Everyone keeps asking if the AI bubble is about to burst.
After watching three hype cycles come and go, I think that's the wrong question.
The right one is quieter, and far more uncomfortable: which AI is a bubble, and which is the real economy being rebuilt underneath the noise?
Because both are happening at once. And if you can't tell them apart, you'll either panic-sell a generational shift or pour money into a mirage. I've seen smart people do both.
Here's how I read it 👇
𝗪𝗲'𝘃𝗲 𝗯𝗲𝗲𝗻 𝗵𝗲𝗿𝗲 𝗯𝗲𝗳𝗼𝗿𝗲 — 𝗮𝗻𝗱 𝘁𝗵𝗮𝘁'𝘀 𝘁𝗵𝗲 𝗽𝗼𝗶𝗻𝘁
In 1999, the internet was "obviously" a bubble. It was. The Nasdaq lost nearly 80% of its value and hundreds of dot-coms vanished.
And yet — the internet went on to swallow the entire global economy. Amazon, the poster child of dot-com excess, is now worth more than most countries' GDP.
So when people ask whether the AI bubble is about to burst, fifteen years in this industry give me a frustrating answer: yes, partly — and it won't matter the way you think.
The bubble bursting and the technology winning are not contradictions. They're the same story at different speeds.
𝗧𝗵𝗲 𝗰𝗮𝘀𝗲 𝘁𝗵𝗮𝘁 𝘁𝗵𝗲 𝗯𝘂𝗯𝗯𝗹𝗲 𝗶𝘀 𝗿𝗲𝗮𝗹
→ Valuations have detached from revenue. A wrapper around someone else's model + a slick demo is raising like it's a sure thing for a decade. That's a lottery ticket with a pitch deck.
→ Infrastructure spend is staggering. Hundreds of billions into data centers, GPUs and energy, betting demand catches up. Some will. Some is built for a future that arrives late, or never.
→ "AI" has become a magic word. The moment a label adds 30% to a valuation, incentives have outrun reality.
→ ROI is lagging. Many enterprises that rushed to deploy generative AI quietly admit the returns aren't there yet. The gap between promise and P&L is where bubbles live.
𝗧𝗵𝗲 𝗰𝗮𝘀𝗲 𝘁𝗵𝗮𝘁 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗻𝗼𝘁 𝟭𝟵𝟵𝟵
→ Real revenue exists, at real scale. The biggest players are earning tens of billions — actual income, not just traffic. The picks-and-shovels layer is wildly profitable today.
→ The technology genuinely works. It's already writing production code, accelerating drug discovery, and running customer operations at companies you use every week. This isn't vaporware.
→ Adoption is faster than any platform shift I've seen. AI products hit in months what took the internet years.
So we have a paradox: inflated valuations sitting on top of a genuinely transformative, revenue-generating technology. That's not a clean bubble. It's a bubble AND a boom, tangled together.
𝗦𝗼 — 𝗶𝘀 𝘁𝗵𝗲 𝗔𝗜 𝗯𝘂𝗯𝗯𝗹𝗲 𝗮𝗯𝗼𝘂𝘁 𝘁𝗼 𝗯𝘂𝗿𝘀𝘁? 𝗠𝘆 𝗵𝗼𝗻𝗲𝘀𝘁 𝗰𝗮𝗹𝗹:
A correction is likely. A collapse of the underlying technology is not.
The froth will come off. Most "thin wrapper" startups won't survive the next downturn. Valuations will compress — painfully — for anyone who confused a press release for a product.
But when the dot-com bubble burst, the internet didn't disappear. It consolidated, matured, and then quietly took over everything.
AI is on the same arc — just faster, and stranger.
The bubble bursting won't kill AI. It will separate AI — the substance from the costume.
𝗪𝗵𝗮𝘁 𝘁𝗼 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗱𝗼 𝗮𝗯𝗼𝘂𝘁 𝗶𝘁
✅ Don't bet on the label. Bet on the cash flow. Ask any "AI company": what survives if we delete the word AI from your deck?
✅ Separate infrastructure from the application layer. Shovel-sellers do fine even when miners don't.
✅ Watch real adoption, not announcements. Usage that sticks beats funding that's announced.
✅ Build for the decade, not the news cycle. The people who built quietly through the 2000 crash won the 2010s.
The question was never really if the AI bubble will burst. Bubbles always burst.
The real question is whether you'll still be standing — and building — when the smoke clears.
I think the ones who treat this as a marathon, not a meme, will be.
What's your read — a 1999-style correction, or something genuinely new?
I'd like to hear from the people building and investing through it. Drop your take below 👇
