By Joseph Moss, International Banker
Artificial intelligence (AI) has, inarguably, become the defining investment theme of the decade. In less than three years, it has propelled the world’s largest technology companies to record valuations, triggered an unprecedented wave of spending on data centres and semiconductors, and transported corporate boardroom discussions from Silicon Valley to Wall Street. With speculation now rife over whether the supposed AI bubble will soon burst, investors are increasingly wondering whether the unprecedented scale of capital flowing into chips, data centres, power infrastructure, cloud computing and more will generate economic returns quickly enough to justify today’s valuations.
Unlike previous technology booms, today’s enthusiasm is built upon genuine commercial foundations. The companies driving the AI revolution are among the world’s most profitable businesses, generating hundreds of billions of dollars in annual cash flows. Demand for computing power continues to outstrip supply; enterprise adoption is accelerating; and AI is increasingly being deployed across industries ranging from healthcare and financial services to manufacturing and energy.
The forward outlooks are perhaps even more eye-watering. Goldman Sachs predicted that the AI infrastructure buildout alone could attract “roughly $7.6 trillion of cumulative capital expenditure between 2026 and 2031”, spanning computing, data centres and power generation. JPMorgan Chase, meanwhile, projected that total global AI-related capital expenditure (CapEx) could ultimately reach $5.5 trillion through the end of the decade.
The debate is no longer centred on whether AI will transform the global economy—few doubt that it will. Instead, investors are now asking whether financial markets have become overly optimistic about the speed, scale and profitability of that transformation. Numerous historical examples exist of genuinely revolutionary technologies that, while ultimately fulfilling their promise, also generated painful market corrections as capital deployment raced ahead of commercial returns. The dot-com bubble at the turn of the millennium, for instance, saw many investors who had financed the internet’s rapid expansion ultimately suffer substantial losses during the ensuing market collapse.
AI could conceivably follow a similar path. While the technology itself appears increasingly indispensable, considerable uncertainty persists about whether today’s extraordinary levels of investment—and the valuations built on them—will be justified by future revenues.
That said, the internet boom of the late 1990s was characterised by widespread speculation in companies with little proven revenue, uncertain business models and minimal profitability. Many firms depended almost entirely on investor enthusiasm rather than sustainable earnings.
Today’s AI leaders present a markedly different picture. Microsoft, Alphabet, Amazon, Meta and Nvidia collectively generate enormous operating cash flows that allow them to finance much of their AI investment internally. Rather than relying on speculative fundraising, these tech behemoths are deploying profits from well-established businesses into what they regard as the next phase of technological development.
This crucial distinction has led many institutions to argue that the current investment cycle rests on considerably firmer foundations than past tech bubbles did. JPMorgan Chase, for example, believing that the economics of the AI buildout remain broadly supportive, recently increased its estimate of global AI-related capital expenditure to $5.5 trillion through 2030, reflecting continued expansion in data centres, semiconductor production and electricity infrastructure. Importantly, it projected that the hyperscalers driving much of this investment would remain highly profitable, with operating cash flows exceeding $900 billion by 2027. From this perspective, rising investment represents a rational response to extraordinary demand rather than speculative excess.
Morgan Stanley similarly argued that AI “is no longer a tech story—it is a macro variable influencing GDP, earnings, credit markets and geopolitics at industrial scale”. The bank forecast that nearly $3 trillion would be invested globally in AI-related infrastructure by 2028, with more than 80 percent of that spending still ahead.
Goldman Sachs, meanwhile, sees AI’s next phase unfolding across the broader economy rather than remaining concentrated among technology companies, with software currently representing less than 0.5 percent of global gross domestic product (GDP). “The real economy—the other ~99.5 percent of the global economy that AI has barely touched, from manufacturing and robotics to defence, construction, and energy—defines the actual scale of opportunity,” according to the bank’s “Harnessing AI for the Real Economy” report.
This helps explain why investors continue to reward companies supplying the infrastructure needed to support AI adoption. Demand for advanced semiconductors, cloud-computing capacity, electricity generation and data-centre construction remains exceptionally strong. Far from being driven solely by speculative enthusiasm, much of today’s investment reflects genuine expectations that AI will become embedded throughout the global economy over the coming decade.
Nevertheless, the strongest arguments supporting today’s AI-investment boom increasingly coexist alongside equally persuasive reasons for caution. The principal concern is less that AI lacks commercial value and more that investments are expanding much faster than the revenues being generated.
Recent months have seen technology companies announce successive increases in capital expenditure, committing hundreds of billions of dollars to expand computing capacity, develop increasingly sophisticated AI models and construct vast networks of data centres. Competition among hyperscalers has increasingly intensified, spurring even greater spending among incumbents. The result has been one of the largest investment cycles in modern corporate history.
While the Bank for International Settlements (BIS) recently observed that the resilience of the global economy has depended heavily on accommodative financial conditions and an investment boom driven by expectations surrounding AI, it also warned that these favourable conditions are becoming increasingly fragile. “The resilience that characterised the global economy through the review period rested on a specific set of conditions—accommodative financial conditions, synchronised monetary easing and an investment boom sustained by expected transformative technological progress from AI,” according to the BIS’ “Annual Economic Report” published on June 28.
“Each is coming under strain,” the BIS also warned, citing four pressure points that stand out in particular: risks of persistent inflation, the sustainability of AI-related investment, growing financial vulnerabilities and weakening fiscal positions.
Markets are gradually becoming less interested in the scale of AI spending itself and increasingly focused on whether those investments can generate sufficiently attractive returns.
Markets are gradually becoming less interested in the scale of AI spending itself and increasingly focused on whether those investments can generate sufficiently attractive returns. Morgan Stanley noted that while around one-fifth of S&P 500 companies have reported measurable benefits from AI adoption, investors have become far more selective, rewarding firms that demonstrate tangible productivity gains rather than simply announcing ambitious AI strategies.
The distinction between investment and monetisation has therefore become critical. Artificial intelligence no longer needs to prove its technological potential. Instead, investors are asking more difficult questions: How quickly will AI generate sustainable revenues? Can productivity improvements justify unprecedented capital expenditure? Will businesses ultimately earn attractive returns on the infrastructure currently being built?
Even investors who are generally optimistic about AI acknowledge that expectations have become increasingly demanding. Goldman Sachs, for one, expects markets to move beyond rewarding AI exposure alone and to increasingly favour businesses capable of demonstrating measurable economic returns from AI deployment.
The AI boom is also increasingly being financed through a complex ecosystem that extends well beyond the balance sheets of the hyperscalers themselves. Banks, private-credit funds, infrastructure investors, pension funds and insurance companies are all providing capital to support the construction of data centres, electricity infrastructure and semiconductor production. In many respects, AI has evolved from a technology investment theme into a broad capital-markets phenomenon, and as such, the greatest risks lie in the financial architecture that is developing AI rather than the technology itself.
Should one or more of the major hyperscalers significantly reduce capital expenditure, the effects would extend far beyond the tech sector. Demand for semiconductors, electricity generation, specialised cooling systems, construction services and data-centre capacity could weaken simultaneously, impacting a wide range of lenders and investors whose exposure to AI may be less obvious.
The BIS has also warned that intense competition among technology firms risks encouraging companies to commit excessive resources to projects with long-term returns that remain uncertain. Should those returns ultimately disappoint, financing conditions could tighten rapidly, potentially transforming today’s investment boom into a prolonged period of adjustment.
Man Group reached a similar conclusion while drawing an important distinction between the technology itself and the financial structures supporting it. “The technology is real. Applications will emerge. The companies solving fundamental cost and distribution problems will thrive,” the UK investment manager wrote in February. “But we believe that the financial architecture built on overleveraged bets, short-duration assets financed with long-duration debt, and circular demand signals is unsustainable.”
Ultimately, compelling evidence supports both sides of the debate. Those who warn of an AI bubble point to rapidly rising valuations, unprecedented capital expenditure, increasingly sophisticated financing structures and growing dependence on exceptionally optimistic assumptions about future revenues. At the same time, the companies leading AI development generate substantial earnings, possess exceptionally strong balance sheets and continue to report robust demand from enterprise customers.
Should businesses successfully translate today’s enormous investments into sustained productivity gains and durable earnings growth, current valuations may ultimately prove justified. Should monetisation develop more slowly than expected, however, investors may conclude that capital has once again outpaced commercial reality.