AI Trades Shift from "Chips" to "Power and Data Centers," Infrastructure Plays like BE and CRWV Gain
The investment narrative around artificial intelligence is undergoing a measurable rotation. While semiconductor and GPU manufacturers captured the bulk of early-cycle returns, the current bottleneck has moved downstream — to electricity generation and data center capacity.
Several data points support this thesis:
Grid constraints are binding. Interconnection queues for new power capacity in major U.S. markets now extend 7 to 10 years. Major technology companies, including $Alphabet(GOOGL)$, have publicly identified grid connection as the primary constraint on data center expansion. Training clusters for frontier AI models routinely require 100+ megawatts of continuous power, a scale traditional utility infrastructure struggles to provision on relevant timelines.
Alternative power solutions are scaling rapidly. $Bloom Energy Corp(BE)$ has positioned its solid oxide fuel cells as a near-term workaround — deployable on-site in approximately 90 days, bypassing grid interconnection entirely while delivering 10–30% higher efficiency than conventional gas turbines. Brookfield committed $5 billion to deploy these systems across AI manufacturing facilities. GS Research estimates 8–20 gigawatts of fuel cell capacity will supply data center electricity by 2030.
$CoreWeave, Inc.(CRWV)$ represents the compute layer of this infrastructure buildout. The GPU-cloud provider, which went public in March 2025, supplies specialized infrastructure for OpenAI, Microsoft, and other AI labs. The company remains unprofitable with negative trailing EPS and trades well below its $187 post-IPO high. However, its recent $8.5 billion investment-grade delayed-draw term loan facility — structured by Blackstone Credit and coordinated by GS and JPMorgan — signals institutional debt market confidence in the underlying business model.
GS Research frames the opportunity as a "Reliability supercycle" — a multi-year investment cycle driven by the need to build redundancy across power, water, labor, networks, and supply chains. The firm identifies six structural drivers: Pervasiveness, Productivity, Price, Policy, Parts, and People. Data-center-exposed power generation equipment companies have outperformed all other AI supply chain cohorts since 2025.
The implication is straightforward: semiconductor demand remains robust, but the marginal investment opportunity has shifted toward the physical infrastructure required to sustain AI workloads at scale. The constraint is no longer the ability to manufacture chips — it is the ability to power them.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
