Nvidia's stock rose more than 800% between 2022 and 2024. A single company — a chipmaker that most people couldn't have named five years ago — became one of the most valuable businesses on earth.
Either artificial intelligence is the biggest economic shift since the internet. Or this is a bubble, and the correction is going to be brutal.
The frustrating truth is: both things can be true at the same time.
What the AI Bubble Chart Actually Tells You
Every major technological bubble in history has followed the same rough shape: a slow build, a vertical run-up fueled by narrative and capital inflows, and then a reckoning — not necessarily because the technology was fake, but because the timeline was wrong.
Look at the dot-com bubble chart from 1995–2002. The Nasdaq ran up 400%+ in five years, then crashed 78%. But the underlying thesis — that the internet would transform commerce, communication, and media — was completely correct. Amazon survived. Google launched during the crash. The companies that didn't survive weren't wrong about the internet; they were wrong about how quickly the profits would arrive, and how much it would cost to get there.
The AI bubble chart has some of the same fingerprints. Valuations front-running revenue. Capital flooding into anything with "AI" in the pitch deck. Enterprise software companies repricing themselves 40-50% higher simply by adding "AI-powered" to their product pages.
But there's a key difference from 1999 that the bears tend to ignore.
The revenues are real.
The Nvidia Question
No serious discussion of the AI bubble holds up without addressing Nvidia directly.
The core skeptic argument: Nvidia is priced for a future where AI demand continues to compound indefinitely. Any slowdown in data center spending, any progress on more efficient training methods, any commoditization of GPU compute — and the multiple collapses.
That argument is not wrong.
But here's what the bears tend to understate: Nvidia's revenues are not speculative. In fiscal year 2025, Nvidia reported $130 billion in annual revenue — the majority from data center GPU sales. These aren't projected revenues. They're booked. Hyperscalers like Microsoft, Google, Amazon, and Meta are spending real money on real chips to build real infrastructure.
The nvidia ai bubble debate, framed properly, is not "is Nvidia's revenue fake?" The question is: will the demand sustain at the pace the market has priced in?
That's a legitimate uncertainty. A P/E multiple that assumes perpetual growth is always a bet on the future, not just the present. And futures can disappoint even when the present is strong.
Sundar Pichai and the $75 Billion Bet
In early 2025, Alphabet CEO Sundar Pichai announced the company would spend $75 billion on AI infrastructure in a single year.
That number became a flashpoint. To bulls, it signaled that the largest companies in the world — with real cash flows and disciplined capital allocation teams — believe the returns are coming. You don't spend $75B on a hunch.
To bears, the sundar pichai ai bubble concern is a different one: what if all of these companies are investing simultaneously, competing for the same AI future, and only a fraction of them can win? The internet era showed that the infrastructure buildout was real — but fiber-optic companies and telcos that overbuilt capacity still went bankrupt. The pick-and-shovel trade can overcorrect.
The honest read on Pichai's bet is that Google is spending $75B because not spending risks existential obsolescence. OpenAI and Anthropic forced their hand. That's a defensive investment as much as an offensive one.
Which means the capex isn't purely a bullish signal. It's also evidence that the incumbents are scared.
When Will the AI Bubble Burst?
This is the question everyone actually wants answered. Here's the honest version.
Nobody knows. And anyone who tells you they do is either lying or selling something.
What history does tell us is what the preconditions for a burst look like — and it's worth running through them.
Capex doesn't convert to revenue. The hyperscalers are spending enormous sums on AI infrastructure. If that spend doesn't translate into products customers will pay for at scale, the return-on-investment thesis breaks. A slowdown in enterprise AI adoption — slower than expected productivity gains, fewer businesses willing to pay premium prices for AI features — could trigger a pullback in data center orders, and Nvidia's revenue growth would decelerate sharply.
LLMs become a commodity. The models themselves — GPT, Gemini, Claude — are increasingly competing on price. If the underlying technology becomes a commodity, the value shifts to applications built on top, not to the infrastructure layer. That's fine for some players. It's bad for anyone priced as if the moat is permanent.
A killer app fails to materialize on schedule. Every technology bubble eventually faces a "where's the money?" moment. The internet's killer app was e-commerce and advertising — but it took a full decade after the 1999 peak to really land. If AI's productivity gains take longer to show up in GDP data and corporate earnings than the market expects, the correction comes, even if the long-run thesis is intact.
A macro shock provides the catalyst. Bubbles rarely burst on their own thesis failing. They usually burst because something external — a rate hike cycle, a liquidity crunch, a regulatory event — removes the fuel. An ai bubble bursting scenario most likely involves an external trigger, not a single "AI doesn't work" moment.
How to Position Yourself
You don't need a perfect read on the bubble's timing to think clearly about your own portfolio. You just need a framework.
1. Distinguish between AI as a technology and AI as an investment thesis. The technology can be real and transformative while the investment valuations are simultaneously inflated. These are not the same thing. Act accordingly when allocating capital.
2. Avoid concentrated exposure to single-narrative stocks. If your portfolio is significantly weighted toward AI-adjacent names that have run 300-500% in three years, you are holding a concentration risk that has nothing to do with the quality of the technology. Diversification exists for exactly this reason.
3. Dollar-cost average into broad-market index ETFs. Index ETFs give you exposure to AI's upside through the companies that will benefit — without the binary risk of single-stock concentration. If AI transforms the economy, your index fund captures it. If valuations correct sharply, you're not wiped out.
4. Treat AI tools as an income lever, not a portfolio strategy. The biggest financial benefit of AI for most people reading this isn't in their brokerage account — it's in their productivity. AI tools are compressing the time cost of writing, research, coding, and design. Solopreneurs who learn to use this leverage will earn more. That's a better play than trying to time a speculative trade.
5. Hold Bitcoin as a macro hedge, not an AI play. Bitcoin's value proposition is independent of whether the AI bubble inflates or deflates. It's a fixed-supply asset in a world of expanding monetary supply — that thesis doesn't change based on Nvidia's next earnings call.
FAQ
Is the AI bubble worse than the dot-com bubble?
It's a different shape. Dot-com companies largely had no revenue — Pets.com, Webvan, and others were burning cash on speculative infrastructure. Today's AI leaders — Nvidia, Microsoft, Google — have massive real revenues. The risk is not zero-revenue speculation; it's whether the multiples applied to those revenues are justified by the growth trajectory ahead. That's a more nuanced problem than 1999, and the correction, if it comes, will likely be more sector-specific than the indiscriminate crash of 2000.
Should I sell my tech holdings now?
That's a timing question, and timing questions are nearly impossible to answer correctly with consistency. Buying the dip vs. timing the market covers this well — research strongly suggests that time in the market beats time out of the market for most investors. If your AI exposure feels uncomfortable, the answer is probably gradual rebalancing, not a wholesale exit.
What would it actually look like if the AI bubble bursts?
A 40-60% decline in AI-adjacent stocks concentrated in a 12-18 month window, driven by a combination of macro shock and earnings disappointment. The underlying companies wouldn't disappear. Just as Amazon survived the dot-com crash, the market leaders in AI likely survive. But investors who bought at peak valuations could wait a decade to get back to breakeven — which is what happened to anyone who bought Cisco at its 2000 high.
Is Nvidia overvalued?
It depends entirely on what growth rate you assume over the next decade. Nvidia's valuation implies continued extraordinary growth in data center demand. If that growth arrives on schedule, the stock could be fairly valued. If AI capex spend plateaus or competition compresses margins, the current price looks expensive. There is no consensus answer — which is the honest truth about any company priced for a specific future.
The Bottom Line
Here's what gets lost in the AI bubble debate: the question isn't whether AI is real. It is.
Nvidia's revenue is real. The productivity gains from AI tools are real. The transformation of software development, content creation, and data analysis is already underway — not theoretical.
The question is whether current valuations accurately price when the full economic impact arrives, who captures it, and how large it actually turns out to be.
On all three of those questions, there is genuine uncertainty. The market has made a bet. That bet could be right. It could also be pricing in a future that takes twice as long to materialize — which is the pattern that has ended nearly every technology bubble in history.
The intellectually honest position: the AI bubble chart looks like other bubbles, but the underlying revenues are stronger than other bubbles. Real innovation. Real earnings. Real hype. And real expectations of future productivity gains that haven't yet shown up in the data. Both things are true simultaneously.
We can't know for sure whether the bubble will burst, or exactly when. But we can build a portfolio that doesn't require us to be right about that call.
Navigate the hype. Own the tools. Stay diversified.
Want a framework for investing in an era of rapid technological change? Join the Wealth Potion newsletter — no noise, no hype, just clear thinking on money and markets.