Cast your mind back to 1999. Pets.com. Webvan. Every company with a dot-com suffix and a vague internet strategy was worth hundreds of millions. Venture capitalists were millionaires on paper by Thursday and richer by Monday. Then March 2000 arrived.
Now look at 2026. AI startups with eighteen months of operating history and minimal revenue are raising rounds at valuations that would make a 1999 venture capitalist blush. The infrastructure bills are astronomical.
Is this different this time? Or are we watching the same movie with different actors?
The Numbers That Should Make You Pause
Global AI investment exceeded $200 billion in 2024 according to multiple industry tracking reports. The hyperscalers — Microsoft, Google, Amazon, Meta — announced over $300 billion combined in AI infrastructure spending for 2025 alone, based on their own public earnings disclosures. Data centre construction, chip orders, and talent acquisition are consuming capital at a pace that has no historical precedent in the technology industry.
At the startup level, valuation multiples are striking. Companies with tens of millions in annual recurring revenue are raising at valuations that imply extraordinary growth sustained for years — multiples that are historically justified only for the fastest-growing companies in the world, and only when growth is both proven and sustainable.
The revenue picture for many AI companies is also more complicated than the headlines suggest. Infrastructure costs are extremely high. The cost per query has not yet fallen to levels that make the business model clearly sustainable at current pricing for all players.
What the Experts Are Actually Saying
MIT economist Daron Acemoglu published "The Simple Macroeconomics of AI" through the National Bureau of Economic Research in 2024. His analysis argued that AI's actual contribution to measured total factor productivity growth may be more modest than investment levels suggest — and that the tasks most amenable to AI automation represent a smaller share of overall economic activity than commonly assumed. (Source: Daron Acemoglu, NBER Working Paper 32487, April 2024)
The counterargument, made by investors and AI companies, is that AI is already embedded in daily life in ways that the internet was not in 1999. ChatGPT has nearly 900 million weekly active users as of February 2026. AI is in your phone's autocorrect, your spam filter, your navigation app, your music recommendations. This is not a technology looking for a use case. It is a technology billions of people use daily without always realising it. (Source: OpenAI announcement, February 27, 2026)
Both arguments have merit. The hype around specific AI companies may be overextended. The underlying technology, and its genuine utility, is not.
"There can be a bubble in the stocks without there being a bubble in the technology itself. The dotcom crash wiped out Pets.com. It did not wipe out the internet."
What Would Bursting Actually Look Like?
A realistic AI market correction would look like this: a wave of startup failures as funding dries up for companies that never found sustainable unit economics. Consolidation as the strongest players acquire the weakest at distressed valuations. Layoffs in the AI sector. A period where the funding environment becomes dramatically more conservative.
What it would not look like: ChatGPT shutting down. AI features disappearing from your iPhone. Google removing AI from Search. These products are too embedded, too valuable, and backed by companies with too much financial resilience to be casualties of a funding correction.
A correction in AI investment markets would hurt venture-backed AI startups significantly. It would barely register for the hundreds of millions of people who use AI tools embedded in major platforms. The technology survives regardless.
Why This Might Be Different From Dotcom
In 1999, the internet was genuinely transformative but not yet genuinely useful to most people in daily life. Broadband was not yet widespread. The infrastructure needed to deliver on the promise was still being built. The stock prices were pricing in a future that was real but a decade away.
AI in 2026 is already useful to hundreds of millions of people today. The infrastructure — cloud computing, mobile internet, data storage — was already built before AI arrived. Unlike the internet in 1999, AI does not require a decade of infrastructure build-out before the value is real. That is a meaningful difference.
It does not mean there is no bubble risk in specific companies at specific valuations. It does mean that the underlying technology and its utility are not speculative in the way that dot-com was speculative.
What You Should Actually Do
For businesses building on AI: prefer integrating with AI capabilities from deeply embedded major platforms — OpenAI's API, Google's Gemini, Anthropic's Claude — over building your core business on top of a single volatile startup whose survival depends on its next funding round.
For individual users: nothing changes. The tools you use daily are not going anywhere. Keep building skills and workflows around AI tools from companies with clear, durable business models.
A correction, if it comes, will hurt investors and AI startup employees. It will barely touch the person using AI to run their business, study for their exams, or draft their emails. Keep building. Keep learning. The underlying shift is real regardless of what the stock market does.
