Despite tariff and immigration policies roiling businesses, the United States economy remains relatively stable, and several economists credit the artificial intelligence industry for the resilience. “Artificial intelligence machines-in quite a literal sense-appear to be saving the US economy right now,” Deutsche Bank’s George Saravelos told clients, arguing that without tech-related spending the country would be near recession this year. Economist Paul Krugman has made similar observations, underscoring how a surge of investment and corporate adoption tied to artificial intelligence is propping up demand.
The capital outlays are substantial. Artificial intelligence firms are pouring money into infrastructure and development, while enterprises are spending heavily on new tools. A flagship data centre in Abeline, Texas, part of the Stargate programme led by Oracle, OpenAI and Japan’s SoftBank, recently came online. Around the same time, Nvidia said it would invest in OpenAI and supply data centre chips, and it became the first US company to hit a trillion-level market value, soon followed by Microsoft, which has cited artificial intelligence as a key driver of business demand. Alphabet and Meta have also increased their commitments to artificial intelligence initiatives.
The concentration of gains has sparked bubble concerns. “Seven companies are pulling more than 400 others forward,” said Campbell Harvey of Duke University, noting that the S&P 500’s recent strength is dominated by a handful of tech firms deeply engaged in artificial intelligence. Still, Carl Frey of Oxford University argues that while share prices look elevated, there is real revenue behind the data centre buildout, adding that today’s environment is “nowhere near tulip mania territory.”
Beneath the market enthusiasm, adoption signals are cooling. Frey warns that early corporate adopters are narrowing efforts to projects that clearly save or make money. IBM and Klarna, for example, replaced thousands of customer service roles with artificial intelligence before reversing some of those moves after finding the tools could not match human performance in all cases. A recent MIT report found that 95 percent of companies using artificial intelligence are not seeing significant revenue acceleration, and US Census Bureau data shows adoption by large companies has begun to slow.
Integration challenges remain a drag. Georgetown University’s Cal Newport says the underlying models in these programmes are “too unreliable” to automate many jobs at scale, and the anticipated rapid displacement has “simply not come true.” A Stanford study nonetheless found entry-level roles in customer service, accounting and software development have fallen 13 percent since 2022 amid large-company adoption of artificial intelligence tools. Whether or not a bubble is forming, a sharp pullback could hurt the broader economy. Frey cautions that unless a bust triggers a financial crisis, the bigger risk is that artificial intelligence has not yet delivered a clear, broad productivity boost that stagnating economies need.
