The sell-off in big tech shares after companies unveiled plans to spend around $660bn on artificial intelligence has rattled markets. Investors have recoiled at the sheer scale of the numbers. Fears of an AI bubble have returned with force.
I understand the unease. When , , and collectively signal a 60% jump in capital expenditure compared with last year, it grabs attention. When Amazon alone flags spending of up to $200bn in a single year, far beyond expectations, it inevitably triggers questions about discipline, returns and excess.
But fear is a poor guide to assessing long-term investment logic. Much of the current anxiety rests on a flawed framing of what this spending represents and how value from AI will actually emerge.
The first mistake is to treat this investment as though it were tied to a single product or service that must quickly prove itself in isolation. That assumption misses the point. What is being built is foundational infrastructure. Data centres, specialised chips and AI platforms form the base layer of modern digital businesses. They sit beneath everything else.
This kind of infrastructure does not announce its value through a neat, standalone revenue line. It shows up gradually, across an entire ecosystem. Better performance. Lower costs. Stickier customers. Stronger pricing power. At the scale these firms operate, even small efficiency gains compound into substantial and durable earnings support.
Another issue is timing. Markets tend to react to cash outlays rather than economic life. Much of this spending is front-loaded. The assets, however, are long-lived. Accounting spreads its costs over many years, even though the investment decision is immediate. Short-term pressure on cash flow tells us very little about long-term profitability.
There is also an assumption that AI must justify itself directly. That is a narrow view. AI does not need to become a standalone profit centre to be valuable. Its impact is felt through improved customer retention, reduced churn and enhanced functionality across existing platforms. Those effects are harder to model, but they are powerful.
The cloud businesses make this clearer. As AI workloads mature, they anchor customers more deeply into ecosystems. Contracts become larger. Switching costs rise. Over time, this dynamic supports stronger margins rather than weaker ones. These cloud platforms already generate exceptional profitability. Advanced AI extends that advantage rather than undermining it.
Some of the spending is also defensive. Scale matters. In AI, relevance depends on compute capacity, data and integration. Falling behind carries strategic risk. Markets may dislike competitive escalation, but for the companies involved, standing still carries greater danger.
History offers a useful perspective. Previous waves of infrastructure investment were also greeted with scepticism. Fibre networks, cloud computing and mobile data all faced periods of doubt when capital spending surged ahead of visible returns. In each case, the spending later proved foundational. The companies that committed early shaped entire markets.
Volatility around AI investment reflects uncertainty about timing, not a collapse in underlying logic. Investors are struggling to price when returns arrive, not whether they arrive at all. That distinction matters.
The bigger risk may lie in underinvestment rather than excess. Failing to build sufficient capacity now could constrain growth later, ceding advantage to competitors willing to commit capital with a longer horizon.
AI investment at this scale is uncomfortable to watch. Big numbers always are. Yet discomfort does not equal irrationality. What is unfolding looks less like a reckless gamble and more like a deliberate effort to secure future earnings power.
Fear may dominate today’s headlines. Time will judge whether patience proves the better strategy.