This week Jerome Powell, Chairman of the Fed, felt that it was necessary to claim, ex cathedra, that AI is not a bubble like other recent bubbles. Several other experts, economists and stock analysts, have been quick to agree. And, while I hope that they are right, it’s worth taking a skeptical look at the situation.
Bubbles are funny things. Economists can’t even agree on what one is — only that you know you’ve been in a bubble after it bursts. I have had a front row seat to two bubbles bursting and it wasn’t pretty.
For most Americans, the bubble that hit closest to home was the housing crash of 2008. The early 2000s felt like an era of unstoppable prosperity. The engine of that boom was housing. Americans treated their homes like ATMs — borrowing against equity, flipping houses, cashing out. Behind the scenes, lending had become dangerously lax.
I saw it firsthand. In 2008, we spun LendingTree out of IAC at the height of the housing crisis, and I spent the next twelve years on the board. I met plenty of good people — and plenty of good bankers — trapped in a bad system. Loans were being approved that borrowers had no plausible way to repay. They were called NINJA loans — No Income, No Job, No Assets. Asking for credit reports or W-2s was frowned upon. Managers told staff, “Don’t ask questions when you won’t like the answers.”
Bubble Time Bomb #1: Bundled Risk
Many of those shaky loans were packaged into “safe” mortgage-backed securities. The theory was diversification — that a bundle of risky loans would somehow become low-risk. In practice, it multiplied exposure. When the defaults came, they almost took down the U.S. banking system.
Everyone was happy until they weren’t. Homeowners, bankers, politicians — all got swept up in the euphoria. Then the U.S. housing market lost an estimated $6 to $9 trillion in value, six million homes were foreclosed, and by 2011 roughly 30% of homeowners owed more than their houses were worth.
A few years earlier came another bubble: the dot-com boom. In the late 1990s, venture investment in internet companies rocketed from about $1 billion in 1995 to $30 billion by 2000. Over 400 startups went public between 1998 and 2000. Most had catchy names but no revenue. When the music stopped, the NASDAQ collapsed from 5,000 to about 1,100, wiping out trillions in paper wealth. I was a participant in the chaos. I was the CEO of a venture-funded, web media startup. We survived. I sold it. But it was absolutely harrowing.
Bubble Time Bomb #2: Valuations Built on Dreams
The dot-com era assumed every company would grow into its dreams. When reality failed to cooperate, the market snapped back fast.
Bubble Time Bomb #3: Intertwined Finances
The era also birthed creative accounting tricks like “round-tripping”: I invest $1 million in your company, you spend $900K on my services, and we both book inflated revenue. For a while it looked brilliant — until it didn’t. Remember AOL-Time Warner? Valued at $350 billion in 2000, mostly because 23 million people were still dialing up to get online. That revenue vanished, and so did the dream.
Yet out of that wreckage came today’s digital economy. The internet infrastructure built during the bubble became the backbone for Amazon, Google, and YouTube. Bubbles, it turns out, sometimes leave useful rubble.
⸻
Déjà vu: The AI Boom
Now it’s artificial intelligence that’s driving the euphoria. Venture investment in AIstartups has exploded from roughly $12 billion in 2015 to over $130 billion in 2024, according to PitchBook and CB Insights. AI-related companies — public and private — now command a combined market value near $20 trillion, up from around $6 trillion a decade ago.
In the first half of 2025, AI startups attracted over half of all global venture capital, and in the U.S. they represented nearly two-thirds of every VC dollar invested. Nvidia’s market cap surged past $4 trillion this summer, briefly eclipsing Apple’s — a symbol of just how concentrated and super-charged this boom has become.
Sound familiar? Several of those bubble time bombs are ticking again, if to a lesser extent:
• Bundled Risk (AI Edition): The new version isn’t mortgage tranches — it’s model dependencies. Thousands of startups are built on a handful of foundational models (OpenAI, Anthropic, Google Gemini). If one falters — say, a licensing change, security failure, or data-cost spike — dozens of businesses could implode overnight.
• Valuations Built on Dreams: Some startups are valued at 20–30× projected revenue for products that don’t yet exist. It’s the same “we’ll grow into it” logic from 1999, dressed in machine-learning jargon.
• Intertwined Finances: AI companies are often both customers and investors in each other. A model provider funds an app-builder that, in turn, spends its funding buying compute time from the investor. It’s a high-tech version of round-tripping — everyone looks richer than they are.
The fates of the major players in the AI business are not independent Microsoft is a major shareholder in OpenAI. Everyone buys from Nvidia. Nvidia invests in Intel. And so on.
⸻
So Why Might This Not End Like 2008 — or Even 2000
There are differences that matter:
1. Real Revenue and Demand.
Unlike most dot-coms, many AI firms already have paying customers and tangible cost savings. Enterprises are deploying AI in customer service, logistics, and software engineering. There’s real revenue — if not always real profits.
2. Infrastructure With Utility.
The GPUs, data centers, and cloud infrastructure being built for AI have cross-industry value — from healthcare to defense to entertainment. Even if hype cools, the assets will still be useful.
3. Less Leverage, More Patience.
So far, AI’s exuberance is mostly equity-funded, not debt-driven. There’s no analog to sub-prime mortgages hiding in the plumbing. If (or when) the market corrects, it’s more likely to deflate than detonate.
4. Broader Adoption.
AI isn’t a single-sector phenomenon; it’s spreading into manufacturing, medicine, energy, and finance. That diffusion reduces the odds of a single “pop” moment.
Still, a correction is inevitable. Multiples this high rarely hold. Some of today’s unicorns will quietly vanish; others will merge or pivot. But that doesn’t mean the promise of AI evaporates. Like the internet after 2002, the useful parts will survive and scale.
Should You Care?
If you’re an investor, yes — selectively. If you’re an average consumer, probably less so. The 2008 bubble hurt ordinary Americans because homes and mortgages sat at the core of family wealth. The 2000 bubble hurt portfolios, not livelihoods. The AI boom, so far, sits somewhere in between. You might not lose your job or your house if it cools — but you could lose your illusions about how fast “the future” arrives.
Even if you are a conservative investor, the odds are good that your mutual funds and pension funds would be hit hard by a drop in the value of the AI industry. According to S&P 500 data, five large tech/AI-exposed companies (Apple Inc., Nvidia Corporation, Microsoft Corporation, Amazon .com Inc. and Alphabet Inc.) together represented about 28.8% of the S&P 500’s total market capitalization. So in short: a relatively small number of “big AI / big tech” companies now constitute roughly one-quarter to one-third of the major U.S. large-cap indexes.
The takeaway: AI will reshape the economy, but not every AI company will survive to see it. The bubble risk is real but not existential — at least, not yet.
As someone who lived through multiple booms and busts, my advice is simple: be excited, but not euphoric. Watch for those bubble bombs — bundled risk, dream-based valuations, and cozy financial interlocks. They’re all out there again, humming softly in the background.
History doesn’t repeat, but it rhymes — and right now, the melody sounds familiar.
Title Image I tried a number of different ways to generate a title image that I liked using both Chat GPT5 and Google Gemini. In this instance Gemini was much more successful. at generating an image that captured the essence of my post