Hard Drives and AI: Seagate Explains Why It Won't Build New Factories Despite Record Demand
Last Monday, at the JPMorgan Global Technology, Media and Communications Conference, Seagate's CEO Dave Mosley responded to a direct question regarding the expansion of production capacity with a statement that the market interpreted as an admission of supply constraints: building new factories would "take too long." The stock closed the session down 6.9%, with intraday lows near 8%, marking the worst daily performance in nearly two months, dragging down Micron, SanDisk, and Western Digital, all falling around 5%. The following day, the pressure extended to Asia, with Samsung down as much as 5.3% and SK Hynix at 5.4%.
However, this information was not entirely unexpected: in an interview with Bloomberg earlier this month, the CEO had already presented the same reasoning, describing investment in new plants as incompatible with the dynamics of technological transitions.
Seagate's Choice: Technological Transition Instead of New Capacity
When asked to clarify what it would take to increase unit or surface capacity at the plants, Mosley outlined the company's industrial logic. "If we took the teams and put them to build new factories or to start new machinery, it would take us too long. We would end up with more capacity, but we would slow down the growth rate of technology," he stated during the conference.
The reference is to the transition to higher density platters: Seagate is moving from the current 3 terabytes per platter to versions of 4 and 5 terabytes, aiming for a compound annual growth rate of the exabytes produced around 25%. The path does not involve new plants but rather the gradual replacement of lines with more advanced machinery within the already operational plants.
The delivery times for critical components support this logic. For wafers destined for magnetic read heads, the manufacturer reports times exceeding nine months, plus a quarter for the production of the complete drive. The company has shifted to a made-to-order production model that provides four or five quarters of visibility on demand confirmed by customers. Mosley openly acknowledged that demand significantly exceeds current capacity.
The Spillover Effect on the Sector and the Distinction Between Storage and Memory
Seagate, as we know, is not a memory producer but a manufacturer of storage units based on hard drives. However, the market narrative has grouped HDD, NAND, and DRAM under the same category of AI infrastructure because data centers for generative workloads absorb both categories with related dynamics: high-capacity hard drives for dataset storage, high-speed memory for training and inference phases. Therefore, market reactions should not surprise: when a leading company in an adjacent segment reports supply constraints that will not be quickly resolved, investor sentiment extends to everything that is somehow related.
In terms of stock performance, Seagate gained about 600% in the twelve months preceding Monday's session, adding over $144 billion in market capitalization. The last quarterly report showed record gross margins driven by data center demand. The price growth in recent months has incorporated aggressive expectations for continued cycle momentum, which Monday's statement somewhat compromised, causing market reactions.
The key technology for Seagate's gamble is HAMR (Heat-Assisted Magnetic Recording), the thermally assisted magnetic recording system that enables the platter densities on which the entire reasoning is based. The Mozaic 3 HAMR platform is already qualified by all expected cloud providers, and the manufacturer aims to achieve 50% of shipped exabytes on HAMR in the second half of 2026.
Seagate's position has a linear industrial logic: each new plant would take teams and capital away from the transition to HAMR, and an excess of installed capacity would become a burden when platter densities render previous lines obsolete. On the market side, there is a supply compression intended to persist. The litmus test for measuring which of the two dynamics prevails is already indicated by the company itself: achieving 50% of shipped exabytes on the HAMR platform by the second half of this year.