Investing.com -- The artificial intelligence investment boom is becoming increasingly difficult to sustain at its current pace as technology companies pour billions of dollars into infrastructure without a corresponding acceleration in revenues, raising the risk of an eventual spending slowdown that could weigh on economic growth, Capital Economics said.

Capital Economics said it remained positive about AI's long-term potential to boost productivity but questioned whether revenues would grow quickly enough to justify the unprecedented capital expenditure. The research firm expects AI investment growth to slow in 2027 before total investment falls by around 10% in 2028, marking a shift from a major driver of economic growth to a potential drag.

The warning comes as the five largest US hyperscalers – Amazon, Alphabet, Microsoft, Meta Platforms and Oracle – prepare to spend a combined $800 billion on capital expenditure this year, with spending expected to reach $1 trillion or more in 2027. Much of the increase since 2023 has been driven by investments in AI infrastructure, including data centres, servers and networking equipment.

However, the revenue hurdle to justify the spending is substantial. Capital Economics estimates that annual AI-related revenues from external customers would need to reach between $1.7 trillion and $1.9 trillion by 2030, compared with an estimated $150 billion to $200 billion currently. That would require revenues to increase roughly ninefold to 12-fold within four years.

The firm identified three potential sources of revenue growth: productivity gains that allow AI providers to capture a share of additional economic output, displacement of existing software and business-services providers, and savings from reduced labour costs. However, achieving the required returns would demand much larger productivity improvements and a more significant transfer of income from workers to capital than the firm currently expects.

Physical constraints could also limit the expansion. Shortages of advanced semiconductors and electricity, alongside rising opposition to data centres over their consumption of land, power and water, could delay projects. Capital Economics said political pressure following Democrats' gains in the US midterm elections could also result in tighter planning restrictions, particularly at the state level.

The investment slowdown could have significant implications for economic growth, particularly in the United States and major Asian technology exporters. AI investment has recently added around 0.3 percentage points to annual US GDP growth, a contribution that could largely disappear next year as spending growth slows. An outright 10% decline in AI investment in 2028 could shave another 0.1 percentage point off US GDP.

Taiwan and South Korea face greater exposure because of their dependence on AI-related manufacturing and exports. Capital Economics estimates that the diminishing boost from AI demand could reduce annual real GDP growth by around three percentage points in Taiwan and one percentage point in South Korea next year, relative to recent rates. The impact could deepen in 2028, reaching 5.5 percentage points in Taiwan and 1.5 percentage points in South Korea.

The report also warned that disappointing investment returns could create financial spillovers as hyperscalers increasingly turn to external financing. While AI infrastructure spending was initially funded largely through companies' own cash flows, debt issuance, equity financing, leases and private credit have become more important. A deterioration in revenues could therefore expose lenders and other investors to losses, potentially tightening financial conditions.

Capital Economics said the risks would remain relatively contained if the investment boom turned soon, with outstanding hyperscaler bonds still amounting to about $500 billion, or roughly 6% of the $8.3 trillion US non-financial corporate bond market. However, a prolonged investment boom financed increasingly through debt could create much larger financial vulnerabilities.

Despite the risks, the firm does not expect an abrupt collapse in AI investment or a severe economic downturn. Existing infrastructure, replacement spending and continued investment should support further AI adoption, even if spending on physical infrastructure moderates. Software and applications linked to AI deployment could also prove more resilient than spending on GPUs, servers and networking equipment.

Capital Economics' central view is that the AI investment boom will eventually adjust to a more sustainable pace rather than signal the end of AI's economic potential. The greater concern, it said, is that a prolonged period of excessive investment could expose financial and economic imbalances, particularly if the productivity gains and revenues needed to justify the spending fail to materialise.