NeutralAVGOBroadcomBroadcom000660SK hynixSK hynixAAOIApplied Optoelectronics, Inc.Applied Optoelectronics, Inc.ALABAstera LabsAstera LabsAMDANETArista NetworksArista NetworksCOHRCoherent Corp.Coherent Corp.CRDOCredo Technology Group HoldingCredo Technology Group HoldingDRAMRoundhill Memory ETFRoundhill Memory ETFEMEEMCOR GroupEMCOR GroupETNEaton Corporation plcEaton Corporation plcFIXComfort Systems USAComfort Systems USALITELumentum Holdings Inc.Lumentum Holdings Inc.MODModine ManufacturingModine ManufacturingMRVLMarvellMarvellMUMicron TechnologyMicron TechnologyNVDANVIDIANVIDIAPWRQuanta ServicesQuanta ServicesSNDKSandisk CorporationSandisk CorporationVRTVertiv Holdings, LLCVertiv Holdings, LLCWDCWestern Digital CorporationWestern Digital Corporation

6d ago on

AI bears are in full swing. I keep seeing people argue that "AI isn't an investment, it's a trade." I strongly disagree. As long as demand persists and the economics make sense, the buildout will continue. If not in one state, then another. If not in the US, then Mexico. If not on Earth, then space. And that last part isn't even a joke anymore. Elon Musk is seriously pursuing orbital data centers. The point is, hyperscalers will keep looking for places to build as long as there's demand and money to be made. Geography, regulations and power availability might change where those facilities are built, but the demand for compute isn't tied to one particular location. Data centers don't just disappear once they're built. They'll need maintenance, upgrades, replacements, new GPUs, more DRAM, networking equipment, storage, cooling and power infrastructure. Then there's sovereign AI, enterprise adoption, new applications and technologies we probably haven't even thought about yet. Will the spending slow down? Absolutely. At some point it will, and some companies will get hurt badly. Some projects won't generate acceptable returns, and some businesses will disappear entirely. But I don't believe AI infrastructure spending is some temporary phenomenon that suddenly ends when this initial buildout is complete. The real challenge, in my view, is figuring out which companies will benefit from all that hyperscaler capex, how much of it they can actually capture, and whether they'll generate attractive returns for shareholders. Because a company can be operating in the right industry, experiencing enormous demand, and still turn out to be a terrible investment. That's where I think investors should spend more time looking. Here are 20 companies I'd watch across the AI infrastructure supply chain: 🖥️ Compute & AI accelerators $NVDA — NVIDIA $AMD — Advanced Micro Devices $AVGO — Broadcom 💾 Memory & Storage $MU — Micron $SKHY — SK Hynix $STX — Seagate $WDC — Western Digital $SNDK — Sandisk 🌐 Networking & Connectivity $ANET — Arista Networks $CRDO — Credo Technology $ALAB — Astera Labs $MRVL — Marvell Technology 🔦 Optical Components $LITE — Lumentum $COHR — Coherent $AAOI — Applied Optoelectronics ⚡ Power & Cooling $VRT — Vertiv $ETN — Eaton $MOD — Modine 🏗️ Construction & Electrical Infrastructure $PWR — Quanta Services $EME — EMCOR $FIX — Comfort Systems USA and many others. Can't possibly fit them all here.