The Forgotten AI Trade That Just Delivered 970% Returns
MarketDash
While everyone chased NVIDIA and GPU stocks, Western Digital quietly became the second-best S&P 500 performer of 2025 with a 970% gain. The real AI story isn't about computing power—it's about where all that data actually lives.
Get Sandisk Alerts
Weekly insights + SMS alerts
Everyone knows the GPU trade by now. NVIDIA makes the processors that power AI. Micron supplies the fast memory those processors need. These are the names dominating portfolios, analyst coverage, and financial media headlines.
But what about the warehouse?
Western Digital (WDC) hit a 52-week high of $307 this morning. Less than twelve months ago, the stock was trading at $28.83. That's a roughly 970% gain, making it the second-best performer in the S&P 500 for 2025, trailing only Sandisk (NASDAQ:SNDK)—which happens to be its own spinoff. Morgan Stanley has slapped a $369 price target on WDC, and BofA just raised theirs to $375.
So what exactly happened here? And the more interesting question: is this story finished, or are we still in the early chapters?
The Storage Problem Nobody Talks About
Think about AI infrastructure like a library. GPUs are the reading room—fast, expensive, buzzing with activity. DRAM and high-bandwidth memory are the shelves right next to the reader's chair. But the vast majority of books in any library sit deep in the back stacks, rarely touched but absolutely necessary. That's where hard drives live. And right now, those stacks are overflowing.
AI training runs produce enormous volumes of data that can't simply be discarded. Model checkpoints, training logs, raw datasets, fine-tuning sets—all of it needs to sit somewhere affordable between sessions. WD's CEO told investors that HDDs make up roughly 80% of storage deployed in hyperscale environments. SSDs are too expensive for this kind of bulk storage. Tape is too slow. Hard drives occupy the sweet spot—cheap enough to deploy by the petabyte, fast enough to retrieve data in seconds.
The numbers support this thesis. Western Digital reported fiscal Q2 2026 revenue of $2.655 billion, shipping 215 exabytes of storage in a single quarter. Gross margins reached 46.1%, up sharply from a year earlier. The company expects full-year fiscal 2026 revenue to roughly double, crossing $12 billion. And it has locked in procurement agreements with its top seven customers for all of calendar 2026, with some deals extending into 2027 and 2028.
When your customers are AWS, Microsoft Azure, and Google Cloud, and they're signing multi-year contracts to secure supply, that's not a typical cyclical hardware story. That starts to resemble infrastructure. One caveat worth noting: these agreements are capacity reservations, not take-or-pay contracts. Customers can renegotiate terms if demand slows. History shows they have before.
The Technology Shift Most Investors Are Missing
Western Digital isn't simply selling the same hard drives it sold five years ago. The company is betting heavily on a technology called HAMR—Heat-Assisted Magnetic Recording—that could fundamentally alter the cost structure of AI storage for the rest of the decade.
Right now, WD's flagship product is a 40TB UltraSMR drive currently in qualification with two hyperscale customers, with volume production planned for late 2026. HAMR drives come next, with ramp production scheduled for 2027. By 2029, Western Digital is targeting 100TB per drive.
To understand why this matters, consider what it means to pack 100TB onto a single 3.5-inch disk. WD's engineers are developing laser technology that could increase areal density from 4TB per platter to 10TB per platter by 2028, fitting up to 14 platters into the same form factor. For a hyperscaler building a new data center, that's a massive shift in rack space requirements per petabyte of storage, and a significant reduction in power and cooling expenses.
Western Digital also announced new 'High Bandwidth Drive' technology that doubles sequential throughput, and a 'Dual Pivot' design that will launch in 2028 and eventually deliver up to 8x bandwidth gains. A separate power-optimized drive uses 20% less electricity. For data centers trying to manage energy costs at AI scale, this is a genuine selling point.
The clever part is that WD built HAMR on the same mechanical platform as its existing drives. Hyperscalers can deploy HAMR and older ePMR drives in the same rack without changing software or storage fleet management. That's not trivial—it eliminates one of the biggest reasons enterprise customers hesitate to adopt new technology.
The Split That Created Two Distinct AI Bets
For years, activist investor Elliott Management pushed Western Digital to separate its HDD and flash businesses, arguing the two were dragging down each other's valuations. The spinoff finally happened in February 2025. WD kept the hard drives. Sandisk (SNDK) took the flash business and relisted as an independent company.
Elliott was right. Sandisk became the best-performing S&P 500 stock in 2025, up 559%. WD came in second at 282%. The combined entity before the split would never have commanded the valuations both parts enjoy today.
But here's what most coverage misses: these two stocks now follow entirely different cycles. WD is tied to AI infrastructure capex—when hyperscalers build data centers, WD sells drives. Sandisk is tied to the broader NAND flash cycle, which includes consumer devices, smartphones, and PCs. When NAND supply tightens (as it is now), Sandisk benefits. When it loosens, Sandisk gets hit.
For investors, that creates an interesting pair trade opportunity. You can own WD as a long-cycle AI infrastructure play and short Sandisk if you think the NAND upcycle is nearing its peak. Or do the reverse if you think consumer tech demand is about to accelerate. Before the split, you couldn't make that distinction.
Western Digital also still holds about 5% of Sandisk shares after selling down its initial 19.9% stake. That's an option on future flash upside that doesn't appear cleanly in most analyst models.
What the Bulls Are Betting On
WD's long-term targets, outlined at its Innovation Day, are ambitious. Management is guiding for revenue growing at more than 20% per year, gross margins above 50%, operating margins above 40%, and earnings per share above $20. The company also approved a fresh $4 billion share buyback on top of $615 million already completed, bringing total buyback authorization to $6 billion.
Storage exabyte demand is expected to grow at more than 25% per year over the next five years. AI and cloud now account for roughly 90% of WD's revenue, up from a company where consumer drives once represented a much larger slice. Institutional ownership has climbed to 92%, as funds that avoided the old conglomerate structure piled in after the spinoff.
This means Western Digital is no longer a cyclical hardware company. It's a near-monopoly supplier of the cheapest and most scalable form of AI data storage, with multi-year customer contracts, a clear technology roadmap, and margins that look nothing like the company's historical profile.
What Could Go Wrong
History has a name for "sold out" storage suppliers: cyclical. The hard drive and flash memory markets have gone through boom-bust cycles repeatedly. The last major bust, in 2023, sent WD's stock below $30 and forced deep losses. The current setup—tight supply, rising prices, multi-year deals—looks different. But similar things were said in 2017 and 2021.
The bear case rests on three concerns. First, if hyperscalers over-order now to lock in supply, they could end up with surplus drives in 2027 and cancel or delay new orders. That digestion phase has ended previous cycles badly. Second, Seagate (STX) isn't just "pursuing the same HAMR roadmap"—it's already executing. Seagate shipped over one million HAMR drives commercially in 2025 and has qualified its drives at Google. WD's HAMR ramp doesn't start until 2027. Western Digital is betting its future on drives that aren't in production yet, against a rival that already has hyperscale customers buying. Third, the SSD threat isn't purely about price. Meta's engineering team recently argued that newer flash drives use roughly 80% less power per terabyte than hard drives—and in markets like Northern Virginia and Dublin, where AI data centers are already running out of power, that gap matters more than cost per terabyte alone. HDDs are still five times cheaper per TB, which keeps them dominant for cold storage today. But the pressure is coming from two directions simultaneously: a rival ahead on technology execution, and a power problem that gets harder to ignore every year.
There's also one detail worth mentioning. WD's CEO Irving Tan sold 17,201 shares on February 2nd, cashing in roughly $5.1 million through a pre-set trading plan. Single insider sales under 10b5-1 plans aren't red flags on their own—executives sell for many reasons. But at a stock trading near all-time highs, it's worth keeping in mind.
One more thing retail investors rarely hear: this story has a historical twin. In 2013 and 2014, Seagate and Western Digital were the "cloud storage winners", with nearly the same narrative, almost word for word. Both stocks rallied hard. Then PC demand fell, SSDs crept into laptops, and exabyte growth slowed just enough to tip the cycle. By 2015, both stocks had given back a large chunk of those gains. The investors who saw it coming weren't smarter. They were just watching the right numbers: exabyte shipment growth rates, channel inventory levels, and whether management's pricing language was shifting from "strong" to "competitive." Those same three signals will likely call the turn in this cycle too, when it comes.
With many analysts covering the stock, the average price target is $265—about 10% below today's price of $294. Most analysts are bullish (Buy/Strong Buy consensus), but the consensus target suggests the market has already priced in substantial good news. Even the bulls see limited room left after the 970% rally.
How Western Digital Stacks Up Against the Competition
Here's how WD compares to Sandisk (SNDK) and Seagate (STX) on the key metrics:
Metric
WD (WDC)
Sandisk (SNDK)
Seagate (STX)
Focus
HDD (pure-play)
NAND flash (pure-play)
HDD (pure-play)
1-yr return
+504%
+650% (since Feb '25 spinoff)
~+90%
Gross margin (latest)
46.1%
41–43% (guided)
~32% (est.)
Revenue growth (YoY)
+28%
+23% (Q1 FY26)
~+20% (est.)
Key tech bet
HAMR → 100TB by 2029
BiCS8 NAND upcycle
HAMR, competing roadmap
Avg. analyst target
$265 (below market)
$280 (~4% upside)
N/A (check current)
Main risk
AI capex slowdown, Seagate competition
NAND oversupply (2027?)
HAMR execution, no flash optionality
The Real Test Ahead
GPU stocks get the headlines. Hard drives get the data. Western Digital has accomplished something unusual: it took a commodity hardware business that everyone wrote off, spun out its most volatile division, locked in the biggest cloud companies on long-term supply deals, and built a technology roadmap that keeps HDDs relevant for at least the next five years of AI growth.
The easy money from $29 is gone. At $295, with the average analyst target below the current price, the stock already prices in a lot. What it doesn't fully price in is the 100TB roadmap, the platform software business WD is quietly building for mid-scale AI customers, and the possibility that AI data creation keeps growing faster than anyone's models suggest.
The AI trade isn't just chips. It's everything that holds the data those chips generate. Western Digital built the warehouse. The question now is whether the rent keeps going up—or whether the next cycle brings it back down.