In partnership with
Every generation gets a technology so exciting that investors stop asking what it costs. Right now it is AI. This year, four companies, Alphabet, Amazon, Meta, and Microsoft, will spend roughly $725 billion between them on AI infrastructure, up about 77% in a single year, and heading toward a trillion in 2027.
That is real money chasing a real revolution. It is also exactly the setup that has produced some of the most expensive mistakes in market history.
So is AI the new dot-com? The honest answer is more interesting than yes or no. Let me show you what history actually teaches, starting with a company most people have forgotten was a cautionary tale.
Alphabet, Amazon, Meta and Microsoft are guiding to about $725 billion of capex in 2026, up roughly 77% in a year.
The 60-Second Version
Big Tech will spend about $725 billion on AI in 2026, up roughly 77% year over year, on pace for around $1 trillion in 2027.
The AI boom has real revenues behind it, which makes it different from the many dot-com companies that had none.
But 'this company is great' and 'this stock is a good buy' are two different questions. Cisco proved it.
A bubble is a story that has outrun the cash flows. The test is always the price, not the technology.
The company that was right, and still cost investors 80%
In 2000, the hottest stock in the world was Cisco Systems. If AI has Nvidia selling the picks and shovels today, the internet had Cisco selling the routers and switches. On the 27th of March 2000, Cisco became the most valuable company on earth, worth about $555 billion.
Then the bubble found the pin. The Nasdaq peaked that same month at just over 5,000 and fell about 78% over the next two and a half years, to around 1,100 by October 2002, erasing more than $5 trillion of value. Cisco itself fell roughly 80%, wiping out about $431 billion. A market bubble had met reality.
Here is the part that should stay with you. Cisco the business was fine. Its revenues kept growing for the next two decades and it stayed profitable throughout. And yet the stock did not reclaim its 2000 high until December 2025, twenty-five years later. The company was right about the internet. The stock, at that price, was catastrophically wrong. A great company can be a terrible investment if you overpay for it.
Why this time really is different (in part)
It would be lazy to stop there and shout 'bubble.' Today is genuinely different from 2000 in one crucial way: the leaders have enormous real revenues and profits. Many dot-com darlings had almost no sales at all. Today's AI leaders print cash.
Nvidia trades around 18 times its sales as of September 2026. Rich, but below the near-30 times Cisco reached at its 2000 peak.
As of September 2026, Nvidia trades at roughly 18 to 20 times its sales. That is expensive by any normal standard, but it is meaningfully below the near-30 times that Cisco reached in 2000, on top of a much larger and more profitable business. So no, this is not a carbon copy of the dot-com mania. The companies are real, and so are the earnings.
Why 'different' is not the same as 'safe'
The trap is to hear 'it is different this time' and relax. Different is not safe. Three things about today rhyme uncomfortably with 2000.
A huge bet on demand that has not fully arrived. The $725 billion of capex assumes AI revenue will show up to justify it. The gap between the spending and the payoff is real, and it is widening.
The assumption that today's leaders stay the leaders. Cisco was the undisputed leader and still fell 80%. Being the best company in a hot sector does not protect the stock.
Valuations that need the future to arrive on schedule. When most of a company's value sits in far-off cash flows, even a delay, not a failure, is enough to halve the price.
None of this says AI will crash. It says the price leaves little room for disappointment, which is the defining feature of every late-stage boom.
How to value a bubble, or avoid one
My rule: a bubble is a story that has outrun the cash flows. You do not need to call the exact top. You need a way to tell whether a price is asking too much of the future. Three checks do most of the work.
Run a reverse DCF. Instead of guessing growth, ask what growth the current price already assumes. If a company has to dominate the entire planet to justify its value, that is your answer.
Watch price-to-sales against history and against the cash actually coming in. A rising multiple on flat cash is a warning.
Separate the technology from the stock. AI can change the world and still be a poor investment at the wrong entry price. Both were true of the internet.
This is just valuation discipline applied when it is hardest to apply, in the middle of a story everyone believes.
So, is it a bubble?
Parts of AI are almost certainly overvalued. Parts may genuinely be the foundation of the next decade. Both can be true at once, and usually are. The job is not to predict the top. It is to refuse to pay any price for a good story.
Remember how the Cisco story ends. Its investors were right about the internet in every way that mattered. It still took them twenty-five years to break even. The surest way to lose money on a revolution is to overpay for it.
Reply and tell me the AI stock you want me to run a reverse DCF on. I will pick one apart in a future issue.
Best regards,
Dheeraj Vaidya, CFA, FRM
Ex-JPMorgan & CLSA
Co-founder, WallStreetMojo
P.S. If you want this plus everything else we teach, our WallStreetMojo All-Access Membership includes the Financial Modeling and Valuation Course along with 50+ courses and certifications covering Investment Banking, Equity Research, AI, and more.
Join Fin, Anthropic, and Clay at Pioneer on October 7th
Pioneer, the summit where CX leaders redefine what’s possible, is on October 7th.
Join leaders from Fin, Anthropic, and Clay for an insightful conversation on the state of AI transformation.
You’ll discover how some of the most innovative minds in CX have transformed their organizations, learn how they think about CX, and hear how they're planning for what's next.
Join the conversation in San Francisco, or tune in virtually.