simplified JBODs

WD supplies Ultrastar Data60 and Data102 multi-drive chassis, JBODs (just a bunch of disks) with up to 102 SAS-interface HDDs inside their 4 RU enclosure, effectively a SAS expander JBOD. These have multiple in and out-of-band options for monitoring, configuration, and control: Redfish-based SPI, IPMI, resource manager and SCSI enclosure services.

Shihab says that large enterprises, below the hyperscalers in size, “don’t have the resources or the years it took the hyperscalers to really take advantage of all of these capabilities from UltraSMR to High Bandwidth Drive, to dual pivot drives, to the 100 TB drives.”

WD wants to make it easier to use its JBODS, with Shihab saying: “We’re adding capability to those to enable them to abstract away all that complexity. So we put in a very simple API that can be plugged into whatever storage software they want, and then we can do the qualification and the work necessary to abstract all of this complexity for them so they can get the capabilities, the performance, the capacity very quickly, as soon as we can deliver them.”

The company is adding an intelligent software layer, through an open API, that will enable companies at 200+ petabyte scale to achieve the same storage efficiency and economics that hyperscalers enjoy today. It will accelerate storage innovation adoption across WD’s UltraSMR, ePMR and HAMR HDD, and flash platforms, and is expected to launch in 2027. The claim is that this will give up-and-coming cloud and enterprise WD customers a clear path to hyperscale economics.

Overall, Shihab said: “WD is challenging conventional storage assumptions and removing the complexity and cost barriers that limit their AI-driven growth. Our capacity, performance, power efficiency, and platform innovations solidify our position as the innovation partner for the AI-driven data economy.”

Comment

WD is blitzing attendees with near and far-reaching announcements at its Innovation Day 2026 event. It is resetting the balance between SSDs and HDDs by adding huge I/O bandwidth increases to disk drives that will enable them to match SSD bandwidth, though not latency, and ensure disk drive storage relevance for decades to come.