From AI chips to power infrastructure
Leopold Aschenbrenner, a 24-year-old investor, has redirected his $5.5 billion AI hedge fund away from AI chip and infrastructure stocks and toward energy infrastructure. The move is based on a clear thesis: as AI data centers continue to scale, electricity supply may become a more critical constraint than computing hardware itself.
Aschenbrenner has sold major AI infrastructure names including NVIDIA and Broadcom, and is now allocating capital to companies positioned to solve the power problem. His largest holding is Bloom Energy, which makes up 20% of the portfolio. Bloom Energy focuses on solid oxide fuel cells that convert natural gas into electricity, offering a modular and efficient option that could fit the fast-growing needs of AI data centers.
Why Bitcoin miners fit the strategy
A notable part of the shift is his investment in Bitcoin mining companies. The thesis is not centered on cryptocurrency exposure, but on the value of the land and electricity access these firms already control. By targeting miners, the fund can potentially tap into existing power capacity and sites suitable for AI infrastructure without going through long licensing and permitting timelines.
This points to an emerging overlap between AI infrastructure and crypto infrastructure. Bitcoin miners have traditionally been evaluated through the lens of energy costs and digital asset cycles. But in an environment where AI labs and data centers are competing for power, those same mining assets may be reassessed as strategically useful infrastructure with broader applications.
A bearish bet on traditional IT outsourcing
Aschenbrenner is also short Infosys, reflecting a negative view on traditional IT outsourcing models as AI automation advances. That position adds another layer to the fund’s strategy, suggesting it is not simply chasing AI enthusiasm, but actively distinguishing between sectors likely to benefit from AI growth and those that may come under pressure.
Overall, the portfolio shift highlights a broader market idea: the next major bottleneck in AI may lie less in chips and more in energy infrastructure. In that context, companies with access to power, land, and deployable energy systems—including some Bitcoin miners—could gain renewed strategic relevance.