America’s biggest banks have found a new growth engine in artificial intelligence. They lend to the tech giants building it, and they use it to cut their own costs. This week’s third-quarter
America’s biggest banks have found a new growth engine in artificial intelligence. They lend to the tech giants building it, and they use it to cut their own costs.
This week’s third-quarter earnings calls showed how much confidence, and how much caution, sits behind that trade.
A fresh note from Bank of America adds to the unease. It says the AI spending surge is following the path of past investment booms that ended in disappointment.
Related: Elon Musk has shopping advice after Amazon blocks his AI
BofA investment strategist Sebastian Raedler says hyperscalers are now expected to spend $1 trillion over the next 12 months, up from $330 billion in early 2025.
Hyperscalers are the cloud giants that run huge networks of data centers.
That's equal to about 3% of U.S. GDP, more than the 2.2% spent during the 1870s railroad boom and the 1.1% during the 1990s tech boom, as per the note.
Raedler writes that “most major capital cycles of the past have ended in disappointment,” even when the technology delivered.
High early profits pull in too much investment, and the extra capacity then drags prices down.
According to BofA, that pattern is starting to show. Average token prices have fallen about 50% since summer. Tokens are the units of text AI models process and bill for.
Data from Ramp AI shows U.S. business spending on AI has dipped slightly since August, even as token use climbs.
Bank leaders are mostly bullish, with some guardrails.
Goldman Sachs CEO David Solomon told analysts the bank is proud of its role in AI financing.
He added that no one knows how demand for capital and computing power will evolve.
Wells Fargo CEO Charlie Scharf called the buildout “a complicated macroeconomic force in its own right,” citing heavy leverage across some supply chains.
Citigroup CEO Jane Fraser is less worried. “There’s simply not enough compute,” she said, arguing that higher interest rates won’t slow the buildout.
Citi also reports internal gains. It says it now builds products up to 40% faster, and AI coding tools have cut its tech vendor costs by 30%.
JPMorgan Chase CFO Jeremy Barnum said the benefits of AI take time to show up at a bank of JPMorgan’s size, so they aren’t visible in quarterly results yet.
