News

Hard Drives Are Sold Out for 2026: AI Data Centers Have Bought Everything

Western Digital and Seagate confirm their entire 2026 HDD production is sold out to AI hyperscalers, with consumer prices surging nearly 50% as the storage industry pivots away from retail buyers.

Hard Drives Are Sold Out for 2026: AI Data Centers Have Bought Everything

It's February, and you already can't buy a hard drive at a reasonable price for the rest of the year. Both Western Digital and Seagate - which together control roughly 80% of the global HDD market - have confirmed that their entire 2026 production capacity is sold out, locked in by a handful of hyperscale AI customers through binding purchase orders and multi-year agreements that stretch into 2028.

This isn't a temporary supply hiccup. It's the storage industry telling consumers: you are no longer our primary customer.

TL;DR

  • Western Digital and Seagate have sold out 100% of their 2026 HDD production capacity to AI data centers
  • Consumer hard drive prices have surged ~50% since September 2025
  • WD now derives 89% of revenue from hyperscaler cloud customers, just 5% from consumer
  • Multi-year contracts already lock capacity through 2027-2028 with no relief in sight

The Numbers

Both manufacturers dropped the news during their Q2 FY2026 earnings calls. The figures paint a picture of an industry that has fundamentally reoriented itself around a small number of very large customers.

MetricWestern DigitalSeagate
Q2 FY2026 Revenue$3.02B (+25% YoY)$2.83B (+21.5% YoY)
2026 HDD Capacity100% sold out100% allocated
Long-term AgreementsThrough 2028Through 2027+
Stock YTD Performance+63.4%~+60% (surged 19% post-earnings)

Western Digital: "Firm POs With Our Top Seven"

WD CEO Irving Tan was blunt during the earnings call: "We're pretty much sold out for calendar '26. We have firm purchase orders with our top seven customers." Those long-term agreements (LTAs) combine volume commitments measured in exabytes with locked-in pricing terms - two customers are signed through 2027, one through 2028.

The company shipped 215 exabytes in Q2 alone, up 22% year-over-year, including 103 EB from its latest-generation ePMR drives with capacities up to 32 TB. It also announced 44 TB drives shipping this year, with a roadmap to 100 TB by 2029.

WD beat earnings expectations with $2.13 EPS (nearly 10% above consensus) on $3.02 billion in revenue.

Seagate: "Fully Allocated"

Seagate CEO Dave Mosley echoed the same reality: "Our nearline capacity is fully allocated through calendar year 2026, and we expect to begin accepting orders for the first half of calendar year 2027 in the coming months."

Seagate posted $3.11 EPS against analyst estimates of $2.81, with revenue of $2.83 billion. Customer discussions already extend into 2028.

Toshiba: The Missing Variable

Toshiba, the world's third major HDD manufacturer, went private in 2023 and no longer publishes detailed capacity forecasts. Industry analysts expect its allocation to be similarly constrained. If that's the case, there's essentially zero unbooked nearline HDD capacity anywhere on the planet for the rest of the year.

Why HDDs? It's Still About Cost Per Byte

You might wonder why AI data centers are buying hard drives at all when SSDs are faster in every measurable way. The answer is straightforward economics: HDDs remain the cheapest way to store data at exabyte scale.

AI workloads produce staggering amounts of data. Training datasets, inference logs, processed backups, scraped web content, multimodal training data (video, audio, images) - all of it needs to live somewhere. Hot data sits on NVMe SSDs close to the GPUs. Everything else goes to spinning rust.

"It is data that is needed to fuel the entire AI process, from training to inference, to enable stronger models and sharper inference results," Tan said during WD's earnings call.

And the SSD alternative isn't getting cheaper either. Enterprise SSD prices have surged over 250% due to NAND flash scarcity, pushing even more cold-storage workloads back to HDDs.

Who's Buying

The buyers are the usual suspects: Google, Microsoft, Amazon Web Services, Meta, and OpenAI, among others. Analyst Sid Nag from Tekonyx put it starkly: "No meaningful open production remains for discretionary buyers except the hyperscalers. HDD manufacturing capacity is now almost exclusively prioritized for large AI/cloud players."

The top ten cloud providers are expected to account for 70% of server capital expenditure in 2026. As we noted in our analysis of the AI memory chip crisis, this pattern of a handful of companies absorbing entire supply chains is becoming the defining dynamic of AI infrastructure spending.

WD's revenue split tells the story: 89% of revenue now comes from hyperscaler AI cloud customers. Consumer represents just 5%. That's not a temporary blip - it's a permanent business model shift.

Where Consumers Stand

This is where the story gets painful for anyone building a NAS, running a home server, or managing storage for a small business.

The Price Reality

Consumer hard drive prices have surged approximately 46% since September 2025. Specific examples:

DriveSept 2025 PriceFeb 2026 PriceChange
Seagate IronWolf 24TB~$300~$500+67%
Consumer/NAS drives (avg)--+30% to +50%
Enterprise/Nearline (avg)--+20% to +40%

Analysts from Bank of America and Morgan Stanley predict pricing won't normalize until early 2028.

It's Not Just Hard Drives

The storage crisis is compounding with broader component shortages. DRAM and NAND flash are similarly constrained. Dell and Lenovo have already raised PC prices 15-20% as these shortages ripple through the supply chain. As Omdia analyst Vlad Galabov warned: "It will be a very tough year for the standard enterprise general purpose server and for enterprise storage."

If you're planning a storage upgrade, the calculus is simple: buy now or wait until 2028. Waiting for seasonal sales in 2026 is unlikely to yield meaningful discounts.

Where It Falls Short

The current situation exposes a structural vulnerability in the storage industry. Three companies control nearly all HDD production globally, and when a single use case (AI infrastructure) absorbs all available capacity, every other buyer gets squeezed.

There's no meaningful plan to expand manufacturing capacity in the near term. Building new HDD fabrication facilities takes years, and manufacturers have little incentive when their current output is already sold at premium prices on multi-year contracts.

For the open-source AI community and smaller companies building AI agents, this adds yet another cost barrier. Running models locally already requires expensive GPUs. Now the storage to hold training data, model weights, and inference logs is getting expensive too.

The irony is sharp: the AI boom that's supposed to democratize technology is, at the infrastructure level, consolidating resources into fewer and fewer hands. If you're not one of the top seven customers, you're competing for scraps.


The storage super cycle is here, and it's not built for you. It's built for the companies training the next generation of AI models - and they've already signed the contracts to prove it.

Sources:

Hard Drives Are Sold Out for 2026: AI Data Centers Have Bought Everything
About the author AI Infrastructure & Open Source Reporter

Sophie is a journalist and former systems engineer who covers AI infrastructure, open-source models, and the developer tooling ecosystem.