America’s Next Bubble Is Already Inflating

The biggest bubble in the United States isn’t in housing, crypto, or even the stock market—it’s in artificial-intelligence hype itself. A recent dispatch by Tech Buzz China founder Masha Ma should have sent chills down Silicon Valley’s spine. Ma, fresh from a field tour of Chinese AI facilities, admits she’s “no energy expert,” yet after weeks of briefings and site visits she writes: “China no longer worries about whether it can power its data centers. The problem is solved.” Another specialist added that China keeps national power reserves at 80–100 percent above peak demand—double what it actually needs—so adding AI workloads doesn’t even register as a grid risk.

Compare that with the U.S., where utilities are scrambling to plug new generators into an aging grid. Permits take years, locals file endless lawsuits, and market rules differ from state to state. The bottleneck isn’t technology; it’s governance.

I flagged this mismatch more than a year ago: the limiting factor for American AI isn’t kilowatts—it’s what you do with the algorithms once you have them. AI needs physical outlets: factories, logistics networks, and, ultimately, consumers. The United States has neither the industrial mass nor the population size to absorb its own breakthroughs. We may end up perfecting the software only to watch it create value somewhere else—most likely in China’s far larger manufacturing and consumer ecosystem.

That irony is already showing up in the GDP numbers. Spending on new AI-focused data centers has overtaken personal consumption as the biggest swing factor in U.S. growth. In 2024 American households shelled out $18.3 trillion—about 70 percent of GDP—while China’s 1.4 billion people spent roughly $12.3 trillion, or 54 percent of its GDP. On paper, the U.S. still looks like the consumption superpower. But those figures are an accounting mirage.

America’s “purchasing power” is largely a financial fiction. Every 10 percent rally in the S&P 500 nudges consumer spending up by half a percentage point, according to Federal Reserve data. Between 2020 and 2022, rising stock and home values added $12 trillion to household balance sheets, which in turn helped lift consumption by roughly 2.1 percent. Yet that wealth effect was underwritten by debt: Americans borrowed an extra $4.6 trillion in consumer credit during the same period, pushing household leverage to 28 percent of total debt.

Now the credit card is maxed out. With households unable or unwilling to borrow more, the U.S. economy needs a new narrative. Enter AI. Wall Street has seamlessly swapped the old lever—consumer debt—for a shinier one: AI infrastructure investment. The trouble is, after eighteen months of nonstop marketing, where are the measurable gains? Higher productivity? Cheaper goods? New industries? So far, the returns look vanishingly small.

The real risk facing American AI, then, isn’t a shortage of electricity. It’s that the country is inflating yet another financial bubble—this time around data centers, chips, and model-training costs—without a credible plan for turning code into broadly shared prosperity. When the hype meets that reality, the bubble may pop long before the power ever runs out.