The era of hoarding GPUs is over; the new AI infrastructure race focuses on fixing the 95% of wasted capacity, putting storage providers at the center of the next boom.
Back
Western Digital gains 7% as AI focus shifts from GPUs to efficiency
Western Digital (WDC) shares surged 7 percent Monday after reporting strong earnings, but the rally points to a deeper shift in the AI infrastructure market: the focus is moving from simply buying GPUs to using them efficiently.
The move is backed by strong fundamentals, with 14 of 16 analysts tracked by TipRanks holding a buy rating on the stock, reflecting confidence in the company’s growth trajectory after a strong earnings report and a 20 percent dividend hike.
The company reported fiscal third-quarter adjusted earnings of $2.72 per share on revenue of $3.34 billion, a 45 percent year-over-year increase. The results beat Wall Street estimates of $2.39 per share and revenue of $3.25 billion. The rally was part of a broader updraft for the sector, with Micron Technology gaining 6 percent and SanDisk shares also rising.
But the real story is the market's pivot from acquisition to optimization. For the last two years, the "GPU scramble" dominated IT budgets. Now, with real-world enterprise GPU utilization averaging a stunningly low 5 percent, according to research from VentureBeat, the bill is due. The focus is no longer on securing capacity, but on squeezing productivity out of billions in underutilized, depreciating hardware.
From Hoarding to Productivity
The blank-check era for AI hardware is over. VentureBeat's Q1 2026 market tracker confirms that "access to GPUs" has fallen sharply as a priority for IT decision-makers, while "cost per inference/TCO (total cost of ownership)" has surged. In any other department, a 95 percent waste metric would be a fireable offense; in AI, it was called "preparedness." Now, it's a line-item emergency.
This forces a fundamental change in how AI leaders are measured. Success is no longer about securing the stack, but about squeezing the stack. This is why the market is rewarding companies that provide the levers for efficiency. The rally in Western Digital, Micron, and SanDisk is not just about a cyclical memory boom; it's a bet on the companies that can help fix the 95 percent problem.
Storage as the Efficiency Lever
Fixing the utilization wall requires a structural overhaul of the efficiency stack, and storage is a primary lever. An idle GPU waiting for data is a wasted resource. High-performance storage architecture is critical to keeping the expensive silicon fed.
This is where Western Digital's strategy of delivering high-capacity HDDs for data centers comes into play. As AI models grow, so does the data required for training and inference. Efficiently storing and accessing that data is a key bottleneck. Newer architectures are solving this by offloading tasks like KV cache management from expensive GPU memory to shared, high-performance storage, a move that improves concurrency and drives down the cost per token. SanDisk's BiCS8 quad-level cell storage, currently in qualification with two major hyperscalers, and Micron's in-demand HBM memory are other key pieces of this efficiency puzzle.
The market is recognizing that the next phase of AI advantage will be built on superior inference economics. Investors are betting that Western Digital, by providing a critical tool to unlock GPU productivity, is well-positioned to capture value in this new, efficiency-focused era.
This article is for informational purposes only and does not constitute investment advice.
