Is 800 billion in capital expenditures not enough? Goldman Sachs: The calculations underestimated non-U.S. and private equity investments; this year's actual investment in AI is expected to exceed 1 trillion.

Global AI investment is far greater than common market perceptions. Goldman Sachs' latest research report indicates that the widely cited prediction of approximately $800 billion in capital expenditures from hyperscale cloud providers significantly underestimates the total. Once investments from private enterprises and non-U.S. companies are included, while excluding non-AI-related expenditures, the total global AI investment is expected to reach $1.019 trillion by 2026. According to the Wind Trading Platform, Goldman Sachs economists Joseph Briggs and Sarah Dong stated in their August 2nd "Global Economic Analysis" report that the commonly referenced figure of about $794 billion for hyperscale cloud provider capital expenditures underestimates the total global AI capital expenditures by approximately $200 billion and overestimates U.S.-based investment by about $200 billion. After adjustments, Goldman Sachs estimates that AI investment within the U.S. will be approximately $581 billion in 2026, with a global total of around $1.019 trillion. Two cross-validation methodsthe adjusted gross profit predictions of publicly listed companies and official national accounts along with trade datapoint to a global AI investment scale of approximately $1.06 trillion and $1.002 trillion for 2026, which aligns closely with the main estimate. For the macro market, this recalibration has direct implications: it suggests that both the size and sustainability of the AI capital expenditure cycle are stronger than previously expected. However, at the same time, Goldman Sachs' leading indicators show that recent growth momentum remains robust, despite trade data from Taiwan and South Korea suggesting a possible moderate slowdown in investment growth for June and July. Four Major Flaws in Commonly Used Indicators The Goldman Sachs report identifies four fundamental flaws in using hyperscale cloud provider capital expenditures as a proxy for AI investment. First, this indicator overlooks critical investments by U.S. private equity firms within the AI ecosystem, as well as capital expenditures from other publicly listed companiesdata from Goldman Sachs' credit team shows that hyperscale cloud providers directly account for only 40% of AI-related supply in 2026. Second, the indicator completely misses investments from non-U.S. enterprises, particularly those from China and other parts of Asia. Third, capital expenditures by hyperscale cloud providers prior to the AI boom exceeded $150 billion, implying that a portion of current spending is unrelated to AI. Fourth, U.S. hyperscale cloud providers operate globally, with a significant portion of their capital expenditures actually occurring outside the U.S. Based on these judgments, Goldman Sachs conducted a multi-dimensional adjustment of standard hyperscale cloud provider capital expenditure data: including capital expenditure forecasts from other publicly listed companies in their AI investment baskets, supplementing key private equity media disclosure data, integrating capital expenditures from non-U.S. AI-related companies, and excluding non-AI-related investments using the 2022 capital expenditure levels as a baseline. Three Approaches Yield the Same Outcome, All Pointing to Over $1 Trillion Goldman Sachs employed three independent methods to estimate the scale of global AI investment, yielding highly consistent results. The primary estimation method (enhanced hyperscale cloud provider capital expenditures) indicates that global AI investment will reach $1.019 trillion by 2026, with $581 billion occurring within the U.S. In terms of geographical distribution, Goldman Sachs allocated based on the location data of publicly disclosed projects from hyperscale cloud providers, estimating that approximately 70% of U.S. hyperscale cloud provider capital expenditures will be directed towards domestic projects, 15% towards Asia, and 9% towards Europe. The first cross-validation method measures final demand increments by tracking adjustments in gross profit predictions of AI-related publicly listed companies relative to the 2022 baseline, resulting in a global AI investment estimate of approximately $1.06 trillion, showing an increase of over $1 trillion in AI-related expenditures since 2022. The second cross-validation method relies on official national accounts and global trade data. According to U.S. national account data, by May 2026, annualized AI-related hardware investment in the U.S. had risen to approximately $463 billion (compared to the 2022 baseline), in addition to about $100 billion in AI-related R&D and intellectual property investments, bringing current annualized AI investment in the U.S. close to $600 billion. For other countries with limited data disclosure, Goldman Sachs estimated global AI investment at about $1.002 trillion using global trade data and historical relationships between U.S. imports and total investments. The average results from the three methods indicate that cumulative global AI investment will reach $1.8 trillion between 2022 and the end of 2026. The Share of Capital Expenditures in GDP is Expected to Continue Rising, in Line with Historical Technology Cycles Regarding the medium- to long-term trend of AI capital expenditures, Goldman Sachs extrapolates based on market consensus expectations of public company capital expenditures, predicting that the share of AI capital expenditures in GDP will continue to increase. Specifically, the share of U.S. AI capital expenditures in GDP is expected to rise from 1.8% in 2026 to 2.5% in 2027 and further to 2.8% in 2028; the corresponding global figures are expected to be 0.9%, 1.3%, and 1.4%. Goldman Sachs points out that these levels align with historical peaks of 2% to 5% in GDP investment shocks during general-purpose technology (GPT) construction cycles; even if the market anticipates a significant upward adjustment in capital expenditure forecasts for 2027, the share of AI investment in GDP will still fall within a reasonable range of historical technology cycles. Goldman Sachs also notes that the timing of any slowdown in AI capital expenditure growth is one of the core sources of uncertainty in the current macro market, recommending a "dashboard" approach to comprehensively track multiple leading indicators, including semiconductor manufacturing equipment imports from Taiwan and South Korea, relevant PMI sub-indices, import prices, as well as memory procurement and GPU leasing prices. Currently, all leading indicators remain at high levels since 2022, indicating that recent growth prospects remain solid. Inflation Erodes Real Investment Increments, Limiting GDP Impact Despite the continued expansion of nominal AI investment, Goldman Sachs cautions investors to pay attention to the erosion effect of cost inflation on real investment increments. Official U.S. data indicates that as of 2026, 8% of the nominal AI-related hardware expenditure increase should be attributed to cost inflation rather than actual investment expansion. If this trend continues into the second half of 2026, the increase in AI-related expenditures will have a smaller impact on real investment compared to 2025. Goldman Sachs also emphasizes that the AI investment's influence on overall U.S. GDP levels remains limited, for two reasons: firstly, U.S. national accounts do not classify semiconductor purchases as investment products; secondly, the high import content of AI hardware has been net deducted in GDP calculations. This means that even as AI capital expenditures continue to grow rapidly, their direct contribution to the overall macro economy is still structurally constrained. This article is reprinted from "Wall Street Watch"; author: Bu Shuqing; edited by GMTEight: Chen Siyu.