The current wave of investment in artificial intelligence is increasingly being compared to the dot-com boom of the late 1990s and the early e-commerce build-out of the 2000s. The comparisons are useful —but only partly. The internet bubble was driven by speculative capital chasing weak business models. E-commerce, by contrast, involved years of losses to build consumer habit, logistics networks and trust before profitability arrived. AI today appears to sit somewhere in between: real demand, real utility, but funded by unusually strong balance sheets and massive free cash-flow pools.
Resembles early e-commerce rather than dot-com bubble
The dot-com boom broadly lasted from 1995 to 2000, when internet valuations detached sharply from business fundamentals. The Nasdaq peaked in March 2000 and fell nearly 78 per cent by 2002. Yet, while valuations collapsed, the internet itself did not. It simply needed time for viable business models to emerge.
That second phase was represented by e-commerce. Amazon was founded in 1994, listed in 1997, and took nearly nine years to post a full-year profit. The company spent most of that time investing in warehouses, fulfilment centres, delivery systems, customer acquisition and pricing. Consumers had to learn a new habit: trusting online payments, accepting delayed delivery and shifting spending habits online. That is the closest parallel to AI today.
Consumers and enterprises are currently forming a new behavioural layer around AI: search through conversation, software copilots, AI-assisted coding, automated workflows, AI agents and productivity tools. Habit formation takes time. E-commerce needed nearly a decade to become mainstream; AI may require a similar runway before monetisation catches up with investment.
AI is being funded by cash-rich incumbents
The biggest difference between the AI wave and the dot-com era lies in the source of funding for the infrastructure.
In the late 1990s, many speculative companies relied on external capital markets. Today, the dominant AI spenders are highly profitable platform companies with large internal cash generation.
Alphabet generated roughly $165 billion of operating cash flow in 2025, held about $127 billion of cash and securities, and still produced more than $70 billion of free cash flow despite elevated capex.
Meta Platforms generated nearly $116 billion of operating cash flow, held more than $80 billion in cash and investments, and produced over $40 billion of free cash flow.
Amazon generated about $140 billion of operating cash flow and held roughly $123 billion of liquidity, although free cash flow compressed sharply because of accelerated infrastructure spending.
This matters enormously. If AI had to depend primarily on venture funding or debt issuance, the cycle would likely end once rates rose or sentiment weakened.
Instead, the current build-out is being funded by companies that built enormous cash engines through search advertising, cloud computing, digital commerce and social media. In simple terms, the excess returns of the platform era are being recycled into AI capex. So, AI can remain economically irrational longer than many sceptics expect — because its sponsors can afford patience.
When will markets know whether AI is a bubble or not?
The key question is not whether AI is real — it clearly is — but whether the current spending earns acceptable returns. The answer: It is unlikely in the next 12 months. It will emerge over a three- to seven-year window.
In the next 2–3 years, markets will tolerate heavy spending if usage growth remains strong. AI assistants, cloud inference demand, coding copilots and enterprise adoption can sustain optimism. Smaller AI startups with weak differentiation may fail first, but hyperscaler capex can continue.
In years three to five, investors will demand monetisation. They will ask whether AI improves search revenue, cloud margins, enterprise software pricing power, ad conversion rates and labour productivity. If revenues lag depreciation and power cost, scepticism will shoot up.
In years five to seven, either AI becomes embedded infrastructure — similar to cloud computing and e-commerce logistics — or the industry faces a major capex reset.
In conclusion, AI today is best understood not as a classic bubble, but a cash-funded habit-creation cycle. Like e-commerce, it may require years of upfront losses or low returns before user behaviour permanently changes. Like in the dot-com era, valuations can overshoot and many players may disappear. But unlike in the past speculative booms, today’s biggest spenders are not fragile startups — they are some of the most cash-generative businesses in corporate history.
That distinction may allow the AI cycle to run longer than markets expect, perhaps for five to seven years, before the final winners and losers are known.
(Karan Taurani is EVP, Elara Capital)
More Like This
Published on April 27, 2026
