NVIDIA Corporation NVDA
AI accelerators
Direct · MAG 7AI
AI investment is a chain of model demand, compute purchases, power, data centers, and software adoption. The risk is that capacity and valuations grow faster than durable cash returns.
Research updatedWhy markets care
DOE’s data-center analysis projected a large increase in U.S. electricity use by 2028. That makes grid connection, power equipment, and energy cost part of the AI investment case, alongside chips and cloud capacity.
OpenAI and Anthropic have announced large funding rounds and infrastructure partnerships. Those commitments support suppliers, but announced capacity is not the same as utilized, profitable capacity.
What to watch
- Hyperscaler capital spending versus realized AI revenue
- GPU and memory supply, utilization, and pricing
- Grid interconnection, power costs, and data-center delivery dates
What could change the view
Inference efficiency can improve quickly; fewer chips per task could reduce unit costs while increasing total usage, so capacity forecasts are uncertain in both directions.
Names in our universe
Tracked watchlist snapshot: April 14, 2026. Exposure paths are research associations, not holdings or return forecasts.
Microsoft Corp. MSFT
Cloud and AI distribution
Direct · MAG 7Amazon.com, Inc. AMZN
Cloud and AI compute
Direct · MAG 7Alphabet Inc (Google) Class C GOOG
Cloud and AI compute
Direct · MAG 7Vertiv Holdings, LLC VRT
Data-center power and cooling
Direct · CLOUD/SEMIEaton Corporation, PLC ETN
Electrical infrastructure
Second-order · HARDWARE / PERIPHERALSConstellation Energy Corporation CEG
Power supply
Second-order · NUKES