Why is Seagate leading the AI storage race? HAMR Mozaic reshaping the economics of data center storage

At the beginning of 2026, a set of data released by the global storage industry exceeded most people's expectations. Seagate (NASDAQ: STX) reported revenue of $2.83 billion in Q2 of fiscal year 2026, up 21.5% year-over-year, with an adjusted EPS of $3.11, up 53.3% year-over-year, both surpassing market expectations. Meanwhile, Western Digital's revenue grew 25% YoY to $3.02 billion, with net profit soaring 296% YoY. The two HDD giants delivered their first ten-year-high performance reports almost simultaneously.

But in most discussions about AI, the recurring mentions are GPU computing power, HBM memory, and Ethernet switches. Mechanical hard drives (HDD), a technology nearly seventy years old, seem far removed from the core narrative of this AI wave. Data provides another answer: Seagate’s data center business accounted for 79% of total revenue in FY2026 Q2, reaching $2.2 billion, up 28% YoY. AI has not bypassed HDD; instead, it is reshaping data center storage architecture in a deeper way—HDD is experiencing a structural demand growth driven by AI.

The Certainty Turning Point in Supply and Demand Structure

Understanding the core of Seagate’s growth this round requires returning to a fundamental question: Is the demand growth for HDDs sustainable?

From the supply side, Seagate CEO Dave Mosley explicitly stated during the earnings call that the company's near-line large-capacity HDD capacity until the end of calendar year 2026 is fully sold out, with customers concerned about “stable supply” rather than price, and orders booked into 2027. This indicates that the supply-demand relationship in the HDD market has shifted from early industry cyclical fluctuations to a structural tightness under supply constraints.

From the demand side, the driver of this change is the reconstruction of storage architecture driven by AI workloads. Massive datasets during training require long-term storage for continuous iteration, while inference-stage AI agents need access to large-scale historical data to perform complex tasks. Mosley pointed out that with the surge in AI-generated video content, the market’s demand for exabyte-level storage will continue to grow. Every stage of generative AI—from data collection, model training, to inference output—generates data that needs long-term retention. HDD, as the main medium for cold and warm storage, sees demand directly linked to the expansion scale of AI infrastructure.

Key indicators: Seagate’s GAAP gross margin in FY2026 Q2 reached 41.6%, and non-GAAP gross margin 42.2%, both hitting record highs. The margin improvement indicates increased pricing power, which stems from customer willingness to pay premiums for high-capacity HDDs under supply-demand imbalance. The report also shows Seagate achieved free cash flow of $607 million in Q2, with a healthy balance sheet.

Comparing HDD with another key storage medium, NAND, can clarify the structural demand. Third-party research firm Dolphin Analysis pointed out that between 2026 and 2027, demand for DRAM, NAND, and HDD will all grow faster than supply, maintaining storage tightness. The difference is NAND can expand capacity through stacking layers without building new lines, but Seagate and Western Digital have been more restrained in HDD capacity expansion, forming a tacit industry “control volume and price” strategy aimed at maintaining supply-demand balance for maximum profitability.

Historically, Seagate experienced two years of profit decline during the previous storage downcycle (2022-2023), leaving a deep industry memory. Under the current duopoly, both companies are unwilling to repeat the disorderly expansion that led to cyclical oversupply. Supply discipline combined with AI-driven structural demand forms the core logic supporting Seagate’s sustained profitability.

Analysts have reached a consensus on this. Citi raised its target price for STX from $740 to $1,150, maintaining a buy rating; Barclays increased it to $1,000; BofA revised upward from $320 to $400. The 12-month average target price of 22 analysts is $847.68, with the highest estimate at $1,150. These upward revisions are driven by AI-driven storage demand, industry supply discipline, and the pricing elasticity brought by high-capacity HDDs.

HAMR Mozaic Technology Roadmap: From Lab to Exabyte-Scale Deployment

Seagate’s Mozaic platform represents the commercialization of HAMR (Heat-Assisted Magnetic Recording) technology and is a key differentiator in its technical roadmap compared to Western Digital.

HAMR’s core logic is to break through the physical limits of traditional perpendicular magnetic recording (PMR). Conventional HDDs’ magnetic domain density is limited by superparamagnetic limits—when magnetic grains become too small, thermal agitation can destabilize data. PMR technology is approaching this physical boundary. HAMR introduces nanoscale laser spots that generate brief high temperatures (less than 1 nanosecond), allowing the magnetic head to write data within this window. Once the laser moves away, the medium cools instantly, stabilizing the data. Essentially, this technology replaces physical space expansion with photon engineering capabilities to achieve a leap in areal density, using energy input instead of physical growth.

The evolution pace of the Mozaic platform clearly reflects Seagate’s mass production capability for HAMR. In 2024, Seagate launched the first-generation Mozaic 3+ platform, with single-disk capacity exceeding 3TB, and near-line HAMR products shipped over 1.5 million units in a quarter. By March 2026, the second-generation Mozaic 4+ platform was officially certified by two large cloud service providers and entered mass production, with single-disk capacity increased to over 4TB, and a 10-disk enclosure reaching 44TB of total storage. An important process feature of Mozaic is the integration of photonic technology: Seagate designs and manufactures its own lasers, embedding them directly into wafer processes, shifting from externally purchased laser modules to in-house wafer-level integration, reducing errors in assembly. This decision shortens customer certification cycles—Mosley mentioned in the earnings call that Mozaic 3 has received certification from all major US cloud providers, and Mozaic 4+ testing at top cloud clients has been faster than expected.

Long-term, Seagate’s goal is to reach 10TB per disk, pushing overall capacity toward 100TB.

Similar to the process evolution path of SSDs, HAMR technology’s capacity growth is driven by areal density increases, making capacity expansion independent of physical size and energy consumption linear growth. For large-scale data centers, this means continuous expansion of effective storage without increasing data center footprint or power load—an advantage amplified in deployments measured in EB.

In the competitive landscape, Western Digital also benefits from AI data center HDD demand but follows a different technical route. In FY2026 Q2, WD’s revenue was $3.02 billion, surpassing Seagate, mainly driven by high-capacity near-line HDDs. Both companies form a duopoly in HDD, with similar gross margins (Seagate 42.2%, WD 46.1%), but Seagate currently leads in HAMR commercialization and single-disk capacity.

Why Do AI-Scale Data Centers Still Rely on HDD?

A common misconception in AI data center storage discussions is to see SSD and HDD as substitutes. In actual operation, their relationship is layered cooperation—SSDs handle hot data and cache layers, while HDDs serve cold and warm data at large scale. This structure balances cost and performance, with the total cost of ownership (TCO) at EB scale being significantly affected.

The unit storage cost gap between SSD and HDD is widening. Vdura’s data shows that from Q2 2025 to Q1 2026, the price of 30TB enterprise TLC SSD surged 257%, approaching $11,000, while HDD prices increased only about 35%. The cost per TB ratio expanded from 6.2x to 16.4x—meaning enterprise SSD per TB cost is now nearly 16 times that of HDD. Vdura further compared three-year TCOs: hybrid storage about $5.99 million, pure SSD about $25.2 million, with the hybrid being only a quarter of the pure SSD.

This cost gap has a significant impact at EB deployment scale. Over 80% of current data center storage capacity still relies on HDD. Global data volume is projected to grow from 72 ZB in 2020 to 527 ZB in 2029, with up to 87% of cloud storage still dependent on HDD. For cloud providers planning tens of EB of capacity, the cost difference between HDD and SSD determines storage architecture choices.

The capacity leap brought by Mozaic 4+ further amplifies HDD’s advantage in TCO. At 1 EB deployment scale, a 44TB disk solution reduces infrastructure investment by about 47% compared to a 30TB solution: HDD units drop from roughly 33,300 to 22,700, saving about 800k kWh annually, and reducing data center space by approximately 100 square feet. Mozaic 4+ also increases rack and wattage capacity, enabling storage expansion without increasing infrastructure footprint or power consumption.

Summarizing the cost structure: Mozaic 4+’s technical approach makes the marginal cost of storage capacity no longer linearly dependent on physical resources, but driven by areal density improvements. For cloud providers expanding AI data centers globally, this allows reconfiguring infrastructure investments—saving electricity and space resources for compute, without proportionally increasing storage costs.

Trading Seagate Stock on Gate: Diversify Your Investment Portfolio

For investors who recognize the “revaluation of AI infrastructure” logic behind Seagate, Gate’s trading platform offers a noteworthy option.

As a leading global digital asset and blockchain service platform, Gate has expanded its trading offerings from cryptocurrencies to US stocks. On Gate, users can conveniently trade high-quality tech stocks like Seagate (STX), achieving unified management of digital assets and traditional securities within a single account.

The core philosophy of Gate’s stock trading is “lower barriers, higher efficiency.” For users accustomed to crypto trading, Gate’s interface and parameter settings are designed in a unified style, making it user-friendly. Regarding fees, Gate supports zero-commission US stock trading, further reducing trading costs.

The process of trading Seagate stock on Gate is straightforward: first, log in and complete KYC; second, search for “STX” or “Seagate” in the “Stock Trading” page; third, choose the order type (market/limit), input the quantity, and confirm to execute. The experience closely resembles spot crypto trading, allowing quick adaptation.

Conclusion

The value chain of AI infrastructure is undergoing a shift from “chip-centric” to “full-stack reconstruction.” While GPU computing power is repeatedly discussed, the value of data storage is systematically underestimated. Seagate’s financial performance and technological roadmap provide verifiable evidence: continuous revenue growth, record-high gross margins, Mozaic 4+ entering mass production, capacity sold out early—these signals indicate HDD is experiencing a “double hit” driven by its technological lifecycle and AI demand—both stock replacement and incremental expansion are releasing simultaneously. Industry supply-demand balance is strengthening pricing power, and HAMR technology is enabling capacity leaps that reshape storage economics.

It’s important to note that all investments carry risks. Potential risks for Seagate include: large-scale HAMR capacity expansion facing yield bottlenecks affecting 44TB+ product delivery; Western Digital’s competitive pressure in high-capacity HDDs eroding margins; and historical storage industry cyclical reversals worth monitoring. The data cited in this article comes from public financial reports and third-party industry research and does not constitute investment advice. Investors should conduct independent assessments aligned with their risk tolerance before making decisions.