Look Forward — 4 November 2025
Data center investments are increasingly moving the macro needle
Investments related to data centers have become the largest contributor to US growth, countering the effects of policy unpredictability on expenditures more generally.
By Satyam Panday and Paul Gruenwald
Highlights
The data center boom powering the AI revolution is clearly moving the macro needle, especially in the US. This goes well beyond the physical construction of data centers.
Preliminary data suggest that investments in data centers and related high-tech activities led to US GDP being about 0.5 percentage point larger in the second quarter of 2025 than it would have been if businesses’ spending on data center and power construction, information processing equipment, software, and research and development had grown in line with the 2011–2022 trend. 2022 marked an inflection point for data center investment due to multiple factors, including the public release of ChatGPT and the passage of the CHIPS and Science Act.
Looking ahead, the key question is the payoff. Will the surge in data center investment generate productivity gains and fulfill the promise of a multiyear boost in growth?
Any boom in AI-driven growth will likely create a windfall and raise distributional issues. Will labor be augmented or replaced? Tensions are inevitable.
Data center construction has exploded in the past few years. This is a global story, with the US in the lead. Data centers are part of a worldwide AI race turbocharged by the launch of technologies such as ChatGPT. This race has multiple dimensions, spanning economics, geopolitics, energy and technology. In a macro context, data center and related high-tech business investment has implications for growth and labor demand, and ultimately productivity.
Data center investment: A new macro hero
The data center investment macro story is about more than constructing physical buildings. In this sense, the term “data centers” underrepresents the macro impact of AI infrastructure build-out. However, in this discussion, we will use the term as shorthand for a broader macro narrative.
Mapping the AI boom to investment has three major components in the national accounts data. The first is the construction of data centers themselves — and the related investment in power — covered under non-residential buildings. The second is investment in the physical “high-tech” equipment needed to operate the data centers, including racks, servers, mainframes and cooling systems. The third is intangible investment, including the software needed to run data center equipment. Each component has distinct dynamics, as we will illustrate below.
The main economic operators in the data center build-out are the hyperscalers such as Microsoft, Alphabet and AWS that provide large-scale cloud computing services. The prominence of hyperscalers in the AI and data center race represents an interesting twist in their macro story. Historically, the tech industry has been viewed as human capital-heavy with a light capex footprint. This was in contrast to manufacturing, which was increasingly capex-heavy and light on employment. With the build-out of large data centers and the race for AI dominance, however, the tech firms’ macro footprint is growing to resemble that of manufacturing.
While the AI race is global, the US is leading the data center story so far. This reflects both the scale of investments to date and the computing capacity necessary to run large language models. S&P Global 451 Research estimates that US data center capacity represents over 40% of the global total, and that figure is projected to continue growing. Asia-Pacific follows, with Europe a distant third.
This paper will focus mainly on the US due to its leading role in the build-out of data centers as well as the availability and timeliness of data. However, the macro analysis in this paper should extend to other economies. A large proviso is that the AI revolution is moving at breakneck speed, so the situation is in flux.
Moving the macro needle
US macro data show that AI-related investments have surged in the past few years. Importantly, this surge has persisted even as higher borrowing costs and policy uncertainty have weighed on other private investment. Business investment related to AI is offsetting weakness in other investment areas and in the economy at large.
The data show important changes in the composition of fixed investment. Since 2020, capital investment in intellectual property products such as software and research and development has constituted the largest component, while investment in structures, which once held the top spot, has declined in the past few decades.
We also see profound compositional changes within categories. Within structures, a notable exception is outlays on data center construction, which are now almost four times higher than the 2021 level, while construction outlays on other structures remain essentially flat. In equipment, we note further evidence of investors prioritizing high-tech spending, with outlays on information processing equipment outpacing other categories. This provides clear, albeit somewhat indirect, evidence of the growing footprint of data center investment strategies. Firms spend more on information-processing equipment today than they do on transportation and industrial equipment combined.
Also related to data centers, spending on high-tech computer and electronics manufacturing facilities has occurred at breakneck speed in the past four years. There is more money flowing into computer and electronics facilities today than all other manufacturing facility types combined. This high-tech factory boom is in addition to the construction of the data centers themselves, suggesting that focusing solely on the latter would lead to an underestimation of the overall effects.
The surge in investment in data center structures should also eventually lead to greater investment in power generation capacity, as data centers are huge consumers of electricity.
The growing importance of data center investment and related high-tech spending is also evident in production data. High-tech only accounts for about 3% of domestic manufacturing in the US, but capacity is expanding rapidly amid the US push for AI dominance. Driven by data center build-outs, the output from domestic high-tech sector industries has sped up well ahead of other non-energy production. Specifically, output from high-tech industries was up nearly 14% year over year in July 2025, while other non-energy production barely eked out a 1% gain.
Data centers as new GDP growth driver
Data centers and related high-tech investment activities have already become a key driver of US growth. Adding up the components outlined above, current estimates suggest that 80% of the growth in final private domestic demand — GDP less net exports, government and inventory — in the first half of 2025 came from data centers and high-tech-related spending. This development is quite astounding. To be sure, the importance of overall spending on data centers and high-tech industries to GDP growth is not straightforward. A large and growing proportion of hardware is imported. Imports of computers and peripherals increased nearly 60% in the past year, according to data from the US Department of Commerce, and part of the jump in such imports reflects front-running of tariffs. Still, even as the boost to GDP from hardware investment is offset by imports, the related outlays on domestically produced software, research and development, as well as structures investment in data centers and power supply, are helping to lift GDP growth.
The spending numbers are big by themselves as a share of private domestic demand, so much so that had it not been for these capital outlays, the rest of final sales to private domestic purchasers in the US economy nearly flatlined in the first half. We estimate that investments in data centers and related high-tech activities led to US GDP being about 0.5 percentage point larger in the second quarter of 2025 than it would have been if businesses’ spending on data center and power construction, information processing equipment, software, and research and development had grown in line with the 2011–2022 trend.
While US GDP growth has remained stronger than expected over the past few quarters, employment growth has weakened. Tariff uncertainty and immigration curbs have weighed down employment, but there is also preliminary evidence that the use of AI may lead to reduced demand for workers in low-skilled professional services. Hiring in the information sector has also declined from its peak in 2022, such that it is now running 3% below its 2015–2019 trend.
To the extent that entry-level and lower-skilled positions are held by younger cohorts, it is notable that nearly all of the recent rise in the unemployment rate has been among 16–24-year-olds. Given the general-purpose nature of the technology, AI could eventually create offsetting jobs in other fields. However, if history is any guide, the transition period is likely to be shaky.
The combination of positive GDP growth and flat employment points to potential early productivity gains from AI, powered by data centers. While we are still in the early days of the AI revolution, anecdotal evidence suggests that firms are already seeing productivity benefits. Decreased demand for lower-skilled services may be the leading edge of a larger trend. And it may take time for these micro-level phenomena to percolate into the macro data.
Nonetheless, the productivity benefits of AI may show more quickly than in the computer revolution of the 1980s. In that era, Nobel Laureate Robert Solow quipped that “computers are everywhere except in the productivity data.”
Sustained growth: After data centers pass the baton
Data center investment is not the macro endgame for AI. Data center investment puts assets in place to generate sustained growth, but it only directly delivers increased output during construction. In terms of macro accounting, data centers will add to the capital stock. But will that capital stock be profitable to operate, and thereby add to sustained GDP growth and employment? Or will there be stranded, unprofitable assets?
GDP growth is ultimately driven by changes in the labor force and changes in productivity. So, the key question is: Will data center investment strategies translate into sustained gains in productivity? And will any gains in productivity come with an augmented labor force or a reduced one? As noted above, there are hints of productivity gains already, but the final word will not be clear for years.
Economists express a wide range of views regarding the likely impact of AI on labor productivity over the next 10 years. These range from 2024 Nobel Laureate Daron Acemoglu’s fractionally positive projection to forecasts of well over 1 percentage point per year in several other studies.
Notably, these estimates project labor productivity growth, but not the growth of the labor force. Therefore, they do not predict whether GDP growth will be higher. Increases in labor productivity can have positive or negative effects on employment: They can result in displacement of workers, offsetting the growth effects of productivity gains, or in augmentation of workers’ output, boosting the growth effects.
Looking forward
The AI revolution — including ongoing data center investment as well as any larger macro payoff down the road — seems set to be a major driver of economic activity for years to come. This is a potentially positive macro story, but it will inevitably raise distributional issues. We conclude by highlighting two of these.
First, what will we do with any macro windfall? The projected multiyear growth and income gains from the AI revolution are poised to create unanticipated economic gains. Some of these could be sizable. As a result, the economic pie will likely be larger than expected, and with it, the potential for citizens’ well-being in affected economies. A relatively recent example of this dynamic was the “peace dividend” after the end of the Cold War in the 1990s. Options for spending the public sector’s windfall include tax cuts, paying down debt, building out or repairing public infrastructure, funding the energy transition, and investing in health as societies age, to name a few.
Second, the AI revolution will also likely create distributional challenges. Despite the hype, transformational changes to economies are rarely beneficial to all. Globalization provides a recent, poignant example. The global economy saw turbocharged GDP growth and trade, but gains were uneven. In many cases, challenges to those who were negatively affected were not adequately addressed, and this led to ongoing political backlash in numerous places.
AI powered by data centers holds great promise, but its effects will not be uniform. This will be true across sectors and across types of workers. For example, in contrast to the prevailing dynamics of globalization, some categories of professional workers appear likely to be negatively affected by AI. The role of labor and wage growth in an AI-enabled economy is therefore a key development to monitor. Public support for AI adoption will likely wane if distributional issues, along with equally important considerations such as governance and trust, are not adequately addressed.
This article was authored by a cross-section of representatives from S&P Global and in certain circumstances external guest authors. The views expressed are those of the authors and do not necessarily reflect the views or positions of any entities they represent and are not necessarily reflected in the products and services those entities offer. This research is a publication of S&P Global and does not comment on current or future credit ratings or credit rating methodologies.