Toshiba is moving to increase production of HDDs for AI data centers. On October 2, the Nikkei reported that the company will invest about ¥60 billion to expand its Philippine plant and double production capacity within fiscal 2027. The investment responds to demand for storing the vast amounts of data used in AI, and Toshiba has already begun shipping samples of HDDs with capacities of up to 34TB. However, higher capacity per drive does not automatically bring the same gains in read/write speed or ease of adoption in existing systems. To understand what the expansion means, it is necessary to look not only at how many drives a factory can make, but also at how well customers can actually use those products in real systems.
The ¥60 Billion Investment and Changes in the Business and Finances
According to the Nikkei's report, the expansion of the Philippine plant is Toshiba's first major investment in its HDD business in about five years. In addition to adding production lines, it will support a new product with up to 40% more storage capacity per drive. The target date is within fiscal 2027, which in Japan's fiscal calendar runs through March 2028. Reading it as a plan to finish "by the end of 2027" would misstate the investment's timeline.
Manufacturing processes will also be reviewed alongside the expansion. According to the Nikkei, inspection work and work inside clean rooms will be automated, holding down the additional staffing that a normal expansion would require by about 40% while also aiming to improve yields. The 40% reduction applies to the new personnel the expansion would otherwise need; it does not mean cutting 40% of existing employees. Rather than adding staff in step with higher output, the aim is to handle the increase through improvements to equipment and processes.
Toshiba's own published financial materials also show changes that matter for understanding this investment. Its fiscal 2025 results materials explain that the HDD business performed solidly on the back of data center demand, with higher revenue translating into higher profit. In addition, at the end of March 2026, the company refinanced the loans it had arranged at the time of going private and relaxed contractual restrictions such as limits on investment. This has increased management's flexibility and put in place an environment in which future investments can be made more freely.
It cannot be said that this improvement in finances directly drove the ¥60 billion investment. Even so, the recovery in HDD demand and the easing of conditions that had constrained investment have coincided. The expansion plan can be seen as a sign that Toshiba is entering a stage in which it ties its development of high-capacity HDD technology to larger production capacity.
From 36 Drives to 30 for 1PB of Storage
Toshiba's M12 series, announced on March 31, 2026, shows how much higher capacity can affect the scale of equipment. With SMR (shingled magnetic recording), it achieves 30 to 34TB, and sample shipments to specific customers began at the end of March. The same series also includes CMR (conventional magnetic recording) models of up to 28TB, with sample shipments planned to begin in stages from the July–September quarter.
Comparing the number of drives needed based on rated capacity alone reveals a notable difference. To secure a rated 1PB, 28TB drives require 36 units, while 34TB drives require 30, a reduction of 6 units, or about 17%.
| M12 recording method | Rated capacity per drive used for comparison | Drives needed for 1PB or more |
|---|---|---|
| CMR | 28TB | 36 |
| Host-managed SMR | 34TB | 30 |
The calculation treats 1PB as 1,000TB, divides 1,000 by each drive's capacity, and rounds up to the nearest whole number. The reduction rate is 6 ÷ 36 × 100, or about 17%. The capacity figures come from Toshiba's March 31 announcement, and 34TB is the upper limit when the variable-capacity format is applied. The comparison uses only rated sample-stage values and does not account for redundancy, spare drives, or capacity lost to formatting.
With fewer drives required, more storage capacity can fit in the same enclosure or rack. This is why Toshiba aims to reduce total cost of ownership through higher capacity. However, the roughly 17% figure is only a calculated reduction in drive count, not a measured value showing that equipment costs or power consumption fall by the same proportion. Actual costs would also change with product pricing and the redundancy scheme used.
Furthermore, the 28TB and 34TB models use different recording methods. Whether the 34TB model's capacity can be used as is in an existing system also depends on how the host side manages the drives. What this comparison shows is the possibility of reducing the number of drives through higher capacity, not that simply swapping drives would cut the count while maintaining the same performance.
The Path to 40TB Runs Through Both Platter Count and Recording Method
The M12 fits 11 magnetic disks in a 3.5-inch enclosure, one more than the previous generation. It switches from conventional aluminum substrates to glass substrates, which can be made thinner, and combines helium filling with Toshiba's proprietary FC-MAMR (flux control microwave-assisted magnetic recording). The design fits more recording surfaces into the same size of enclosure while also raising the recording density of each disk.
In an announcement in October 2025, Toshiba further said it had successfully verified technology for mounting 12 disks. Combined with MAMR, the company plans to bring a 40TB-class HDD to market in 2027, and it is also looking at combining it with HAMR (heat-assisted magnetic recording) in the future. The successful verification of a 12-disk configuration, the M12's sample shipments, and the plan to launch a 40TB-class product are each at a different stage. The disclosure of the factory expansion does not mean mass production of 40TB-class HDDs has already begun.
SMR, meanwhile, overlaps parts of the recording tracks like roof tiles to write more data in the same area. While this increases capacity, random data rewrites can reduce transfer speeds. The M12 uses host-managed SMR, a mechanism that limits performance degradation by having the server or storage system efficiently manage data rewrites.
In other words, a manufacturer being able to produce high-capacity drives is a separate matter from customers being able to use that capacity effectively. Even as demand grows for storing large volumes of AI training data and logs, the same configuration cannot necessarily be applied as is to uses involving frequent small rewrites. How data is placed and how often it is updated will be key conditions when adopting high-capacity HDDs.
Caution is also needed with the power-saving figures. Toshiba says the M12 cuts power consumption per unit of capacity by about 18% from the previous generation, but this metric is W/TB, calculated by dividing active idle power consumption by storage capacity. It does not mean a data center's overall power consumption can be cut by 18%. Separating power efficiency per unit of stored capacity from the power needed for actual processing helps avoid either overstating or understating the benefits of higher capacity.
Aiming for a 30% Share: Competitors' Mass Production and Customer Qualification
Competitors are also pushing HDD capacities higher. On March 3, 2026, Seagate announced that it is shipping in volume its HAMR-based Mozaic 4+ drives of up to 44TB to two major cloud providers. It said general sales will proceed as production capacity expands, so this 44TB model cannot currently be compared as a product anyone can buy. Even so, it has moved from the evaluation stage with specific customers to volume shipments, a different stage from Toshiba's 40TB-class product today.
In Western Digital's (WD) February 3 announcement, two customers were evaluating a 40TB UltraSMR product, with mass production scheduled for the second half of 2026. For HAMR products, it plans to begin production in 2027. These are also plans as of the announcement date, and it cannot be assumed that they subsequently proceeded as scheduled. When comparing companies, it is necessary to line up not just maximum capacity but also how far customer qualification and volume shipments have progressed.
In the same announcement, WD also outlined technology that reads and writes in parallel using multiple heads, as well as a concept for adding independently operating actuators. As drive capacities grow, it becomes more important whether the large amount of data stored on a single drive can be accessed at sufficient speed. Each company is improving transfer bandwidth as well as capacity, so a product's usability cannot be compared by TB figures alone.
Toshiba's medium-term target reported by the Nikkei is to raise its capacity-based market share from the current level of just over 10% to 30%. However, doubling production capacity does not by itself double market share. Total shipped capacity in the market will also grow, and competitors will bring out high-capacity products of their own. It also has not been confirmed whether the reported "doubling of production capacity" refers to the number of drives or total storage capacity. For this reason, future supply cannot be estimated by multiplying in the capacity increase of the new products.
For the expansion of the Philippine plant to broaden procurement options, production lines coming online is not enough. High-capacity products must move into mass production and actually be adopted in customers' systems. If the production capacity expanded in fiscal 2027 translates, under those conditions, into a real increase in usable storage capacity, Toshiba's investment would strengthen the supply capability for long-term storage of AI data.