Dot-Com vs AI Bubble: Is It Different This Time?

“This time is different” are the four most dangerous words in finance. They are usually spoken right before a market collapse to justify insane valuations.  But when comparing the AI boom of 2025 to the Dot-Com bubble of 2000, the phrase might actually be theoretically true. The AI boom does not look like a replay of the Dot-Com crash; it is a completely new type of financial cycle.

Image by Alexa from Pixabay

In 2000, we had a Valuation Bubble where stock prices were detached from reality. In 2025, we might be in the midst of a Capacity Bubble (infrastructure spending is detached from current utility). To begin with, let’s take a quick look at the two companies at the heart of the respective booms, Cisco (NASDAQ:CSCO) in 2000, and Nvidia (NASDAQ:NVDA), the current AI bellwether. Separately, see Why Broadcom May Be One of the Best AI Bets.

In 2000, Cisco briefly became the world’s most valuable company with a market cap of over $500 billion, priced for decades of unrealistic growth. Nvidia today is a very different beast—high-margin, cash-rich, and still reasonably valued relative to its expansion.

Ask yourself – Is holding NVDA stock risky? Of course it is. High Quality Portfolio mitigates that risk.

Here’s the Cisco vs Nvidia Snapshot:

The most common mistake investors make is assuming that because Nvidia’s market cap stands at over $4 trillion, it is “expensive” like Cisco was. The data proves otherwise.

• P/E: 200x for Cisco, based on its peak stock price in 2000 and its earnings for the year. This compares to about 38x forward earnings for Nvidia
• Growth: CSCO 55% revenue CAGR from 1997 to 2000, Nvidia 70% CAGR over last 3 years on a much larger base.
• Cash Flow: About $1.3 billion per quarter for Cisco, vs a staggering $25 billion per quarter for Nvidia.
• Customers: Fragile dot-com startups vs cash-loaded Big Tech — many of Cisco’s clients collapsed; Nvidia most definitely won’t. Related: Nvidia Built An AI Moat. Can Rivals Find The Drawbridge?

Cisco was a bubble because the stock price was fake. Nvidia may be a bubble because the demand could be temporary.

1. Capacity vs. Utility: The Great Mismatch

So, the problem isn’t the price of the stock; it may be the utility of the product.

We are currently witnessing a historic disconnect between Capacity (the infrastructure being built) and Utility (the value being extracted from it).

• Capacity (The Spend): Microsoft, Google, and Meta are collectively spending well over $200 billion a year on AI data centers and chips. OpenAI has somehow committed to over $1.2 trillion in spending in the coming years.
• Utility (The Return): Sequoia says AI needs $600B new annual revenue to justify it. Reality? OpenAI is doing $20 billion run-rate.
• The Reality: Currently, the actual revenue from end customers paying for AI services (like coding assistants or chatbots) is a fraction of that. OpenAI has about 800 million weekly users — only 5% to 10% pay ($20–$200/month). Over 90% freeload.

2. The Financing of the Spending

The second major difference is how this bubble is being funded. In 2000, the bubble was funded by Vendor Financing – companies like Cisco lent money to unprofitable startups so those startups could buy Cisco routers. When the startups failed, the money vanished.

Today, the financing is split into two very different buckets:

A. The “Safe” Bucket: The Hyperscalers Microsoft, Google, and Amazon are paying for their capacity with Operating Cash Flow. They are the most profitable companies in history. If the AI bet fails, they will lose money, but they will not go bankrupt. They are solvent customers.

B. The Risky Bets:

• The Neo-Clouds: A prominent subset of this group — firms like CoreWeave and Lambda Labs — buys billions of dollars of GPUs to rent them out. They operate on thin margins, volatile demand, and high leverage. Much of this sector uses debt secured directly by the GPUs. That’s precarious because these chips depreciate fast as each new generation makes the previous one less valuable.

• The Core Risk: If rental prices drop due to excess capacity, the collateral value falls. That can trigger forced selling of GPUs to repay lenders, dumping hardware into the market and crushing prices further. It’s a classic negative feedback loop.

• Circular Investment Model: On top of this, chipmakers themselves are intensifying the cycle. Nvidia is investing massive sums into AI leaders such as OpenAI, who then use that capital to buy Nvidia’s own chips. Nvidia may invest up to $100 billion into OpenAI

• Equity-Linked Subsidies: AMD-OpenAI have a relatively strange arrangement — where AMD grants OpenAI warrants for a potential 10 percent stake at $0.01/share — effectively acts as a subsidy. It potentially allows a cash-constrained, but very high profile buyer to finance multi-gigawatt purchases using the supplier’s own soaring equity, blurring customer and owner and reinforcing bubble concerns.

3. How It Could All Unravel

The Final Verdict

Is this a bubble? Quite likely. Is it the Dot-Com bubble? No.

• 2000 was a bubble of Junk: Besides the major network and software infrastructure companies which still exist today, investors ended up also investing in shares in fake companies with no business models. Think Pets.com.
• 2025 is a bubble of Anticipation: Investors believe they are funding the greatest infrastructure project in human history, but this is coming before the utility and economics have been fully proven.

The Trefis High Quality (HQ) Portfolio, with a collection of 30 stocks, has a track record of comfortably outperforming its benchmark that includes all 3 – the S&P 500, S&P mid-cap, and Russell 2000 indices. Why is that? As a group, HQ Portfolio stocks provided better returns with less risk versus the benchmark index; less of a roller-coaster ride, as evident in HQ Portfolio performance metrics.

Invest with Trefis Market-Beating Portfolios

See all Trefis Price Estimates