Calling AI a bubble can conceal two different questions. Are some companies, projects, and assets priced far beyond the returns they will produce? And is the underlying technology destined to disappear? The dot-com era shows that the first answer can be yes while the second remains emphatically no.
A BofA Global Research survey found 54 percent of fund managers believe AI stocks are already in bubble territory, while 38 percent disagree. Valuations, multi-billion-dollar infrastructure programmes, and heavy chip spending have all moved faster than proven revenue in parts of the market.
Yet investors who recognise overheating have not necessarily left. UBS strategists found about 90 percent of investors who see a bubble still hold AI-related assets. That tension captures the market: participants fear a correction but also fear missing a technological shift that could reshape industries for decades.
The Dot-Com Comparison Cuts Both Ways
Cisco Asia-Pacific president Ben Dawson compares the current cycle with the internet’s early years: excitement attracts capital, weak ideas multiply, a correction arrives, and the surviving technology continues changing society. Some AI companies and business models may fail without invalidating the broader transformation.
The buildout also resembles the fibre boom. Venture capital and sovereign wealth funds are financing capacity before demand is fully established. If adoption develops slowly, data centres or other assets could be stranded. If applications mature quickly, today’s spending may look like necessary preparation.
Cisco infrastructure executive Simon Miceli views the expansion as the industrialisation of AI and asks whether the world is building quickly enough. Goldman Sachs economist Joseph Briggs likewise considers infrastructure spending economically sustainable, while acknowledging that the winners are hard to identify when customers can switch providers and technology changes rapidly.
Policy and Capital Are Amplifying the Cycle
Governments treat AI as economic capacity and national security, not merely another software market. US administrations have encouraged speed through policy and investment signals. China directs capital toward domestic companies to reduce reliance on American technology. Europe has paired regulation with initiatives such as the AI Continent Action Plan and a €1 billion Apply AI fund.
Public involvement can accelerate innovation, but it can also intensify competition before commercial demand is clear. Supply constraints add another variable. ABB chief executive Morten Wierod does not see a bubble, although construction and supply-chain limitations could slow new data-centre deployment.
The Bank of England has warned that a loss of confidence could produce a sharp market correction with a material impact on the UK financial system. IMF chief economist Pierre-Olivier Gourinchas offers a counterweight: because AI investment is not primarily debt-financed, a downturn may be less likely to become a systemic financial crisis.
Businesses Need Value, Not a Market Forecast
For an operating company, correctly timing the investment cycle is less important than connecting AI to a real problem. The internet companies that survived the dot-com collapse did not endure because every valuation was rational; they built services customers continued to use after the enthusiasm cooled.
Early-stage AI is particularly exposed. Singaporean sovereign wealth fund GIC’s chief investment officer Bryan Yeo has warned of startups receiving enormous multiples despite modest revenue. OpenAI chief executive Sam Altman also expects some investors to lose large sums while others make substantial gains.
A correction would force a useful distinction between owning AI exposure and creating AI value. Businesses should ask whether a deployment improves revenue, cost, speed, quality, or resilience—and whether that benefit survives the full expense of data, infrastructure, oversight, and integration.
The most plausible outcome is not a choice between “bubble” and “revolution.” Financial excess can burst while useful models, chips, networks, and applications remain. The central challenge is to avoid confusing a lasting technological direction with a guarantee that every company travelling in that direction deserves its current price.
