Artificial intelligence not only changes software and computing power, but also places completely new demands on the storage infrastructure. At Innovation Day 2026, Western Digital showed how consistently the company is responding to this change. With a clear roadmap, new HDD technologies and an expanded platform approach, WD is positioning itself as a central infrastructure partner for AI applications. The focus is on scalability, cost-effectiveness and energy efficiency – factors that need to be increasingly considered together in the data-driven age.
- New storage roadmap with a focus on AI workloads
- HDD capacities from 40 TB today to 100 TB and more in the pipeline
- Significantly increased HDD performance as an alternative to expensive flash
- Platform API for simpler and more efficient storage provisioning
Western Digital in the age of AI: capacity growth to 100 TB and beyond
With the rapid increase in AI-generated data, the question of scalable storage capacity is taking center stage. Western Digital is pursuing a two-pronged technology approach consisting of ePMR and HAMR, which addresses both short-term and long-term requirements. Two hyperscale customers are currently qualifying UltraSMR ePMR hard disks with a capacity of 40 TB – an industry record. Series production in larger quantities is planned for the second half of 2026.
At the same time, HAMR-based HDDs are also undergoing customer qualification. This technology will be scaled up from 2027 and forms the basis for a leap in capacity to up to 100 TB by 2029. It is particularly noteworthy that WD plans to expand ePMR to up to 60 TB using HAMR innovations without increasing energy consumption. Both technologies are based on a common architecture, simplifying manufacturing and enabling a smooth transition for customers.
For operators of large data centers, this approach primarily means planning security. Capacities can be expanded gradually without having to fundamentally adapt existing infrastructure. There is no need for a forced technology change, while the scaling speed increases with rising data demand. Western Digital thus addresses one of the central problems of modern AI infrastructures: constant growth with controllable costs.
More performance, less energy: New HDD architectures for AI workloads
In addition to pure capacity, the performance of HDDs is also becoming increasingly important in the AI context. Western Digital presented two new technologies that are intended to significantly reduce the gap between classic hard disks and flash memory. The background to this is the high cost factor of flash, which is six to ten times more expensive than HDDs depending on the scenario and also has a limited service life.
The so-called High Bandwidth Drive technology enables simultaneous reading and writing via several heads on different tracks. This allows the data bandwidth to be doubled compared to conventional HDDs, and even increased by a factor of eight in the future – at minimal additional cost and without additional power consumption. The first systems are already being validated by customers.
WD also presented its dual pivot technology. It uses a second, independently operating actuator within a 3.5-inch HDD and is designed to double sequential I/O performance without sacrificing capacity or requiring extensive software modifications. In combination, both approaches achieve up to a four-fold increase in sequential performance. This means that the I/O performance per terabyte remains at the current level even with HDDs with 100 TB, which significantly reduces the need for additional SSDs. Dual-pivot HDDs are currently in the laboratory and are expected from 2028.
Energy-optimized storage and platform strategy as a bridge to hyperscale
Another focus is on energy-optimized HDDs for so-called “cold” AI data. Although this data is not used permanently, it must be available within a short period of time. They are too active for tape drives, but too expensive for classic flash or capacity HDDs. Western Digital closes this gap with specially optimized hard drives that significantly reduce power consumption while keeping access times below one second. Customer qualification of these drives is planned from 2027.
WD is also expanding its platform business. A new, open software layer with API should also enable companies with medium data volumes to access hyperscale economies from 2027. This will allow storage innovations to be introduced more quickly, qualification risks to be reduced and existing architectures to be retained. The focus is clearly on simplification and faster added value for customers beyond traditional hyperscalers.
Conclusion
With its new roadmap, Western Digital shows how storage technologies can be consistently aligned with the requirements of the AI age. Capacities of up to 100 TB, significantly increased HDD performance, energy-optimized designs and an open platform approach form a coherent overall picture. Many of the technologies presented are already undergoing customer validation, with the first series products to follow from 2026. WD is thus setting new standards for an economical and scalable AI storage infrastructure at an early stage.
