May 21, 2026
The lazy comparison says today is safer because valuations look lower on the surface. During the dot-com era, median PEs in many technology names pushed toward absurd territory, roughly 150 or higher in some cases, while today many of the dominant AI leaders trade closer to 35–45 times earnings. On paper that appears far more rational.
But bubbles are never just about valuation. They are about psychology, capital flow, and feedback loops.
That is where the comparison becomes far more interesting.
AI Bubble Psychology | Dot-Com Was Open Madness. AI Is Structured Optimism.
During the late 1990s, much of the insanity sat directly on the balance sheet. Many companies had almost no earnings, weak cash flow, and business models built more on presentation slides than operating leverage. Investors were buying future possibilities with almost no present economics attached.
Today is different.
The dominant AI firms generate enormous free cash flow. NVIDIA, Microsoft, Oracle Corporation, and the hyperscalers are real businesses with genuine profitability, fortress balance sheets, and enough cash generation to finance infrastructure at nation-state scale.
That changes the emotional structure completely because investors can now say, with some justification, “this time there is substance underneath the excitement.”
And they are not entirely wrong.
The technology is real. Productivity gains are probably real. Long-term transformation is likely real too.
But real technology has never prevented capital cycles from overshooting.
AI Capital Spending Cycle | The AI Loop Is Reflexive
What makes this cycle dangerous is not fake revenue. It is reflexive demand feeding itself through the ecosystem.
The structure currently looks something like this:
- hyperscalers order enormous volumes of AI chips
- chipmakers report explosive growth
- chip profits finance more infrastructure and startup investment
- startups consume cloud credits aggressively
- cloud providers report accelerating AI demand
- enterprise software firms increase AI spending defensively
- private capital subsidizes adoption
- rising valuations lower capital costs
- lower capital costs fund more spending
Then the loop repeats.
This is not fraud in the traditional sense. It is a self-reinforcing liquidity cycle where spending itself validates the growth narrative temporarily.
That distinction matters.
Defensive AI Capex | FOMO Capex Is Real
A meaningful portion of current AI spending does not appear driven by fully proven end-demand economics. It looks defensive.
The psychology increasingly sounds like:
- “We cannot afford to fall behind.”
- “Deploy first, monetize later.”
- “Scale before competitors lock the market.”
That creates what can only really be described as FOMO capex.
The closest historical analogy may not even be internet stocks directly, but the late-1990s fiber optic buildout. The internet absolutely transformed the world, but capital allocation detached from realistic near-term returns. Infrastructure was massively overbuilt relative to immediate monetization.
The infrastructure eventually mattered enormously.
Many investors still got destroyed.
That distinction is critical because markets constantly confuse “important technology” with “safe valuation.”
Those are not the same thing.
Mega-Cap AI Ecosystem | The Bubble Is More Sophisticated This Time
The current setup is structurally stronger than dot-com in some ways because the leaders are not fragile startups alone. A handful of mega-cap firms now generate enough free cash flow to finance entire ecosystems around themselves.
That did not exist to the same degree during the internet bubble.
Today the dominant firms act almost like semi-central banks inside the AI ecosystem:
- financing infrastructure
- subsidizing adoption
- funding startups
- supporting cloud expansion
- validating ecosystem demand
The structure therefore appears more stable in the short term because real earnings cushion the narrative.
But that may actually create a more systemic long-term risk because the optimism becomes harder to challenge while the cycle remains strong.
Reflexive Valuation Loop | The Real Danger Is Circular Optimism
The danger is not necessarily collapse through fraud or insolvency. It is valuation gravity eventually colliding with normalized growth.
The cycle increasingly feeds itself:
- high valuations justify enormous spending
- enormous spending creates explosive revenue growth
- revenue growth validates high valuations
- investors continue funding the expansion
Until eventually the system matures.
That is usually where the shift happens:
- AI capex growth slows
- enterprise adoption becomes less explosive
- pricing pressure appears
- infrastructure buildouts normalize
- investors realize future growth was partially pulled forward
Then multiples compress even if the companies themselves remain dominant and highly profitable.
That is the part most investors never fully internalize.
Tech Revolution Drawdowns | Great Companies Can Still Collapse 50%
A genuine technological revolution does not protect investors from repricing phases.
The railroad boom transformed transportation permanently. The radio boom changed communication. The internet reshaped commerce. Smartphones altered daily life globally.
Yet many leaders inside those revolutions still experienced devastating drawdowns once expectations outran monetization curves.
That pattern repeats because markets always overshoot before they normalize.
The crowd consistently makes the same psychological mistake:
“If the technology changes the world, the stock price must always justify itself.”
History says otherwise.
AI Bubble Durability vs Immunity | The Important Difference
This does not mean the AI boom collapses tomorrow. In fact, structurally it may persist longer precisely because the underlying companies are so financially powerful.
That makes the bubble more durable.
But durability and immunity are not the same thing.
The current AI cycle resembles a more mature version of dot-com psychology, one backed by real earnings, real infrastructure, and real adoption, yet still vulnerable to the oldest force in markets:
expectations expanding faster than sustainable returns.
And when that gap finally matters, the repricing can still be brutal even if the technology itself succeeds beyond imagination.
That is the paradox most people miss.
The technology can win completely while investors still overpay for the privilege of believing in it too early.
