The artificial-intelligence boom is entering a more difficult phase. After years in which surging demand for chips, data centres and cloud services helped power markets, investors are now confronting a less forgiving question: how much capital can the industry absorb before expected returns arrive?
Recent developments point to a collision between three forces. AI investment remains a major source of economic growth, but borrowing costs are rising, regulators are probing the industry and financial markets have begun punishing companies most exposed to the data-centre buildout. The result is not necessarily the end of the boom. It is a transition from enthusiasm based largely on forecasts to scrutiny based on cash flow, productivity and accountability.
Growth has depended on extraordinary investment
Technology spending has become an important pillar of the US economy. US real GDP expanded at a 3.4% annualised rate in the third quarter of 2026, with growth supported by technology investment and consumer spending, according to Deloitte. Real consumer spending rose at a 3.8% annualised rate, while services spending increased 4.1%.
That resilience has helped sustain the AI trade. Semiconductor companies, cloud providers and data-centre developers have benefited from expectations that businesses will continue buying computing capacity. But the same figures also sharpen the risk for monetary policymakers: stronger demand can prolong inflation and delay interest-rate cuts. Markets recently placed an 81.1% probability on a US Federal Reserve rate increase in December, up from 66.9% the previous day, Deloitte reported.
Higher rates matter because AI infrastructure is expensive long before it produces measurable profits. Data centres require land, electricity, advanced chips and long-term financing. A company can therefore report rapidly rising demand while still facing years of heavy capital expenditure. When bond yields rise, the value investors assign to distant earnings falls.
Markets are questioning the infrastructure model
The clearest warning has come from the market itself. The Financial Times reported that data-centre construction stocks lost $4.2 trillion in value as the AI trade unravelled. Such a fall does not prove that AI demand is disappearing, but it shows that investors are reassessing the assumptions embedded in valuations.
One interpretation is that the correction is healthy. Supporters of continued investment argue that current spending is comparable to earlier infrastructure revolutions: the returns from electricity, telecommunications and the internet also took time to emerge. AI could reduce administrative costs, accelerate research and raise productivity across sectors that have barely begun deploying it.
The opposing view is that the industry has developed a circular financial dynamic. Chipmakers sell equipment to companies building AI capacity; cloud providers then need customers to use that capacity; investors value the providers on the assumption that future demand will justify present spending. If corporate customers delay adoption or discover that productivity gains are smaller than advertised, the weakest projects could become stranded assets.
Politics is moving from promotion to oversight
Government attitudes are shifting as the sector becomes more economically and strategically important. In Washington, Republican lawmakers are debating how far oversight should extend across chipmakers and the wider AI industry after a scandal involving Nvidia, according to Politico. Former Nvidia chief Jensen Huang has agreed to testify before House lawmakers, while legislators have asked intelligence agencies to examine whether his relationship with Alice Huang created national-security concerns, Politico reported separately.
The political argument is contested. Industry advocates warn that aggressive regulation could weaken domestic firms in competition with China, slow innovation and push investment offshore. Critics counter that national-security risks, market concentration and opaque corporate relationships make hands-off policy untenable. Europe is facing a related dilemma: lawmakers have called for action to reduce dependence on a small number of foreign technology companies, according to Politico Europe.
These debates are likely to expand beyond safety rules. Governments will also decide who pays for electricity-grid upgrades, whether tax incentives should support data centres and how much public infrastructure can be committed to private AI projects. Those choices will determine whether the boom produces broad economic gains or mainly enriches a narrow group of firms and investors.
What comes next
The next phase will be defined by evidence. Investors will look for revenue from paying customers rather than demonstrations, while policymakers will demand clearer information about energy use, data governance and national-security exposure. Companies with strong balance sheets and identifiable productivity gains may continue expanding; projects dependent on cheap financing and perpetual demand may be postponed or cancelled.
The broader economy faces a delicate balance. US growth remains robust, but inflation pressures are complicating monetary policy. In Europe, inflation accelerated and the Bank of England raised rates for the first time since 2023, according to Deloitte. A higher-rate environment can restrain AI investment even as governments view the technology as essential to competitiveness.
The AI story, then, is unlikely to resolve into either a clean triumph or a sudden collapse. The more plausible outcome is consolidation: fewer speculative projects, greater political oversight and a sharper separation between technologies that deliver measurable value and those sustained mainly by expectation. The boom can continue, but its next justification will have to be economic performance.