The article discusses the potential challenges in the AI infrastructure sector, particularly regarding data centers and AI compute pricing. Gabe Marshank of Stansberry Research highlights the buildout of data centers, expected to reach $4.5 trillion by 2030, requiring about $1.7 trillion in annual revenue to achieve adequate returns. Investors may be overlooking the potential for decreasing AI compute prices, which could mirror historical collapses in similar commodities.
Marshank argues that while the technology itself is promising, the business models driving investment in data centers may not be sustainable. He compares the current situation to the bandwidth bubble of the late ’90s, where oversupply led to a significant drop in prices and numerous company failures.
Additionally, he touches on trends in energy production, emphasizing “behind-the-meter” power generation as a faster alternative to traditional grid connections, particularly through companies like Atlas Energy Solutions and Liberty Energy. Marshank also mentions geothermal energy as an overlooked opportunity in the energy sector, advocating for investments in companies that can effectively navigate its engineering challenges.
For investors, he suggests focusing on individual securities rather than macro trends, establishing clear sell criteria in advance to manage risks effectively.
