Bubbles have an interesting characteristic: countdowns to their collapse usually begin before they've even finished inflating. Perhaps it’s residual fear from recent history. It wasn’t too long ago that the dot-com bubble ballooned and crashed spectacularly, leaving thousands of companies flattened in its wake.
The Gartner Hype Cycle is a graphical model that estimates new technology’s maturity, adoption, and social application over time, with each stage having a memorable name: Technology Trigger, Peak of Inflated Expectations, Trough of Disillusionment, Slope of Enlightenment, and Plateau of Productivity. According to the 2025 Hype Cycle, generative artificial intelligence was in the Trough of Disillusionment, and AI agents were at the Peak of Inflated Expectations. This is a warning that the tech is already losing its shine while the next wave of promises is still cresting.
We can draw parallels between the ongoing AI bubble and the dot-com era, with both meeting bubble metrics like hype-driven valuations, investment per employee, and market concentration. History suggests that a market correction will follow, but there are just enough differences for optimists to claim that the situation is similar, but not the same. Unlike the internet boom of the late 90s, AI growth is backed by well-known, profitable entities with solid earnings; inflation is stubbornly higher; and most importantly, AI is already integrated and yielding productivity gains in several industries.
Ironically, this comprehensive integration is also what makes the risk of the bubble popping too consequential to ignore. The numbers are overwhelming; Big Tech is projected to spend USD 650 billion on AI infrastructure this year. However, the resource-hungry nature of AI means pouring immense capital into it alone does not ensure success or sustainability. The planned infrastructures are already straining labour, chip production, and energy supplies. While data centres have started their expansion, their popularity proportionally has started going downhill. The centres are famous for their massive water consumption, and nobody wants their neighbourhood drained and electricity bills surging. Concerns persist about whether the actual returns on these massive investments will justify the expense. Capital expenditure outpacing revenue generation is precisely when bubbles come to form.
The argument for a bubble existing becomes harder to deny when OpenAI CEO Sam Altman himself suggested that there is one. When the person most incentivised to deny the existence of a bubble acknowledges it, it is a signal worth analysing. For peripheral economies, these signals should trigger alarm bells.
Bangladesh is still in the lower end of AI readiness, as tech literacy remains restricted to urban and digitally privileged groups. This won’t insulate it from the fallout, though. Countries at this stage won’t feel the sharp loss of overvalued assets, but they will definitely feel the foundations of their digital economic growth eroding. Institutions which have been routing increasing volumes of development finance toward digital infrastructure will shrink their commitments when the global appetite for “digital transformation” cools down.
The most severe threat will be indirect but also quite predictable. In March, it was reported that Bangladesh’s merchandise exports fell for the seventh consecutive month in February. Further impediments posed by global oil shocks will cause significant hits. Most consumers will reduce their discretionary spending on apparel in a tech-driven recession in the West. With the Ready-Made Garment (RMG) sector accounting for most of the nation’s export earnings, it could be a sign of a tumultuous period.
Trade bearing the brunt of the impact has historical precedent, as can be demonstrated by how Bangladesh’s economy reacted to the worldwide financial crisis in 2008. The more direct victims of an AI reckoning would be the IT sector, particularly the freelance workforce and computer science majors who are getting priced out of their own field.
It's worth noting that bubbles don’t always “pop”; sometimes they “deflate”. Big Tech monopolies have funded most of the AI infrastructure until now, not banks, making a broad financial collapse improbable. AI is also increasingly becoming a national security asset and reinventing modern warfare to the point where governments have incentives to subsidise it. In the deflation scenario, a recession will simply be a controllable market correction. The difference between a pop and a deflation is largely a matter of speed and severity, not of immunity. The uncertainty itself remains the pressing policy problem.
References:
1. CNBC (August 18, 2025). OpenAI’s Sam Altman sees AI bubble forming as industry spending surges.
2. Reuters (February 23, 2026). Big Tech to invest about $650 billion in AI in 2026, Bridgewater says.
3. The Daily Star (March 3, 2026). Exports drop for 7th straight month on garment slum.
