Global AI investment is far larger than commonly perceived. Goldman Sachs' latest research report indicates that the widely cited capital expenditure forecast of roughly $800 billion for hyperscale cloud service providers is systematically underestimated. When factoring in investments from private companies and non-U.S. firms, and stripping out non-AI-related spending, total global AI investment is projected to reach $1.019 trillion by 2026.

According to Goldman Sachs economists Joseph Briggs and Sarah Dong, in an August 2nd Global Economics Analysis report, the frequently referenced figure of approximately $794 billion in hyperscaler CapEx underestimates total global AI spending by about $200 billion while simultaneously overestimating U.S. domestic investment by a similar amount. After adjustments, Goldman estimates U.S. AI investment in 2026 at roughly $581 billion, with the global total reaching approximately $1.019 trillion.

Two cross-validation methods—one tracking listed companies' gross profit forecast revisions and the other using official national accounts and trade data—point to global AI investment of around $1.06 trillion and $1.002 trillion, respectively, closely aligning with the primary estimate. For the macro market, this recalculation carries direct implications: it suggests the AI CapEx cycle has greater magnitude and persistence than previously expected, with Goldman's leading indicators showing robust near-term growth momentum, although trade data from Taiwan and South Korea hints at a possible moderate slowdown in investment growth during June and July.

Four Key Flaws in Commonly Used Indicators

The report identifies four fundamental flaws in using hyperscaler CapEx as a proxy for AI investment. First, this metric overlooks investments from U.S. private companies playing a critical role in the AI ecosystem, as well as other listed firms' CapEx—Goldman's credit team data shows hyperscalers directly account for only 40% of AI-related supply in 2026. Second, it completely excludes non-U.S. corporate investment, particularly from China and other Asian regions. Third, hyperscaler CapEx already exceeded $150 billion before the AI boom, meaning some current spending is unrelated to AI. Fourth, U.S. hyperscalers operate globally, with a significant portion of their CapEx occurring outside the United States. Based on these observations, Goldman made multi-dimensional adjustments to standard hyperscaler CapEx data: incorporating other listed companies' CapEx forecasts into its AI investment basket, supplementing key private firms' media disclosures, adding non-U.S. AI-related companies' capital expenditures, and using 2022 CapEx levels as a baseline to exclude non-AI investments.

Three Methods Converge on Over $1 Trillion

Goldman employed three independent methods to estimate global AI investment, with results showing high consistency. The primary estimate (enhanced hyperscaler CapEx) indicates $1.019 trillion in global AI investment for 2026, with $581 billion within the U.S. Geographically, based on location data from announced hyperscaler projects, Goldman estimates approximately 70% of U.S. hyperscaler CapEx goes to domestic projects, 15% to Asia, and 9% to Europe. The first cross-validation method tracks final demand increases by measuring revisions in AI-related listed companies' gross profit forecasts relative to 2022 benchmarks, yielding a global AI investment figure of about $1.06 trillion in 2026, with cumulative AI-related spending increases exceeding $1 trillion since 2022. The second cross-validation method relies on official national accounts and global trade data. U.S. national accounts show AI-related hardware investment reaching an annualized $463 billion (above 2022 baseline) by May 2026, plus roughly $100 billion in AI-related R&D and intellectual property investments, bringing current U.S. AI investment annualized to nearly $600 billion. For other countries with limited data disclosure, Goldman uses global trade data and historical relationships between U.S. imports and total investment to estimate global AI investment at approximately $1.002 trillion. The average of the three methods suggests cumulative global AI investment from 2022 to end-2026 will reach $1.8 trillion.

AI CapEx as Share of GDP Expected to Rise, Aligning with Historical Tech Cycles

Regarding the medium-to-long-term trajectory, Goldman extrapolates from market consensus forecasts for public companies' CapEx, expecting AI capital expenditure as a share of GDP to keep rising. Specifically, U.S. AI CapEx as a share of GDP is projected to climb from 1.8% in 2026 to 2.5% in 2027, and further to 2.8% in 2028; globally, the figures correspond to 0.9%, 1.3%, and 1.4%. Goldman notes these levels are consistent with historical general-purpose technology (GPT) construction cycles where peak investment impacts on GDP ranged from 2% to 5%. Even with significant upward revision potential for 2027 CapEx forecasts, AI investment as a share of GDP remains within a reasonable range for historical tech cycles. Goldman also points out that the timing of an AI CapEx growth slowdown is a key source of macro market uncertainty, recommending a "dashboard" approach to track multiple leading indicators, including semiconductor manufacturing equipment imports from Taiwan and South Korea, relevant PMI subcomponents, import prices, memory procurement, and GPU rental prices. All leading indicators currently remain at high levels since 2022, suggesting robust near-term growth prospects.

Inflation Erodes Real Investment Gains, Limited GDP Boost

Despite expanding nominal AI investment, Goldman warns investors about cost inflation eroding real investment increases. Official U.S. data shows that 8% of the nominal AI-related hardware spending increase so far in 2026 should be attributed to cost inflation rather than actual investment expansion. If this trend continues through the second half of 2026, the boost to real investment from higher AI-related spending in 2026 will be smaller than in 2025. Goldman also emphasizes that AI investment's impact on overall U.S. GDP remains limited due to two measurement biases: first, the U.S. national accounts do not count semiconductor procurement as investment; second, the high import content of AI hardware is netted out when calculating GDP. This means even if AI CapEx continues to grow rapidly, its direct contribution to the broader macro economy faces structural constraints.