The recent selloff in technology and artificial intelligence-related equities has raised questions about the durability of the AI infrastructure investment cycle.
Political uncertainty, escalating capital spending and concerns over investment returns have contributed to the correction. Yet market sentiment appears to have deteriorated more rapidly than the underlying demand for computing capacity.
At their respective lows during the recent correction, the Nasdaq-100 had fallen nearly 15 per cent from its previous peak, while the Philadelphia Semiconductor Index had declined almost 39 per cent from its June high.
Democratic victories in U.S. Congress and growing opposition to data-centre development have heightened concerns over regulation, electricity costs and permitting. Hedge fund deleveraging and the unwinding of crowded AI positions have further amplified the selling.
These risks warrant scrutiny. But short-term disruptions to individual projects should not be confused with a breakdown in the structural forces driving this global investment cycle.
A global investment cycle
AI infrastructure is not exclusively an American growth story. Hyperscalers are expanding computing capacity across Asia, Europe and the Middle East as businesses and governments accelerate AI adoption.
Microsoft, for example, announced US$10 billion of investment in Japan through 2029, covering AI infrastructure, cybersecurity, and workforce development, alongside plans for more than US$10 billion of capital and operating spending across the Middle East through 2030.
The International Energy Agency projects global data-centre electricity consumption will nearly double, from 485 terawatt-hours in 2025 to 950 terawatt-hours by 2030.
This expansion points to substantial demand for semiconductors, power generation, cooling systems, and electrical infrastructure. U.S. political developments may influence where and when investments occur, but they do not eliminate the need for computing capacity worldwide.
Investment will follow opportunity
Within the United States, states and municipalities compete to attract investment, jobs, and tax revenues. Developers facing restrictions in one jurisdiction can evaluate alternatives offering reliable power, transmission capacity, suitable sites, and more predictable approvals.
Relocation is neither immediate nor costless. Grid interconnections, equipment lead times and financing requirements can create delays, and some projects may be cancelled. Opposition to data centres also crosses party lines, particularly where electricity affordability, land use and water consumption are at issue.
Nevertheless, constraints in individual jurisdictions are not equivalent to a nationwide halt in development. They may redistribute investment rather than eliminate demand.
The strategic race with China
AI leadership carries consequences for productivity, economic competitiveness, cybersecurity, and national security. Competition with China gives the United States a powerful strategic incentive to maintain investment in advanced computing, regardless of which party controls Congress.
Democratic control of the incoming Congress introduces additional policy uncertainty, but presidential veto authority and legislative constraints limit the scope of immediate federal changes.
State and local permitting decisions can present more direct obstacles. Distinguishing these different sources of regulatory risk is essential to assessing their impact on corporate earnings.
The investment cycle is not the investment case
According to Boston Consulting Group, Amazon, Alphabet, Meta, Microsoft, and Oracle are projected to invest nearly US$3.9 trillion in AI capital expenditures between 2026 and 2030.
The extraordinary scale of these commitments underscores the long-term nature of the infrastructure buildout, supporting demand across the semiconductor, electrical equipment, and power infrastructure industries.
These substantial outlays also raise legitimate questions about future revenue, financing requirements, and returns on invested capital.
A durable industry growth cycle does not make every AI-related stock an attractive investment. Established businesses with substantial free cash flow, strong balance sheets and visible earnings growth differ fundamentally from speculative companies reliant on external funding or unproven projects.
Valuation is equally important. Strong secular demand cannot compensate for paying too much for anticipated earnings. Conversely, a sharp share-price decline does not necessarily signal deterioration in a company’s operating outlook. The opportunity lies in separating changing market sentiment from changing business fundamentals.
Fundamentals matter more than headlines
History provides useful perspective on severe corrections. My analysis of Bloomberg data over the past 20 years shows that following Nasdaq-100 drawdowns of 15 per cent or more, the index delivered positive 12-month returns in 75 per cent of non-overlapping episodes, with an average gain of 21.2 per cent and median gain of 31.4 per cent.
The long-term investment opportunity in AI infrastructure remains compelling, but successful investing requires distinguishing structural growth from speculative excess.
Periods of indiscriminate selling can create opportunities in financially sound businesses with sustainable earnings growth and reasonable valuations.
Portfolio decisions must also reflect diversification, liquidity requirements, investment horizons, and risk tolerance, within the context of each client’s broader total wealth plan.
