Hundreds of billions are flowing into AI infrastructure, and the shape of the curve already rhymes with 1999.
At the turn of the millennium, we witnessed the Dotcom bubble.
Investors were pouring billions into Internet startups with little more than a “.com” in their name. Valuations skyrocketed, expectations exploded — until reality caught up. 💥
The bubble burst.
Hundreds of companies disappeared.
And yet — the Internet itself didn’t just survive. It thrived.
Two decades later, the digital economy is worth a multiple of everything that was ever invested back then.
Today, I see clear parallels with Generative AI.
The hype is extraordinary.
Hundreds of billions are being spent on data centers, GPUs, and model training — in some cases on a scale that no realistic business model could justify.
Even major players are funding each other’s growth in circular systems that feel uncomfortably similar to 2000.
An AI bubble is a market phase in which investment in AI-related assets — chips, data centers, model training, and startups — outpaces the technology’s near-term capacity to generate proportional economic value.
But here’s the catch:
Just because we’re in a bubble doesn’t mean the underlying technology isn’t real.
AI will absolutely change how we live and work — the question is simply how fast and at what cost.
Humans have always been bad at judging innovation:
- ⚡ We overestimate it in the short term — expecting overnight transformation.
- ⚙️ And we underestimate it in the long term — missing the quiet revolutions that follow.
Every breakthrough — from electricity to the Internet — has gone through the same “Hype → Crash → Plateau” pattern.
So common, in fact, that it has a name: Amara’s Law, or the Hype Cycle.
And AI might currently be climbing toward its (first) big fall.
From my own experience, I can say: the impact of AI is already tangible though.
I already see AI reshaping how we collaborate, communicate, and even how we think. — even this post was crafted with the help of (precisely directed) AI.
At the same time at my agency, we evaluate every use case critically.
We deploy AI only where it delivers measurable, sustainable value.
Because long after the hype has faded, the real differentiator will not be who used AI first — but who used it wisely.
Let’s not get blinded by the noise.
Let’s build the systems, ethics, and human judgment that make AI truly valuable — after the bubble bursts.
Where would you place AI on the curve today? Are you building for the hype — or for what comes after?
Read next:
- Optimize Your Website for Generative AI: 15 Proven Tips for GAIO — a practical guide to making your content visible to ChatGPT and other LLMs today.
- Vibe Coding for Designers: The Real Risk Nobody Talks About — the same hype-vs-reality pattern playing out at the tool level.
FAQ
What is an AI bubble?
An AI bubble is a market phase in which investment in AI-related assets — chips, data centers, model training, and startups — outpaces the technology’s near-term capacity to generate proportional economic value. The key point is that a bubble is not a verdict on whether the technology is real; it describes a gap between price paid and near-term returns. The Dotcom era, the 2008 housing market, and today’s AI cycle all share that gap. The technology kept advancing after every previous bubble. So will this one.
Will the AI bubble burst like the Dotcom bubble?
The mechanics rhyme, but the timing is anyone’s guess. In the Dotcom era, capital flooded infrastructure — fiber, servers, last-mile connectivity — that turned out to be real and useful, even though most of the companies building on top of it did not. AI looks structurally similar: data centers, GPUs, and foundation models are being built on assumptions about demand that may take a decade to materialize. Expect a sharp correction, not a slow fade — and expect the underlying technology to keep advancing regardless.
What is Amara’s Law?
Amara’s Law, coined by futurologist Roy Amara, is the observation that we tend to overestimate a technology’s effect in the short run and underestimate it in the long run. It is the mechanism behind Gartner’s Hype Cycle — peak of inflated expectations, trough of disillusionment, slope of enlightenment. The pattern repeats with every major innovation: electricity, the automobile, the Internet, and now AI. Knowing the pattern exists does not make you immune to it, but it does help you ignore the noise when everyone else is panicking.
Is it still worth investing in AI after the bubble bursts?
The technology will outlast the bubble, so the real question is what you invest in. The companies that survived the Dotcom crash were not the ones with the loudest pitch in 1999 — they were the ones that figured out how to translate infrastructure into durable products. The same will be true of AI: the model layer will likely commodify, while applied AI — domain-specific tools, AI-native workflows, agentic systems — compounds for decades.
What should companies do today to prepare for the post-bubble AI economy?
Start by separating AI experiments from AI strategy. Then measure value against concrete business outcomes, not demos. Finally, build internal literacy so your team can tell signal from noise. The goal is not to use AI everywhere, but to know where AI delivers measurable, sustainable value and where it does not. That distinction is what separates the winners from the casualties when the music stops.