Is AI the Next Dot-Com Bubble?
Why today’s hype looks a lot like 1999 and how to spot the 1 percent of companies that will survive
Déjà Vu: When “.com” Became “.ai”
In 1999, merely having a domain name ending in .com was enough to secure millions in funding for a startup. Profits were secondary, revenue a bonus. Remember Pets.com, Webvan, and eToys? They all vanished when the market corrected in 2000.
Fast-forward twenty-five years and substitute "internet" with "AI." Pre-revenue firms are now commanding billion-dollar valuations. Character.ai, for instance, raised $150 million at a $1 billion valuation without a single sale. Public markets reflect this fervour, with unprofitable "AI plays" trading at 20-to-60 times revenue levels last seen during the dot-com mania.
Even Robin Li, Baidu’s CEO, describes this period as "1999 all over again." A similar fear-of-missing-out (FOMO) is compelling investors to funnel cash into anything with an "AI" label. Tools are being released faster than users can adopt them, and venture capital spreadsheets project infinite growth and profitability, eventually.
The parallels in hype are undeniable. However, will the ensuing crash be identical? That hinges on one critical factor: demonstrable value.
The Harsh Reality Behind the Buzz
Most projects don’t deliver
IBM’s survey of 2,000 CEOs revealed that only 25 percent of AI initiatives met their ROI targets, and a mere 16 percent scaled beyond a pilot phase. In short: three out of four projects fail.
Most startups won’t survive
Li predicts that 99 per cent of AI companies will disappear within 18 months. Why? Because novelty is not a sustainable business model. A flood of chatbots, code copilots, and image generators launched in 2023, but how many do you still use daily? Consumers are already indifferent to "AI-PC" stickers that no one requested.
Even winners burn cash
OpenAI’s ChatGPT reached 100 million users in record time, yet insiders suggest it could burn $8 billion this year simply to keep its GPUs operational. Popularity does not equate to profit; at some point, margins must materialise.
When resources inevitably tighten, companies that offer only a "cool demo" will be the first to falter. As Warren Buffett famously remarked after the dot-com bust: "When the tide goes out, you discover who’s been swimming naked."
Who Lives, Who Dies: ROI > Novelty
History’s survivors; Amazon, Google, eBay all share two fundamental traits:
They solved a painful, expensive problem (shopping, search, auctions).
They created a sustainable business model (real revenue > real costs).
The Arms Race of FOMO
If logic suggests most firms will fail, why is capital pouring in?
Big Tech is terrified of missing the next Google. Meta plans up to $72 billion in AI capital expenditures this year (2025). Microsoft, Google, and Amazon are engaged in a fierce GPU buying war.
Venture funds perceive limited exit opportunities elsewhere; AI seems inevitable, prompting them to inflate valuations for anything with a model.
Corporate boards fear irrelevance; 64 per cent of CEOs cite FOMO as their primary driver for AI adoption.
Similar to the gold rush of 1849, the safest profits are not found in digging for gold, but in selling the shovels. Nvidia, which produces most of the world’s AI chips, has added over $2 trillion in market capitalisation in two years, echoing Cisco’s router boom in 1999. Meanwhile, many prospectors will discover only dust, not gold.
A Pragmatic Playbook: How Not to Get Burned
Whether you are an investor, builder, or buyer, ask three critical questions before committing:
If we removed the letters “AI,” would anyone care? Hype alone cannot be the sole appeal.
Does it replace or improve an existing process far better than the status quo? It must be faster, cheaper, or enable something previously impossible.
Can we prove tangible ROI within one budget cycle? If payback is vague or exceeds 18 months, anticipate CFO resistance.
For Investors
Focus on infrastructure, vertical SaaS with a regulatory moat, or horizontal tools tied to indispensable workflows (e.g., data labelling, MLOps).
Demand a path to gross margins >60 per cent. GPU expenses are unforgiving.
For Builders
Lead with the problem, not the model. Customers purchase outcomes, not neural networks.
Start with a single high-value use case; expand only after demonstrating proven traction.
For Enterprise Buyers
Pilot quickly and measure results ruthlessly. Discontinue what fails to meet cost-saving or revenue-lift targets.
Beware of vendor lock-in to proprietary models unless the ROI gap is substantial.
Remember: The metrics that matter are traction, real users, and real revenue. Everything else is noise.
After the Bubble (Because It Will Pop)
When dot-com optimists resurfaced in 2002–2003, they found a transformed landscape:
The pretenders had vanished.
The infrastructure (affordable bandwidth, web standards) had matured.
Survivors like Amazon had captured significant market share at bargain prices.
Expect an identical pattern for AI:
Mass extinction of copycats when capital dries up.
Consolidation around platforms that provide indispensable utilities.
Second-wave innovation built on cheaper models, stable APIs, and clearer regulations.
In essence, AI will undoubtedly change the world, but through a handful of enduring winners, not the hundreds currently vying for headlines.
Conclusion: Bubble? Yes. Opportunity? Even Bigger.
Is AI today a bubble? In the investment sense, almost certainly. Valuations outpace fundamentals, FOMO fuels irrational spending, and 99 per cent of startups will disappear.
However, bubbles are not solely destructive; they build essential infrastructure for future progress. The dot-com bust laid fibre-optic cables and gave rise to cloud giants. The AI boom is laying the groundwork for compute clusters, machine-learning frameworks, and an AI-first culture poised for automation.
When the froth dissipates, companies that solve real problems and deliver provable ROI will emerge stronger than ever. Your task is to distinguish the enduring wave from the transient foam.
So, the next time an “AI-powered” demo appears in your feed, ask:
Would I pay for this if it didn’t say AI?
If the answer is no, keep scrolling. If yes, you may have discovered one of the 1 per cent.
Seen an AI tool that’s all sizzle and no steak? Share it in the comments. Let’s collaboratively identify which ideas will sink or swim before the bubble decides for us.
