Morgan Stanley CEO: AI Infrastructure Will Cost $10 Trillion to $15 Trillion Over the Next 10 Years

In the next 10 years, AI infrastructure will cost $15 trillion. Source: Bandaozongheng. Published: 2026-11-04 10:54. AI market data generated poster. It will boost GDP growth by 50 to 70 basis points.

A set of forecast figures recently put forward by Morgan Stanley Chairman and CEO Ted Pick provides a new reference point for the scale and macroeconomic impact of the global AI investment wave. He said that in the next 10 years, capital expenditure in the AI field may require $10 trillion to $15 trillion, and drive GDP growth by 50 to 70 basis points. This estimate elevates AI infrastructure from an industry topic to the level of a macroeconomic variable, triggering widespread market discussion.

Looking at the numbers themselves, $10 trillion to $15 trillion in ten-year capital expenditure is an astonishing scale, equivalent to an annualized $1 trillion to $1.5 trillion, comparable to rebuilding several large national-level infrastructure projects every year. And the 50 to 70 basis points of GDP pull means AI investment is expected to contribute about 0.5 to 0.7 percentage points of annualized incremental growth to the global economy over the ten-year horizon. Against the backdrop of generally slowing growth in major economies, this contribution is regarded by many institutions as structural support worthy of attention.

Ted Pick made the above remarks on the occasion of a recent public event when discussing the long-term impact of AI. As the head of a top global investment bank, his remarks represent both the observation perspective of the capital market and Wall Street's cautious weighing of AI investment returns and industry cycles. Morgan Stanley has long tracked technology and infrastructure investment, and its estimate of AI capital expenditure is backed by a large amount of tracking data on capital expenditure by technology giants, data center construction, and power supporting facilities.

From the perspective of industry reality, this forecast is not groundless. Global technology giants are rapidly increasing their investment in AI computing power, with data center construction, GPU and AI chip procurement, network and storage supporting facilities, and power supply becoming the main destinations of capital expenditure. A number of leading cloud vendors and computing power companies have raised their capital expenditure guidance for several consecutive quarters, treating AI infrastructure as a must-win battleground that determines future competitive positioning. Morgan Stanley's estimate is precisely a scaled extrapolation of this investment wave.

From an economic logic perspective, the pull of AI capital expenditure on GDP is mainly realized through several channels. First, the implementation of physical investment such as data centers and chip manufacturing is directly counted in fixed capital formation, forming current demand; second, the penetration of AI into production efficiency and new business forms may improve total factor productivity in the medium and long term; third, the upstream and downstream industrial chains around the computing power ecosystem, including equipment, software, power, and engineering services, will drive broad employment and supporting investment. The estimate of 50 to 70 basis points combines the pull of direct investment with indirect multiplier effects.

Of course, this forecast is also accompanied by obvious controversy. Opponents worry that if capital expenditure of such scale cannot obtain matching returns on the revenue side, it may repeat the overinvestment and overcapacity of the internet bubble period. There are already many voices warning that the pace of AI application implementation and commercialization may lag behind the expansion of computing power supply. If the investment return cycle is lengthened, the sustainability of capital expenditure will face a test. Ted Pick himself also emphasized that the key lies in how to transform investment into actual productivity improvement, rather than simply piling up computing power.

From the perspective of the industrial chain, regardless of whether the final scale of AI investment falls at the upper or lower end of the range, its driving effect on semiconductors, power, data centers, optical modules, and other segments is already very clear. The continuous iteration of chip computing power, the increase in data center power density, and the accompanying demand for power and heat dissipation constitute the main line of AI infrastructure investment. For the semiconductor industry, the long-term prosperity of AI capital expenditure means that upstream equipment, materials, advanced packaging, and even segmented areas such as power management and precision components will all receive sustained order support.

It is worth noting that the distribution of AI infrastructure investment is uneven, and power supply is becoming a hard constraint restricting the expansion of data centers. Computing power projects in multiple regions face approval or construction delays due to grid capacity and energy consumption quotas. Supporting investment in power infrastructure is becoming a part of AI capital expenditure that cannot be ignored, and to a certain extent it also amplifies the overall investment scale estimate.

Compared with large-scale infrastructure cycles in history, AI capital expenditure shows more distinct characteristics of being technology-intensive and fast-updating. Traditional infrastructure has long depreciation cycles and relatively gentle diminishing marginal returns, while AI computing power facilities face rapid technological iteration. Chip replacement and architecture upgrades may cause some early investment to face value impairment within a relatively short period of time. This both amplifies the scale of investment and raises the risks involved. How to balance the scale, pace, and return of investment is a common issue faced by the industry and enterprises.

This article is reproduced from media reports or online platforms and represents the author's personal position or views. Our reproduction is solely for sharing and does not mean that we approve or endorse it. If the source is incorrectly marked or your lawful rights and interests are infringed, please contact customer service in a timely manner. As a neutral platform service provider, we will promptly correct, delete, or handle it in accordance with the law. Comments: No user comments yet. Latest News More Related Recommendations

Data center revenue up 175% year-on-year, ON Semiconductor acquires Ferric with heavy investment to strengthen AI power supply 2026-11-04 10:56

India packaging opportunity, do small and medium-sized OSATs still have a chance? 2026-11-04 10:37

Anthropic's computing power scale will exceed 3.5 gigawatts by the end of 2026, with next year's target doubling 2026-11-04 10:29