Remember the market’s surprise of 2025? It wasn’t another meteoric rally for Nvidia that stole the headlines; it was storage makers. Western Digital’s stock became one of the year’s breakout winners as cloud providers and hyperscalers poured money into data centers to feed generative-AI workloads. That shift — from big-iron glamour to the quieter suppliers behind the scenes — is shaping investor thinking as 2026 begins.
The quiet frenzy behind the servers
Big models need three things: compute, power and somewhere to keep the mountains of data they chew through. In 2025, that translated into an unexpected leadership group in the S&P 500. Sandisk and Western Digital shot up, Seagate climbed, and companies you’d normally call industrial — from power providers to cable and fiber makers — joined the party.
Investors who had assumed the AI trade meant only GPUs found themselves chasing what traders call “pick-and-shovel” plays: the parts of the supply chain that grow when hyperscalers expand capacity. It’s a classic market pivot. Nvidia and the cloud giants remain central, but percentage gains have cooled; the more interesting upside now looks like it could come from the infrastructure that enables their growth.
Where portfolio managers are placing bets
Not everyone is searching the same corners. Some managers are doubling down on retail companies that will benefit from AI-enhanced shopping experiences, while others load up on the vendors that physically build and power data centers.
Eric Clark, CIO at Accuvest Global Advisors, sees a year where retail and AI overlap. His top picks include Amazon and Mercado Libre for their platform effects and reach, and he keeps an eye on discount and value retailers such as Walmart and TJX for durable consumer demand. Clark also likes ServiceNow and Intuit as software plays that help enterprises actually deploy AI — not just buy the hype.
That split — software and platforms on one side, hardware and infrastructure on the other — is sensible. A retailer’s recommendation engine matters, but that engine needs servers, storage and steady electricity to hum.
Why the “picks-and-shovels” story matters
There’s a practical reason to consider the suppliers: valuations. Hyperscalers and marquee chip names are large, expensive, and already priced to perfection in many cases. The companies that sell racks, storage arrays, fiber, power gear and cooling can still grow revenues quickly as hyperscalers scale out. As one portfolio manager put it, look for the places “that money is being spent.”
But remember the lessons of cyclical demand. After a pandemic bump, some suppliers saw sharp reversals once the rush cooled. The same could happen here if hyperscalers pause or slow capex, or if a technology shift reduces demand for a subset of hardware.
The retail-AI fusion: more than targeted ads
AI in retail isn’t only about product recommendations. Clark and others point to broader structural changes: smarter inventory forecasting, finance and payments embedded into marketplaces, and AI-driven in-store experiences that increase cart sizes. Some tech innovations are already changing how customers discover and buy. Google’s recent work to add agentic booking and better shopping tools hints at how search and commerce will tighten together in practice — and that matters for retailers and payments businesses alike. See developments in Google’s AI Mode booking features.
Meanwhile, the tools that generate images, product mockups, and creative assets are evolving rapidly. Microsoft’s jump into in-house image models shows how big tech is layering capabilities that retailers and creators can adopt — often behind the scenes in marketing and product pipelines. Developers and product teams will chase those capabilities, which could be another tailwind for software vendors that integrate them. Read about Microsoft’s new image model here: MAI-Image-1.
Risks: hype, capex cycles and crowded valuations
A sober column in late 2025 reminded investors there’s talk of a bubble: the same exuberance that pushes related stocks higher can reverse quickly if growth disappoints. Two things to watch:
- Capex cadence: Hyperscalers’ spending plans matter. If they pause, pick-and-shovel names can wobble fast.
- Execution and competition: Software hopefuls must actually help companies roll out AI in messy, real-world settings — which is why some managers favor firms that act as integrators.
- Diversify within the theme: mix hyperscalers and semiconductor leaders with storage, power, cabling and software integrators.
- Favor businesses with multi-year contracts or sticky revenue — those smooth capex swings.
- Look for retail platforms that marry commerce with finance and AI-driven personalization; those are likely to sustain higher margins.
There’s also regulatory and geopolitical risk around where data centers get built, who supplies semiconductors, and how trade restrictions evolve. And don’t ignore the inventory cycle — too many vendors scaling production at once could create a glut, just as happened after past demand spikes.
How investors might think about positioning for 2026
If 2025 taught the market anything, it’s that opportunities often hide in the supporting cast. A few practical approaches:
Finally, keep an eye on the weird and the new. Companies, and even ideas like putting datacenters off-planet, are on the table for longer-term thinkers — projects such as Google’s Project Suncatcher suggest how far the industry is willing to look for capacity and advantage.
If 2025’s winners taught investors anything, it’s that the market’s spotlight can move quickly. For 2026, many professionals aren’t betting on a single star — they’re building exposure to the whole ecosystem that lets AI actually run in the real world.