The humble hard disk drive (HDD)—long overshadowed by flash storage—is making a comeback, driven by the insatiable data appetite of artificial intelligence. Recent announcements from major HDD head manufacturers suggest that the industry will ramp up both unit shipments and overall storage capacity by the end of the decade, reversing years of declining volumes. This shift signals that even as AI workloads demand blazing-fast access, the sheer scale of data generation still requires cost-effective, high-capacity magnetic storage.
HDD manufacturers are investing in new production lines for recording heads, the critical components that read and write data. These investments indicate confidence that hyperscale data centers—the backbone of AI—will need more HDDs, not fewer, to store the trillions of parameters and petabytes of training data. While NAND flash and SSDs are indispensable for active processing, HDDs remain the cheapest option for bulk storage, especially for cold and warm data that doesn't need instant retrieval.
The announcement comes from suppliers like TDK and Headway Technologies, which produce the thin-film magnetic heads used in modern HAMR (heat-assisted magnetic recording) drives. These heads enable areal densities beyond 2.5 Tb/in², pushing single-drive capacities past 30 TB. Industry analysts at Coughlin Associates (the article's author) note that total HDD exabyte shipments could double by 2030, driven entirely by AI and machine learning workloads.
Why now? AI model training generates massive intermediate datasets. Inference services log every query for retraining and auditing. Video surveillance, autonomous vehicles, and scientific simulations all produce floods of data that must be stored economically. Cloud giants like AWS, Google Cloud, and Microsoft Azure are expanding their data center footprints, and they consistently buy HDDs in bulk. Western Digital and Seagate, the two dominant HDD makers, have already launched 30+ TB drives and are working on 50 TB+ models.
Key figures: According to a recent report by IDC, the global datasphere will grow from 120 zettabytes in 2024 to 291 zettabytes by 2027. HDDs currently hold about 60% of the world's data, and that share is projected to remain stable as flash growth fills higher-performance tiers. The head manufacturing ramp could add 10–15% more HDD units annually by 2028, reversing a long-term decline of about 4% per year in unit shipments since 2015.
Analysis: This trend challenges the narrative that SSDs will completely replace HDDs. While SSDs win on speed and power, they cost 5–10x more per terabyte. For AI data lakes where latency tolerance is minutes or hours, HDDs are economical. Moreover, HAMR and other advanced technologies are extending the HDD roadmap well beyond 2030. Investors have taken notice—Seagate's stock has climbed over 40% in the past year on AI storage optimism.
Outlook: By 2030, expect HDD unit shipments to stabilize or grow modestly, while capacity per drive soars. The bottleneck will be head supply and rare earth magnets, not demand. Watch for earnings reports from Western Digital and Seagate, and investment announcements from HDD head suppliers. If AI workloads continue to double annually, the HDD renaissance has only just begun.
Technology
↑ Positive
Hard Disk Drive Unit Shipments Could Grow To Support AI Workloads
Recent HDD head manufacturing announcements indicate that the industry will increase unit shipments in addition to overall storage capacity shipments by decade's end.
Thomas Coughlin, Contributor
Forbes
3 min read
5/10
Key Takeaways
- HDD unit shipments declined about 4% annually from 2015 to 2024, but new head manufacturing investments could reverse that trend by 2028, with annual unit growth of 10–15%.
- Head manufacturers TDK and Headway Technologies have announced capacity expansions for HAMR (heat-assisted magnetic recording) heads, enabling areal densities beyond 2.5 Tb/in².
- Global datasphere is projected to reach 291 zettabytes by 2027, with HDDs still holding ~60% of stored data, according to IDC.
- Seagate and Western Digital have launched 30+ TB HDDs and are developing 50 TB+ models specifically targeting AI data lakes.
- AI workloads such as model training, inference logging, and video analytics generate massive intermediate datasets that favor HDDs for cost-effective bulk storage.
Frequently Asked Questions
HDD shipments are expected to grow because AI workloads generate massive amounts of data that needs to be stored cost-effectively. HDDs offer the cheapest cost per terabyte, making them ideal for bulk storage of training data, inference logs, and archived datasets.
Analysts project that HDD unit shipments could grow 10–15% annually by 2028, driven by new head manufacturing capacity and increased demand from hyperscale data centers for AI storage.
HAMR stands for heat-assisted magnetic recording. These heads use a laser to heat the disk surface, allowing data to be written at much higher densities. New HAMR head investments enable HDD capacities to exceed 30 TB and eventually reach 50 TB, which is critical for AI data lakes.
SSDs are preferred for high-performance tiers like hot data and active AI model training, but HDDs remain dominant for bulk and cold storage due to their much lower cost per terabyte. Both technologies coexist in modern AI data centers.
Seagate and Western Digital are the two major HDD manufacturers directly benefiting. Head suppliers like TDK and Headway Technologies also gain from increased production orders. Cloud providers such as AWS, Microsoft Azure, and Google Cloud are the primary buyers.
AI model training generates terabytes of intermediate data per run. Inference systems log every query for monitoring and retraining. Additionally, video analytics, autonomous driving, and large language model scaling all contribute to exponential data creation that requires economical storage.
Topics
Original source
www.forbes.com
Discussion
Join the discussion
Sign in to post a comment or reply.
No comments yet. Be the first to share your thoughts!