The dot-com bubble of the late 1990s wiped out trillions of dollars in market value and left a generation of technologists scarred by the experience. Companies with no revenue, no customers, and no clear path to profitability were valued at billions of dollars — until they weren't. The crash was swift, brutal, and in retrospect, inevitable.
Now, with AI companies raising record rounds, incumbents pivoting their entire strategies around artificial intelligence, and the word "AI" appearing in nearly every earnings call regardless of relevance, the comparisons are inevitable. Is the AI boom the next bubble? Will it pop — and if so, how hard?
The Parallels Are Real
The similarities between the dot-com era and today's AI frenzy are hard to ignore. In the late 1990s, investors poured money into internet companies on the premise that being online was itself a business model. The technology was real — the internet did change everything — but the valuations were completely detached from economic fundamentals.
Today, a similar dynamic is playing out. Companies are raising massive rounds not because they've demonstrated sustainable revenue, but because they're building on AI. "Powered by AI" has become the new ".com" — a suffix that unlocks capital regardless of what's actually being built. The underlying technology is genuinely powerful, but many of the companies being funded are yet to demonstrate they can turn that power into durable, defensible businesses.
But the Differences Matter
The dot-com bubble was characterized by companies selling to consumers who hadn't yet adopted the internet at scale. Infrastructure barely existed. E-commerce required convincing people to enter credit card numbers into a website, which many found terrifying. The market had to be built from scratch alongside the companies trying to monetize it.
AI is different in a critical way: the demand is already there. Enterprises are actively buying AI products today. OpenAI reportedly hit $1 billion in annual revenue faster than almost any software company in history. The spending isn't speculative — it's a response to genuine productivity gains that organizations are documenting in real time. When the underlying product genuinely works and organizations are paying for it, that's not a bubble dynamic.
The infrastructure layer is also far more mature. Unlike the internet era, where building a website meant negotiating arcane protocols and unreliable hosting, AI developers today can access world-class compute and model infrastructure via APIs. The barriers to actually shipping have never been lower.
Warning Signs to Watch
That said, not all AI investment is created equal. The warning signs worth monitoring are the same ones that preceded the dot-com collapse: excessive valuations for companies with no clear path to profitability, copycat products racing to capture a market that may not be large enough for all of them, and a narrative that treats AI adoption as inevitable without accounting for implementation complexity.
Compute costs remain a significant challenge. Training and running large models is expensive, and gross margins in AI products are often much thinner than traditional software. Companies that are discounting aggressively to capture market share now may face painful unit economics reckoning later. The companies that survive will be those who can demonstrate genuine customer retention and sustainable margins — not just impressive benchmark scores.
Reasons for Optimism
History rarely repeats exactly. The dot-com era crashed because the technology was real but the market wasn't ready. AI is in an inverse situation: the market is ready, enterprises are buying, and the technology is improving at a pace that keeps expanding what's possible. Even a significant correction in valuations wouldn't necessarily mean the technology stops being useful or the industry stops growing.
The likely scenario isn't a 2000-style implosion but a more gradual consolidation. The hundreds of AI companies currently funded will compress into a smaller number of category leaders. Some layers of the stack — particularly foundation model development — will consolidate significantly as compute costs make it impossible for all but the best-capitalized players to compete. The companies with the clearest paths to genuine value creation will survive; the ones riding purely on narrative will not.
The Companies Most at Risk
Not all AI companies face equal exposure. The segment most at risk is the middle layer: companies that are neither building foundational models nor delivering a uniquely differentiated end-user experience, but simply wrapping existing APIs with a thin product surface. These companies are vulnerable from two directions simultaneously — from the foundation model providers who keep expanding their own product offerings, and from competitors who can replicate their product at minimal marginal cost.
The companies best positioned for long-term survival share a few common characteristics: they have proprietary data that competitors cannot easily acquire, they operate in a domain where AI needs to be deeply specialized and their team has genuine domain expertise, or they've built a workflow tool so embedded in their customers' operations that switching costs are prohibitively high. Pure prompt engineering on top of a generic model is not a competitive moat. Distribution, proprietary data, and deep integration are.
Enterprise buyers are beginning to understand this distinction too. Early AI spending was exploratory — companies bought AI tools to figure out what AI could do for them. The next phase of enterprise purchasing will be more disciplined, with procurement teams scrutinizing ROI and demanding demonstrable, measurable productivity gains. That shift in buyer behavior is itself a natural filter that will separate genuinely useful AI products from the ones that sold primarily on excitement.
What a Correction Would Actually Look Like
If a significant correction does occur, it's worth thinking clearly about what form it would take. The dot-com collapse was sudden and catastrophic — driven by a specific trigger that cascaded across the entire internet sector at once. The AI landscape today is more structurally complex, and any correction would likely be more gradual and uneven across different segments of the market.
The most plausible scenario is a wave of quiet failures over the next two to three years. Startups that raised at peak valuations in 2023 and 2024 will find their next funding rounds unavailable at equivalent multiples. Without the ability to raise on narrative alone, they'll face pressure to demonstrate real revenue and sustainable gross margins — something many currently lack. Some will find genuine product-market fit and survive. Many will not, and will either shut down or be acqui-hired for their engineering talent rather than their product.
Public market sentiment could act as an accelerant. If a major AI company misses earnings expectations substantially or reveals that customer retention is weaker than anticipated, the broader narrative around AI monetization could shift quickly. Markets have a tendency to overcorrect in both directions, and the same momentum that inflated AI valuations could deflate them rapidly if confidence breaks.
What this would emphatically not mean is that AI stops being important or transformative. The internet didn't become less significant after the dot-com crash. Amazon, Google, and PayPal — all survivors of that era — went on to become among the most valuable companies in history. The AI survivors of any near-term correction will likely follow a similar trajectory over the decade to come.
The Verdict
Will the AI bubble pop? Probably — in the sense that some portion of the current valuation landscape is disconnected from economic reality and will correct. Dozens of well-funded AI startups will fail to find product-market fit and will quietly shut down or get acqui-hired in the next few years.
But will AI itself fail? Absolutely not. The technology is too useful, too embedded in enterprise workflows, and too rapidly improving to suffer the fate of pets.com. What we're likely to see is not a crash but a culling — a market that gets more rational as it matures, rewards genuine value creation, and punishes hype-driven excess. That's not a bubble popping. That's an industry growing up.
For consumers and businesses evaluating AI tools right now, the practical implication is straightforward: focus on the products that solve a real problem for you today, from companies that have a credible path to still being around in five years. The technology works. Be selective about which bets on it you're making.
Sheldon writes about AI strategy, emerging technology, and the business dynamics shaping the software industry. He founded Dear Tech to provide honest, consumer-first analysis in a space dominated by hype. He has been following the AI industry since the early transformer era and writes from the perspective of someone who uses these tools every day.