Hyperscaler Decade
Best of
@favoritetakes
+2.3%
since Sep 28
The idea
I believe cloud platforms are on their way to tens of trillions in combined value over the next 5-10 years, and that the spending it takes to get there flows through a whole supply chain even faster. So this basket anchors on the hyperscalers themselves, then spreads across the chips, data center gear, power providers, and landlords that capex is forced through, weighted toward large established names at every layer. I keep a small cash reserve to add on weak stretches rather than chase, since this is meant to be held for years, not traded around headlines.
What it owns
- The core platforms whose infrastructure spending drives the entire supply chain for the next decade. · MicrosoftIt operates Azure and enterprise AI services as core cloud platforms expected to grow with the hyperscaler buildout., AlphabetIt operates Google Cloud while designing its own TPU chips, gaining advantage from both cloud revenue growth and lower internal computing costs than competitors., AmazonIt operates AWS, the largest cloud platform, and funds massive data center construction to support that service., Meta PlatformsIt is spending heavily on AI infrastructure to improve advertising and products, directly embodying the capital flows that sustain the hyperscaler buildout., OracleIt operates Oracle Cloud with large AI contract backlog to convert, positioning it as a smaller platform with room to gain share if hyperscaler growth accelerates.36%
- Processors and memory form the largest single line item in hyperscaler capital spending and determine which companies capture the most value. · NVIDIAIt manufactures the default processors for hyperscaler AI training and inference, capturing spending across the entire cloud ecosystem., Taiwan SemiconductorIt manufactures nearly every advanced AI chip regardless of designer, making it a central tollbooth in the hyperscaler supply chain., BroadcomIt provides custom networking and compute chips that hyperscalers use to reduce dependence on off-the-shelf processors., Advanced Micro DevicesIt supplies AI processors as a second source to hyperscalers who want alternatives to single-supplier pricing., Micron TechnologyIt produces high-bandwidth memory that AI systems consume in large quantities, with its growth tied to data center memory demand outrunning supply.27%
- Switches, cooling systems, and electrical equipment that convert raw processors into functioning data centers at scale. · Arista NetworksIt manufactures the networking fabric that connects thousands of AI chips into single machines for hyperscaler customers., Vertiv HoldingsIt supplies cooling and power distribution equipment for data centers, with demand rising as AI racks require more thermal management than previous generations., EatonIt supplies electrical distribution equipment that moves power from the grid into data centers as part of a larger diversified industrial business.11%
- Data centers require firm, reliable electricity at volumes the existing grid was not designed to supply, making power the binding constraint on growth. · Constellation EnergyIt operates the largest nuclear fleet in the country and supplies carbon-free baseload power to hyperscalers under long-term contracts., GE VernovaIt manufactures gas turbines and grid equipment needed to deliver power to data centers, with order backlog sold out years in advance., VistraIt generates and sells power at market prices, with margins expanding if data center electricity demand in its regions grows faster than new supply.11%
- Companies that own data center buildings hyperscalers lease when construction cannot keep pace with demand. · EquinixIt owns data center buildings where companies interconnect to cloud platforms, benefiting from rising occupancy and rents as AI demand fills capacity., Digital RealtyIt leases entire data center campuses directly to hyperscalers, with revenue growth tied to the volume and rental rates of that leasing.9%
- Cash6%