Nvidia gets the headlines, but servers, memory, networking and real-estate plays are where institutional flows will broaden. Here are seven stocks investors often miss — and the risks to watch.
Illustration by IMF Alpha editorial · Reviewed by Pedro Marini
Why this matters now
Nvidia has become shorthand for AI investing. That concentration is a double-edged sword. As the largest cloud providers push billions into GPUs, everything behind the chips — memory, boards, networking, cooling, power and the real estate that houses it all — is suddenly under intense pressure and, yes, poised for outsized returns. The system that turns compute demand into racks and datacenters is where the real, less glamorous money shows up.
Quick take: if you want AI exposure without owning a single chipmaker, look down the value chain: servers, DRAM, networking silicon, fab tools and data-center landlords. Some of those plays follow demand more directly and carry very different valuation dynamics than Nvidia.
Seven stocks worth watching (and why)
SMCI — Super Micro Computer (servers): Supermicro builds GPU-dense systems. For investors wanting to express conviction about incremental GPU installs, SMCI has been the fastest lever. Think of it as the shipbuilder for a GPU navy. Growth can be uneven, but margins are widening as custom, high-density designs command a premium.
MU — Micron Technology (memory): GPUs need huge pools of high-bandwidth memory. Memory is cyclical, true, but the baseline for DRAM demand has shifted up as memory per GPU session explodes. Micron is the most direct U.S. DRAM play.
MRVL — Marvell Technology (networking and accelerators): High-speed networking and packet processing are the plumbing that lets GPU racks scale. Marvell’s switches and accelerators are turning up inside AI-optimized topologies more often than not.
AMAT — Applied Materials (chipmaking equipment): As foundries chase denser nodes and advanced packaging for AI accelerators, equipment suppliers see the orders sooner than many expect. Applied captures upgrades and capacity additions across multiple fabs.
LRCX — Lam Research (etch and deposition tools): Complementary to Applied, Lam is indispensable for advanced logic fabrication. If capex keeps rising, these two benefit disproportionately.
DLR — Digital Realty (data-center REIT): GPUs draw lots of power, and not every cloud owner wants to run every datacenter. Digital Realty offers a relatively stable earnings stream with direct exposure to where AI compute actually lives.
NVDA — Nvidia (platform leader): Not a hidden gem, just the anchor. Even if you worry about concentration, Nvidia’s platform and software moat keeps demand for the rest of the chain elevated. Ignore it at your peril.
The list is not exhaustive. But it sketches a diversified way to play the hardware side of the AI build-out.
The data, in brief
Risks and counterpoints
How to use these ideas
Signals I’d watch
The takeaway
Nvidia is the headline; the real trade for many investors is the supply chain that turns demand for models into physical racks. Less glamorous, often messier, and—once the hype subsides—frequently more rewarding. If you own a single AI stock, fine. Just be conscious of what you’re leaving out: the memory sticks, the switches, the machine rooms and the tools that make the chips run.
How banks, cloud vendors and chip makers are betting on fake-but-faithful data to train models while dodging privacy landmines—and why that bet has limits
Firms are moving large language models behind the firewall to protect alpha, cut cloud bills and limit hallucinations — but model risk and regulation are accelerating the debate
From instant voice banking to fraud checks that never leave the handset, on-device models are reshaping fintech economics, user privacy, and who wins in the chip race.