Energy Level
“Energy Level” – A cartoon that illustrates AI’s next bottleneck, energy, as investors rotate from chips to power and grid infrastructure.
Artificial intelligence’s first real constraint was compute. Training modern models requires massive parallel processing, and GPUs quickly became the only practical way to scale AI workloads. Nvidia captured that moment by pairing high-performance chips with a powerful software ecosystem. For investors, however, the GPU trade has become crowded. Nvidia’s rise and the valuation expansion across semiconductor suppliers mean much of the obvious upside in AI compute is already reflected in prices.
Increasingly, capital is rotating toward energy, which many see as the next structural constraint. AI data centers consume enormous amounts of electricity. A single large training cluster can require power comparable to a small city. The existing grid was built around human consumption patterns, not thousands of GPU clusters running continuously. As AI deployment accelerates, access to electricity, grid permits, and suitable industrial sites is quickly becoming the next bottleneck.
That shift is visible in the strategy of Leopold Aschenbrenner, founder of an AI-focused hedge fund. Known for his essay Situational Awareness, Aschenbrenner launched his fund with roughly $255 million in late 2024 and reportedly grew it to about $5.5 billion within a year after early bets on the GPU cycle. Recent filings show a pivot away from chip manufacturers toward infrastructure. Today, his largest position, roughly 20% of the portfolio, is in Bloom Energy, which builds modular fuel-cell systems capable of generating electricity directly at data centers. He has also built significant exposure to GPU cloud operator CoreWeave and to power-connected infrastructure such as Core Scientific, whose mining facilities already hold valuable energy permits and industrial-scale sites.
Every major technology cycle eventually runs into a constraint that reshapes where capital flows. In the early phase of the AI boom that constraint was chips. In the next phase, it may be electricity. As Satya Nadella, the CEO of Microsoft, observed, “every AI workload ultimately runs on infrastructure.”
Sources:
Sharon Goldman (Mar 05, 2026) – Why Leopold Aschenbrenner’s AI hedge fund is betting big on power companies and Bitcoin miners to fuel the ‘superintelligence’ race – Fortune
Fred Hubler (Mar 11, 2024) – AI Is Power Hungry: Invest Now In Energy And Infrastructure – Forbes
Prafull Sharma (Mar 10, 2026) – Digitalization And AI: The New Asset Infrastructure Transforming The Energy Sector – Forbes
