Goldman Sachs believes that the widely cited figure of approximately $800 billion in AI-related capital expenditures significantly understates the true scale, as it omits investments by non-U.S. companies and private firms and includes some non-AI spending. After adjustments, the firm estimates that global AI investment will reach $1.02 trillion by 2026, indicating that both the magnitude and duration of the AI capital expenditure cycle are stronger than market expectations.
Global AI investment is significantly larger than commonly perceived. A recent Goldman Sachs research report indicates that the widely cited forecast of approximately USD 800 billion in capital expenditures by hyperscalers is systematically underestimated. Once investments by private companies and non-U.S. firms are included, and non-AI-related spending is excluded, total global AI investment is projected to reach USD 1.019 trillion by 2026.
According to the ZuiFeng Trading Desk, Goldman Sachs economists Joseph Briggs and Sarah Dong stated in their August 2 Global Economic Analysis report that the frequently referenced figure of approximately USD 794 billion in hyperscaler capital expenditures both underestimates total global AI capital spending by roughly USD 200 billion and overestimates U.S.-based investment by about the same amount.
After adjustments, Goldman Sachs estimates that AI investment within the United States will amount to approximately USD 581 billion in 2026, while the global total will reach around USD 1.019 trillion. Two cross-validation approaches—revisions to gross profit forecasts for listed companies and official national accounts and trade data—point to global AI investment of approximately USD 1.06 trillion and USD 1.002 trillion, respectively, closely aligning with the primary estimate.
For macro markets, this revised estimation has direct implications: it suggests that both the scale and duration of the AI capital expenditure cycle are stronger than previously anticipated. Goldman Sachs’ leading indicators show that recent momentum remains robust, although trade data from Taiwan and South Korea signal a modest deceleration in investment growth in June and July.
Four Key Flaws in Commonly Used Metrics
The Goldman Sachs report identifies four fundamental flaws in using hyperscaler capital expenditures as a proxy for AI investment.
First, this metric overlooks investments by U.S. private companies, which play a critical role in the AI ecosystem, as well as capital expenditures by other publicly listed firms. According to Goldman Sachs’ credit team, hyperscalers directly account for only 40% of AI-related supply in 2026.
Second, the metric entirely omits investments by non-U.S. companies, particularly those in China and other parts of Asia.
Third, hyperscalers’ capital expenditures exceeded USD 150 billion even before the AI boom began, indicating that a portion of current spending is unrelated to AI.
Fourth, U.S.-based hyperscalers operate globally, and a significant share of their capital expenditures actually occurs outside the United States.
Based on the above assessment, Goldman Sachs has made multi-dimensional adjustments to capital expenditure data for major hyperscale cloud providers: incorporating capital expenditure forecasts of other publicly listed companies in its AI investment basket, supplementing with media-disclosed data from key private companies, adding capital expenditures from non-U.S. AI-related firms, and using 2022 capital expenditure levels as a baseline to exclude non-AI-related investments.
All three approaches converge on the same conclusion: over USD 1 trillion.
Goldman Sachs employed three independent methodologies to estimate global AI investment, yielding highly consistent results.
The primary estimation methodology (enhanced hyperscale cloud provider capital expenditure) indicates that global AI investment will reach USD 1.019 trillion in 2026, of which USD 581 billion will be spent within the United States. Geographically, Goldman Sachs allocated spending based on disclosed project locations of hyperscale cloud providers, estimating that approximately 70% of U.S. hyperscale cloud providers’ capital expenditures will flow into domestic projects, 15% into Asia, and 9% into Europe.
The first cross-validation method measured incremental end-demand by tracking revisions to gross profit forecasts of AI-related public companies relative to the 2022 baseline, resulting in an estimated global AI investment of approximately USD 1.06 trillion in 2026, representing cumulative AI-related spending increases of over USD 1 trillion since 2022.
The second cross-validation method leveraged official national accounts and global trade data. U.S. national accounts data show that, as of May 2026, annualized U.S. AI-related hardware investment had risen to approximately USD 463 billion (relative to the 2022 baseline), with an additional USD 100 billion in AI-related R&D and intellectual property investment, bringing the current annualized total of U.S. AI investment to nearly USD 600 billion. For countries with limited data availability, Goldman Sachs used global trade data and historical relationships between U.S. imports and total investment to estimate global AI investment at approximately USD 1.002 trillion.
Averaging results across the three methodologies suggests that cumulative global AI investment from 2022 through the end of 2026 will reach USD 1.8 trillion.
Capital expenditure as a share of GDP is expected to continue rising, consistent with historical technology cycles.
Regarding the medium- to long-term trajectory of AI-related capital expenditure, Goldman Sachs extrapolates from market consensus estimates of public companies’ capital expenditure, projecting a continued increase in AI capital expenditure as a share of GDP.
Specifically, AI-related capital expenditure as a share of U.S. GDP is projected to rise from 1.8% in 2026 to 2.5% in 2027 and further to 2.8% in 2028; on a global basis, the corresponding figures are 0.9%, 1.3%, and 1.4%, respectively.
Goldman Sachs noted that the aforementioned level aligns with the historical peak investment impact of 2% to 5% of GDP observed during general-purpose technology (GPT) build-out cycles. Even if market forecasts for 2027 capital expenditures have significant room for upward revision, AI investment as a share of GDP would still remain within the historically reasonable range for technology cycles.
Goldman Sachs also pointed out that the timing of a slowdown in AI-related capital expenditure growth is one of the key sources of uncertainty in the current macro outlook, and recommended adopting a 'dashboard' approach to track multiple leading indicators comprehensively, including semiconductor manufacturing equipment imports from Taiwan and South Korea, relevant PMI subcomponents, import prices, and memory procurement and GPU leasing costs.
All leading indicators are currently at elevated levels relative to those seen since 2022, indicating that near-term growth prospects remain robust.
Inflation erodes real investment gains, limiting its contribution to GDP growth.
Although nominal AI investment continues to expand, Goldman Sachs cautioned investors to pay attention to the erosion of real investment growth caused by cost inflation.
Official U.S. data indicate that, year-to-date in 2026, 8% of the increase in nominal AI-related hardware spending can be attributed to cost inflation rather than real investment expansion. If this trend persists into the second half of 2026, the boost to real investment from higher AI-related spending in 2026 will be smaller than in 2025.
Goldman Sachs also emphasized that AI investment’s impact on overall U.S. GDP remains limited, due to two measurement biases: first, U.S. national accounts do not classify semiconductor purchases as investment goods; and second, the high import content of AI hardware is netted out when calculating GDP. This implies that even sustained rapid growth in AI capital expenditures will face structural constraints in its direct contribution to aggregate economic output.
Editor/lambor