Tether CEO Paolo Ardoino says Big Tech’s AI spending spree has four structural problems. And they could matter more than most people think.
He laid them out in a July 4 post on X. Here they are:
1. AI is priced too cheap. Companies are subsidizing compute to win customers. That makes growth look good. But if they raise prices, users might bolt. If they don’t, margins stay thin.
2. Profit timelines don’t match. Data centers and GPUs cost billions upfront. AI profits could take years. That gap between capital spending and returns is getting bigger — and harder to justify.
3. Hardware ages fast. AI chips lose relevance in 3 to 5 years. But the debt financing them assumes much longer payback. Companies may have to replace expensive gear before it’s paid off.
4. Open-source is eating their lunch. Free AI models are getting good fast. If they’re good enough, why pay for commercial AI? That weakens pricing power across the board.
Ardoino’s warning lands amid growing unease. Chinese hedge funds have called AI stocks a “super bubble.” JPMorgan projects $5.5 trillion in AI spending by 2030. Alphabet, Amazon, Meta, and Microsoft are on track to spend $720 billion this year alone.
The Bank of England warned last October that AI valuations look dot-com-bubble-ish. Infra spending may need trillions — much of it debt-financed.
To be fair, the optimists have a point too: today’s biggest AI spenders already have strong earnings and real businesses. That’s different from 1999.
But Ardoino’s argument cuts through the noise. The risk isn’t just market hype. It’s inside the economics of AI infrastructure itself. If pricing, hardware lifecycles, and competition don’t align, the math stops working.
Worth paying attention to.
