AI may be heading into its dot-com moment—a bubble bound to burst. But just like the internet, the real innovations will outlast the hype. The key is understanding the differences between the current bubble and it's dot-com predecessor, identifying what is bubble-proof, and having the conviction to act before the “safe bets” become obvious to everyone.
The launch of ChatGPT in 2022 heralded a quantum leap in our understanding of Generative AI and the potential it holds. Speculation and massive investment are currently dominating the space, but several forces are driving AI toward a correction:
"AI everything" is being pushed as a solution to every conceivable problem, with claims of out-of-the-box superintelligence, the ability to replace entire departments' worth of employees, and save companies millions of dollars. While there are numerous proven successes, the hype vs. reality will soon become apparent and disappoint many.
Too much venture capital is chasing too many thin, undifferentiated companies with suspect business models, many of which won’t survive.
Explosive demand for compute power and energy needed to train and operate large language models (LLMs) is outpacing supply, making large-scale AI adoption extremely expensive and stressful on existing energy infrastructure.
These dynamics look a lot like the dot-com buildup of the late 1990s.
The first internet bubble peaked in 2000 and then burst spectacularly, wiping out nearly $5 trillion in market value and 75% of the tech-heavy Nasdaq composite index. Despite generating more than $1.5 trillion in revenue (in 2024 dollars) in 2000, the internet still couldn’t sustain the wildly inflated valuations of the multitude of startups, many of which existed only as vaporware with no real path toward profitability. Even with real commerce happening (nearly $300 billion in e-commerce revenue that year), speculation on unproven business models was rampant.1
Generative AI today is generating less than $10 billion in revenue—a tiny fraction of internet-era numbers. Yet the costs of AI are enormous. According to Morgan Stanley, the transformative potential of generative AI could demand $2.9 trillion in global data center spending through 2028—$1.6 trillion for chips and servers and $1.3 trillion for infrastructure. 2
At first glance, the discrepancy seems alarming: the internet was pulling in trillions when its bubble burst, while AI is burning trillions in investment with only a few billion in direct revenue. Shouldn’t that be a flashing red light warning of a more disastrous crash looming?
It’s true that AI’s revenue-to-cost gap is much wider than during the dot-com boom. But there are critical differences:
In short: the discrepancy is real, but it isn’t purely speculative. The scale and nature of investment—combined with the fact that AI is being embedded into the very core of enterprise workflows—makes this moment different from 2000.
In the late ’90s, search engines were everywhere: AltaVista, Lycos, Excite, Yahoo!, Ask Jeeves. Most of them got massive traffic, but their business models were fuzzy, ad clutter was rampant, and their technology was shallow. They rode the wave of speculation, but when the dot-com bubble burst, many of them collapsed or faded into irrelevance.
Google, by contrast, had a few things going for it that separated it from the speculative herd:
That combination—innovation + clarity of focus + actual ROI for advertisers—made Google one of the few companies to grow stronger in the ashes of the bubble.
Google wasn’t the only one, either. Amazon survived the crash by sticking to fundamentals (efficient logistics, customer obsession) while Pets.com and others collapsed. Salesforce doubled down on the SaaS model when it was still seen as risky, laying the foundation for an entire industry.
The parallel to AI today is sharp: there are hundreds of AI-powered “wrappers” and me-too products right now, many chasing hype with little defensible intellectual property. But the long-term bubble-proof winners will be those with real technical differentiation, embedded workflows, and business models tied to measurable results.
As the excess burns off, the useful will stick. The AI that lasts will be:
Adoption strategy matters too. A recent MIT study found that purchasing AI tools from specialized vendors and building partnerships succeeds about 67% of the time, while internal builds succeed only one-third as often.3 The companies that treat AI like a muscle to be built—by piloting now, aligning reliable partners, testing ROI, and scaling strategically—will be the ones ready when the hype cycle clears.
AI is still in its early days. With a widely expected correction coming, it's easy to see why many companies are unsure about the correct way forward and are hesitant to take meaningful steps toward AI transformation. A McKinsey survey on the state of AI shows that while 92% of companies plan to increase investment in AI, only 1% of executives view their generative AI rollouts as "mature." Fewer than a quarter have established clearly defined roadmaps to develop AI solutions.4
There are numerous hurdles to effective AI adoption - concerns about data accuracy and availability, lack of internal AI expertise, unclear investment needs and ROI ambiguity, and data privacy concerns.5 But history tells us that waiting carries its own risks. During the dot-com era, the companies that sat on the sidelines lost ground to competitors who experimented, learned, and iterated. The same pattern is repeating with AI.
The winners of the next decade won’t be those who waited for certainty; they’ll be the ones who built muscle early—piloting AI responsibly, learning from missteps, and embedding AI into the DNA of their business. In other words, the bold and pragmatic will come out ahead.
If history repeats, many AI companies will disappear when the bubble corrects. But just like Amazon, Google, and Salesforce, the winners that endure will reshape entire industries.
Here’s the difference: in 2000, leaders could afford to wait until the smoke cleared. Today, AI is moving too fast—and the experience competitors are gaining now won’t be easy to catch up on later.
The companies experimenting today aren’t just buying tools—they’re building institutional knowledge, trusted data pipelines, and the ability to separate hype from ROI. Those capabilities will define who thrives in the post-bubble landscape.
For executives, the real question isn’t whether the AI bubble will burst. It’s whether you’ll be one of the organizations positioned to take advantage when it does.
Waiting for perfect clarity is the riskiest move you can make. The companies that lean in now—pragmatically, strategically, and with ROI in mind—will be the ones defining what comes next. Everyone else will be playing catch-up.
Sources: