How Significant Are Changes in Overseas AI Investment?

Author: CICC Insight

Summary

Recently, overseas AI capital expenditure has continued to expand significantly, creating spillover effects on the Chinese economy through the global AI industry chain. On one hand, the construction of overseas data centers and AI infrastructure has driven demand for servers, optical modules, PCBs, communication equipment, and related components, supporting the export of related Chinese products. On the other hand, the growth in overseas orders has also improved the revenue expectations of domestic AI industry chain companies, leading to increased investment in capacity expansion, which in turn boosts fixed capital formation. Looking ahead to 2027, according to market consensus, overseas AI capital expenditure is expected to continue to grow, but the year-on-year growth rate may decline from the high levels of 2026. If the growth rate of overseas AI capital expenditure slows down, what marginal impact will it have on the Chinese economy?

I. Overseas AI Capital Expenditure Continues to Expand, but Growth Rate May Decline from Q4 2026

We primarily focus on the impact of changes in U.S. AI capital expenditure. In the second quarter of 2026, the capital expenditure of five representative cloud vendors—Amazon, Alphabet, Microsoft, Meta, and Oracle—grew by 86.5% year-on-year, continuing a rapid expansion trend (Chart 1). However, according to FactSet consensus expectations, the scale of overseas AI capital expenditure, represented by cloud vendors, is expected to continue to grow, but the year-on-year growth rate may slow down from Q4 2026. We believe that, in addition to high base factors, the slowdown in growth may also be related to the gradual emergence of constraints such as financing, physical limitations, and security governance.

Chart 1: Market Expectations for Year-on-Year Growth Rate of Overseas AI Capital Expenditure to Slow in Q4 This Year

Note: The overseas AI capital expenditure for Q3 2026 and beyond is based on FactSet consensus forecasts.

Source: FactSet, CICC Research Department;

First, as the capital stock expands rapidly, corporate investment returns and financing constraints are gradually increasing. Over the past two years, cloud vendors have rapidly expanded their capital expenditure, putting pressure on free cash flow (Chart 2) and increasing reliance on external financing. Against the backdrop of relatively high long-term interest rates and rising credit spreads for some cloud vendors, the increase in financing costs may further raise the return threshold for new investments. Meanwhile, the large capital stock formed in the early stages also means that future depreciation and amortization pressures will rise. As the commercialization of AI still needs further realization, the recent gap between ROIC and WACC has also narrowed (Chart 3), leading corporate capital allocation to gradually shift from the previous focus on computing power and scale to a greater emphasis on verifying investment returns, which may result in a slowdown in the growth rate of AI capital expenditure.

Chart 2: The Proportion of Capital Expenditure of the Five Major Cloud Vendors to Operating Cash Flow is Rising

Source: FactSet, CICC Research Department

Chart 3: The ROIC and WACC of Cloud Vendors are Gradually Narrowing

Source: FactSet, Bloomberg, CICC Research Department

Second, physical constraints are beginning to limit the speed of project implementation. The expansion of data centers places higher demands on grid capacity, electricity prices, water resources, and land (Chart 4). Since 2025, disputes in some regions of the U.S. regarding electricity and water use for data centers and community costs have intensified, and constraints have been strengthened through the suspension of approvals and increased entry requirements, forcing a slowdown in the expansion of data center projects. The expansion of overseas AI infrastructure may face pressure characterized by "strong demand but limited project implementation."

Chart 4: The Expansion of Data Centers Places Higher Demands on Power Supply

Source: IEA, CICC Research Department

Third, security governance risks may lead to a slowdown in capital expenditure. As AI's ability to autonomously complete complex tasks rapidly improves, the number of risk events in actual operations is also increasing. The importance and urgency of AI security governance are rising simultaneously. Recently, Anthropic CEO Dario Amodei suggested that the pace of enhancing the capabilities of cutting-edge models should be moderately slowed down to allow more time for third-party evaluation and risk prevention. If similar initiatives are further transformed into industry coordination or regulatory requirements, the training cycles for large-scale cutting-edge models may be extended, thereby marginally constraining the growth rate of training computing power demand.

II. Export Channels: A Slowdown in Overseas AI Capital Expenditure May Reduce Its Contribution to GDP Through Exports

If overseas AI capital expenditure slows down, what impact will it have on the Chinese economy? We believe that the most direct impact comes from exports. The construction of overseas data centers requires a large number of servers, optical modules, PCBs, network devices, and electronic components, in which the Chinese supply chain has a strong participation. Therefore, the previous round of overseas AI investment upturn has shown a significant synchronization with the improvement in Chinese exports.

From quarterly data, we find that the growth rate of overseas AI capital expenditure generally leads the year-on-year growth of Chinese exports and AI chain product exports by about one quarter (Chart 5). Our calculations show that for every 10 percentage points increase in year-on-year overseas AI capital expenditure, the year-on-year growth of Chinese AI chain product exports increases by about 2.8 percentage points, corresponding to a transmission coefficient of about 0.28. According to current consensus expectations, the year-on-year growth of major cloud vendors' AI capital expenditure is expected to decline from 89.9% in 2026 to 38.6% in 2027, a decrease of about 51.3 percentage points. Correspondingly, the year-on-year growth rate of Chinese AI chain product exports is expected to maintain positive growth but may marginally slow down by about 10 percentage points. However, we expect that overseas AI capital expenditure will still maintain positive growth next year, providing some support for Chinese AI chain exports, although the contribution may decline compared to this year.

Chart 5: The Growth Rate of Overseas AI Capital Expenditure Generally Leads Chinese Exports by About One Quarter

Note: The overseas AI capital expenditure for Q3 2026 and beyond is based on FactSet consensus forecasts.

Source: FactSet, Wind, CICC Research Department

III. Investment Channels: The Impact of Slowing External Demand is Relatively Lagged

In addition to directly driving exports, overseas AI demand may also further impact the investment in the AI industry chain by increasing orders and capacity utilization rates for domestic AI industry chain companies. Therefore, the impact of overseas AI capital expenditure on domestic investment is essentially an investment effect induced by external demand.

Based on data from A-share AI industry chain listed companies, we weighted fixed asset investment by the proportion of overseas income to construct an investment indicator related to overseas demand in the AI industry chain. Our calculations indicate that changes in overseas AI capital expenditure have a longer transmission lag on domestic AI industry chain investment, with overseas AI capital expenditure generally leading AI industry chain investment by about one year (Chart 6), with a transmission coefficient of about 0.49. This also aligns with the economic logic that companies require a certain amount of time from receiving overseas orders, confirming capacity gaps, to expanding factories and purchasing equipment.

Chart 6: Overseas AI Capital Expenditure Generally Leads AI Industry Chain Investment by About One Year

Note: We weighted fixed asset investment by the proportion of overseas income to construct an investment indicator related to overseas demand in the AI industry chain. Note: The overseas AI capital expenditure for Q3 2026 and beyond is based on FactSet consensus forecasts.

Source: FactSet, Wind, CICC Research Department

From the perspective of marginal impact on economic growth, unlike the export channel, due to the longer lag in investment transmission, the high growth of overseas AI capital expenditure in 2026 may still support domestic investment into 2027. According to current estimates, we expect that as the slowdown in overseas AI capital expenditure gradually transmits, its contribution to domestic investment may decline compared to previous levels by 2028.

In summary, from the perspective of impact channels, the slowdown in overseas AI capital expenditure will not have a synchronous impact on the Chinese economy: the export channel reacts quickly, while the investment channel reacts with a lag. Therefore, we expect that the main impact in 2027 may first manifest as a weakening of export-driven growth, while the expansion investments brought by previous overseas orders will still provide some buffer, and by 2028, the lagged impact on the investment channel may gradually become apparent.

Finally, it should be noted that our calculations are static partial equilibrium analyses, primarily identifying the impact of changes in overseas AI demand. If domestic AI capital expenditure continues to grow rapidly, especially if domestic cloud vendors, computing power infrastructure construction, and domestic substitution-related investments continue to expand, it may provide some hedge against the slowdown in overseas demand. Additionally, if the penetration rate of AI in enterprise operations increases, domestic AI investment estimated using cloud vendors may be underestimated, and the actual resilience of AI industry chain investment may also be stronger than the results derived solely from changes in overseas demand in this article.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.