圣路易斯联储分析AI热潮期间大型科技公司的现金与投资变化
Hyperscalers’ Cash and Investment during the AI Boom
圣路易斯联储分析显示,10家大型科技公司的现金资产比从2018年的约三分之一降至2026年初的16%,同期其他上市公司约为11%。这些公司的资本支出占资产比在2024年短暂回落后升至2026年初近13%,研发占资产比则从2023年约11%的峰值降至约9%;两项投资合计占资产比从2018年初的12.5%升至2026年初的22%。
材料呈现大型科技公司现金资产比下降、投资强度上升及投资构成变化,可供了解AI相关投资的企业财务趋势。
KEY TAKEAWAYS
- “Hyperscalers” — large tech firms involved in data centers and computer chips — have shifted financial priorities during the AI boom, reducing their share of cash relative to assets, while directing more spending toward investment.
- Investment activity has grown much faster among major technology companies than in the broader corporate sector.
- The mix of investment is changing: Capital spending has become the main driver of investment growth since 2024, while research and development remains elevated but has eased from its 2023 peak.
Major U.S. technology companies like Amazon.com Inc. and Microsoft Corp. have sharply increased their investments during the artificial intelligence (AI) boom. How did this affect the cash holdings and other spending of these tech companies? In new St. Louis Fed research, we document the impact on their balance sheets and a shift in investment composition: Capital spending has accelerated since 2024, while research and development (R&D) spending relative to assets has declined from its 2023 peak.Julian Kozlowski, Masataka Mori, Juan M. Sánchez and Nicholas Sullivan, “Financing Artificial Intelligence: Cash, Debt, and Corporate Investment,” Federal Reserve Bank of St. Louis Review, Fourth Quarter 2026, Vol. 108, No. 13, pp. 1-20.
In a two-part blog series, we outline our research findings, which focus on 10 technology firms: Alphabet Inc., Amazon.com, Apple Inc., Broadcom Inc., Intel Corp., International Business Machines Corp., Meta Platforms Inc., Microsoft, Nvidia Corp. and Oracle Corp. We refer to this group as “hyperscalers,” using a broad definition that includes semiconductor suppliers alongside cloud and data center operators. We date the start of the AI era to 2018, following advances in deep learning and several years before ChatGPT’s release in late 2022.
This blog post examines how these companies have been allocating their resources by looking at their cash holdings, capital expenditures and R&D. The second post will examine how accelerating investments are affecting operating cash flow, borrowing and payments to shareholders.
Decline in Cash Holdings Relative to Assets
The first figure shows that hyperscalers’ cash holdings — measured by the cash-to-assets ratio — have declined substantially relative to their assets.Each group’s cash-to-assets ratio is the average, weighted by total assets. Their cash-to-assets ratio fell from approximately one-third in 2018 to 20% in late 2022 and 16% in early 2026. Other firms — all other publicly traded companies — experienced a much smaller decline, from roughly 14% in 2018 to 11% in early 2026. Consequently, the gap between the two groups narrowed from approximately 19 percentage points to 5 percentage points.
The decline among hyperscalers was already well underway before ChatGPT’s release in November 2022 and continued afterward. Note that the decline in the cash-to-assets ratio does not necessarily imply an absolute decline in the dollar value of their cash, since their asset base has also grown.
Rise in Capital and R&D Expenditures
In the years leading up to 2018, hyperscalers’ capital expenditures were approximately 4% to 5% of assets, compared with 3% to 4% among other firms, as shown in the second figure. The gap widened after 2018 as hyperscalers’ capital expenditure-to-assets ratio rose to approximately 9% by 2023. Following a temporary decline to about 7% in 2024, the ratio increased sharply, reaching nearly 13% in early 2026. Meanwhile, other firms’ ratio remained near 3%. The recent acceleration in capital spending relative to assets is therefore concentrated among hyperscalers.
Hyperscalers were more R&D-intensive throughout the sample, as shown in the third figure. Before 2018, their R&D expenditures were approximately 7% of assets, compared with roughly 4% among other firms. The gap widened after 2018, with hyperscalers’ R&D-to-assets ratio reaching about 11% in 2023. Although the ratio subsequently declined to approximately 9% in early 2026, it remained above its pre-2018 level. Other firms’ R&D expenditures remained close to 4% of assets throughout most of the period.
Together, the second and third figures show a change in investment composition for hyperscalers. Both capital expenditures and R&D increased relative to assets after 2018. More recently, their paths have diverged: The capital expenditure-to-assets ratio accelerated sharply after 2024, while the R&D-to-assets ratio declined from its 2023 peak. Capital expenditures have therefore driven the most recent increase in investment intensity.
Conclusion
These large technology firms’ cash-to-assets ratio has been declining while they devote more resources to investment. Summing both investment types — R&D and capital expenditures — shows an increase in the ratio from 12.5% in early 2018 to 22% in early 2026. Capital expenditures have driven the most recent increase. These patterns raise a financing question: How are firms supporting this expansion? Our next post will examine operating cash flow, borrowing and payments to shareholders.
Notes
- Julian Kozlowski, Masataka Mori, Juan M. Sánchez and Nicholas Sullivan, “Financing Artificial Intelligence: Cash, Debt, and Corporate Investment,” Federal Reserve Bank of St. Louis Review, Fourth Quarter 2026, Vol. 108, No. 13, pp. 1-20.
- Each group’s cash-to-assets ratio is the average, weighted by total assets.
来源:圣路易斯联储 · 经济分析 · stlouisfed.org