It's a question you hear everywhere now, from news anchors to family dinners. It's also a question where it's easy to sound certain and hard to actually be right. This page lays out what the evidence says as of late October 2026, without hype and without doom, so you can make up your own mind.
This is general information, not financial advice.
The short answer
Two things can be true at once. AI is a real, useful technology that millions of people use every day. And the amount of money being spent to build it may be running ahead of the money it earns, which is what people usually mean by "bubble." Nobody can reliably say whether, when or how sharply that gap closes.
What people mean by "bubble"
A bubble is when prices or spending race ahead of the value something can realistically produce, often fueled by excitement and borrowed money. When expectations reset, prices fall, sometimes sharply.
Importantly, bubbles can form around technologies that turn out to matter. The dot-com crash of 2000 wiped out many companies, yet the internet went on to change daily life. "Bubble" is about prices and spending, not about whether a technology is useful.
The case that AI is overheated
The worries focus on how much is being spent, how it's being paid for, and how much revenue is coming back.
- Huge spending. J.P. Morgan Global Research projects about $900 billion of AI capital spending in 2026 alone and $5.5 trillion by 2030 (J.P. Morgan, Oct 19, 2026).
- More of it is borrowed. The Bank for International Settlements reports that the five largest U.S. cloud companies issued $222 billion of bonds by September 2026, and that more borrowing is moving through complex off-balance-sheet vehicles. It notes that "good" and "bad" credit booms look similar in real time until productivity gains show up, or don't (BIS Bulletin 140).
- A revenue gap. Research firm Gartner says more than $3.5 trillion of AI infrastructure spending could be exposed to impairment by the end of 2030 (Gartner, Oct 14, 2026). As reported by Fortune, Gartner estimates AI data centers would need about $8.5 trillion in cumulative revenue through 2030, against about $3.6 trillion it expects AI to generate for them (Fortune, Oct 21, 2026; Fierce Network).
- Business returns are uneven. A widely shared 2025 MIT Project NANDA report found that 95% of organizations it studied saw no measurable profit-and-loss return from generative AI. It's worth knowing the sample was small (52 interviews and 153 survey responses) and preliminary (MIT NANDA, "The GenAI Divide: State of AI in Business 2025", July 2025, PDF copy; coverage: Fortune).
- Investors are getting pickier. CNBC reported in October 2026 that markets are now separating companies earning cash from the boom from those spending heavily on hope, with heavily indebted "neocloud" providers seen as most vulnerable (CNBC, Oct 21, 2026).
The case that AI is real
On the other side:
- People actually use it. AI chat assistants, coding helpers and writing tools have moved into everyday work. Our own tool directory lists more than 200 tools, most with free plans.
- Cloud revenue is growing. J.P. Morgan points to rising cloud revenue run-rates and contracted future revenue as the strongest data supporting the spending (J.P. Morgan).
- Skills command a premium. Upwork's 2026 research found freelancers doing AI work earn 34% more per hour than those who don't (Upwork Future Workforce Index 2026).
- Real limits slow the buildout. Power supply, chip and memory shortages, and local permitting are constraining how fast data centers get built, which can act as a brake on overbuilding (J.P. Morgan).
Both can be true
The most balanced reading is that AI looks like past "general purpose" technologies such as railways and electricity: genuinely transformative, and also prone to overinvestment along the way. The BIS notes that U.S. data center investment over 2025–32 is estimated to exceed, as a share of GDP, the railroad, electrification and telecom booms (BIS). Some of those booms paid off; others ended in crises. It's too early to tell which this is.
What it could mean for everyday people
If AI spending cooled sharply, you might notice:
- Investments. Many retirement funds hold big tech stocks. A sharp fall would show up there. This is a reason for the usual advice (diversify, think long term), not for panic.
- Tool prices and free plans. If companies need to earn more, free plans could shrink and prices could rise. Some tools could disappear, so keep copies of important work.
- Jobs. Tech hiring could slow. But the tools themselves wouldn't vanish. Learning to use them remains useful either way; see our job tracker.
- Electricity bills. Data center growth is already part of the debate about local power prices (J.P. Morgan). See our explainer on AI and the environment.
What we're watching
- Whether AI revenue grows fast enough to cover spending.
- How much of the buildout relies on debt.
- Whether businesses report measurable returns, not just pilots.
- Power and chip constraints.
We'll update this page as the picture changes. If you'd like the calm version in your inbox, our newsletter covers it.
Common questions
Is AI a bubble?
Spending may be ahead of revenue, which is the bubble worry, but AI itself is a useful technology. Both can be true. No one can reliably predict if or when a correction happens.
Will the AI bubble burst?
Nobody knows. Analysts disagree, and timing markets is notoriously hard.
If there's a crash, will AI tools disappear?
The big tools are unlikely to vanish, but prices, free plans and smaller tools could change. Keep backups of your work.
Should I sell my AI stocks?
We can't give financial advice. Consider talking to a licensed adviser and staying diversified.
Is it still worth learning AI?
Yes. Even after the dot-com crash, internet skills stayed valuable.