Seagate Mosaic 3+ hard drive
Seagate announced shipments of sample Exos M hard drives, HDDs, to customers with storage capacities up to 36 TB. These HDDs use the company’s Mozaic 3+ heat-assisted magnetic recording technology, HAMR, where lasers are used to reduce the magnetic fields from the write head required to write a high density on stable magnetic recording media by clamping high. Seagate also said it is ramping up production of Exos M HDDs up to 32TB to a leading cloud service provider.
These 36 TB HDDs have 10 drives and so they offer 3.6 TB per drive. This will likely result in about 1.8 TB of areal density per square inch, a 13% improvement over previous 32 TB HDDs. The chart below shows Coughlin Associates’ HDD areal density shipping history. This chart shows a significant increase in HDD areal density in 2023 and 2024, and if this trend continues, it shows that HAMR allows for increased HDD areal density, and thus increased capacity per HDD, to increase up to 15- 20% per year. Seagate projects that 10 TB per drive is achievable, enabling a 100 TB HDD with 10 drives.
The history of increasing product density in the area of hard drives
Seagate says this product will offer significant advantages for data center storage in terms of scale, total cost of ownership and durability. The company says these drives can provide 300% more storage capacity within the same data center footprint, a 25% cost reduction per TB and a 60% energy reduction per TB compared to previous generation products, for a 30TB Exos M Mosaic drive compared to a 10TB Exos X10 HDD. Seagate says 10TB is a common drive capacity currently used in data centers and needs upgrades today.
A Recon Analytics survey, commissioned by Seagate and conducted in November 2024 with 1,062 participants, showed some interesting trends in how AI is creating increasing demand for digital storage, and thus the need for cost-effective storage such as HDDs. 61% of infrastructure buyers who primarily use cloud storage for AI data management said they expect storage requirements to at least double by 2028 due to an increase in retention times from 6 months to forever, 73% used daily or weekly LLM checkpoints and 80% considering replication data for highly or moderately significant AI.
95% of storage buyers using AI or planning to say they are taking steps to meet growing storage demands, including 61% adopting more scalable storage, 56% implementing data management software, 49 % using compression techniques and 55% improving the existing storage infrastructure. Recon Analytics research finds that wherever AI is adopted, existing conservation practices will need to improve to realize AI’s full potential.
Cloud storage is expected to remain the primary storage medium for AI with 65% of data stored in the cloud versus on-premises in 2024 and rising to 69% by 2028. 90% of respondents who have adopted AI believe that storage longer data improves the quality of AI results. Of which 93% claim that data storage requirements have changed due to the implementation of AI and the ability to improve models by incorporating checkpoints. The importance of data replication in a company’s AI data management strategy also increases the amount of storage a company uses.
Storing temporary modeling results is also driving storage demand. 73% of respondents say AI training is driving increased data retention as they are backing up their previously saved checkpoint data on a daily to weekly basis. Compounding the retention impact of maintaining AI checkpoints, infrastructure buyers also need to consider how long they will retain each checkpoint as part of LLM training. Of those surveyed who save checkpoints daily (28% of respondents), 32% keep data for more than 12 months while 29% keep it for six to 12 months.
Seagate announced that they are shipping HAMR 36TB evaluation drives to customers. These drives offer significant scale, power savings and storage cost advantages over previous generations of HDDs. A study shows that AI will drive demand for storage solutions such as these.