“DeepSeek’s power implications for AI training punctures some of the capex euphoria which followed major commitments from Stargate and Meta last week,” Jefferies infrastructure analyst Graham Hunt and his colleagues wrote in a note to clients Monday. “With DeepSeek delivering performance comparable to GPT-4 for a fraction of the computing power, there are potential negative implications for the builders, as pressure on AI players to justify ever increasing capex plans could ultimately lead to a lower trajectory for data center revenue and profit growth.”
Investors fear that the proliferation of cheaper, more efficient models may hurt the prospects of technology companies — and their suppliers — that are spending tens if not hundreds of billions of dollars on artificial intelligence investments.
Just last week, both Altman and Mark Zuckerberg, the founder and chief executive of Meta, announced huge new investments in artificial intelligence infrastructure.
Altman’s OpenAI is part of Stargate, the joint venture with Microsoft and SoftBank that got a splashy White House-based announcement and promises to invest $500 billion in artificial intelligence infrastructure. There was already some skepticism of these numbers, with Altman-nemesis Elon Musk charging that certain members would be unable to fulfill their ends of the deal, Microsoft Chief Executive Satya Nadella told CNBC from Davos, “I’m good for my $80 billion.”
Zuckerberg, meanwhile, said late last week that his company was building a data center “so large it would cover a significant part of Manhattan,” which would require 2 gigawatts of electricity to power. (For scale, reactors 3 and 4 of the Vogtle nuclear plant in Georgia are a little over 1 gigawatt each.) He also said that Meta had planned up to $65 billion of capital expenditure this year.
These escalating announcements have been manna to investors in any company that provides the building blocks for large artificial intelligence systems — namely chips and energy, with companies like Nvidia, the chip designer, and power companies and energy infrastructure companies posting some of the best stock market performances last year.
But exactly how cheaper artificial intelligence plays out in terms of real investment remains to be seen. Late Sunday night Redmond, Washington-time, Nadella posted a link on X to the Wikipedia page for Jevons Paradox. The idea dates from 19th century Britain, and posits that increased efficiency in using a resource (in Jevons’ case, coal) could actually accelerate its depletion, as the resource becomes cheaper for the same economic output, encouraging more use of it (in Jevons’ case, iron).
“Jevons paradox strikes again!,” Nadella wrote. “As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of.”
Investors in chips and energy companies are hoping that’s the case; at least so far, the market doesn’t appear to agree.