US Capital Spending Tells Us The AI Economy Is Taking Off
In the first of a multi-part series, Julius Probst, PhD compares the current AI investment boom to those that accompanied other tech revolutions.
Photo Credit: Campbell
After years of mediocre growth, the United States may be entering a period of stronger economic expansion. As Artificial Intelligence (AI) is starting to have a noticeable effect on economy-wide investment, GDP growth, worker productivity and the labor market, one cannot help but wonder whether we are seeing the rise of a new General Purpose Technology (GPT).
What is a General Purpose Technology, you might wonder?
Economists define it as a technology that has a transformative effect on the economy (and society), propelling growth for years or even decades to come. Examples include the railway boom of the late 19th century, electrification of households and businesses starting in the 1920, the automobile (pre-WW2), and the Dotcom boom of the 1990s.
In this piece, we will outline four different mechanisms through which GPTs transform the economy and labor market. But does the AI boom meet these requirements? Let’s find out:
Fact 1: Every new GPT has coincided with a stock market boom
Every economic cycle that has seen the emergence of a GPT has also coincided with a stock market boom. Investors get excited (sometimes overly so) about potential revenue generated through the deployment of the new technology during the economic upswing.
In the late 19th century, investors poured money into railway stocks. Eventually, the market corrected with railway stocks crashing.
The story repeated itself during the “Roaring Twenties” as investors got overly excited about electrification, the automobile, and the economy in general. And again, with the Dotcom bubble.
With today’s AI boom, investors are increasingly wary that U.S. stock prices are historically elevated and potentially in bubble territory.
Fact 2: Capital investment surges
GPTs do not just affect financial markets, though. They have a substantial effect on the tangible economy as capital spending into real assets surges. While stock market frenzies typically end abruptly, the deployment of the new technology creates an infrastructure boom with positive effects for growth long after the bubble pops: The construction of the railway network led to a big economic expansion thanks to surging interstate trade. Electrification and the automobile revolution were the growth story of the 1920s and ‘30s, while the internet and personal computers created the growth boom of the 1990s.
When looking at capital spending today, it surely looks like AI has created a significant investment boom. Annual construction spending on data centers increased from about $10 billion to $40 billion in the last three years alone (+300%). Spending on electricity plants — the infrastructure needed to power those data centers — increased by about $30 billion annually relative to 2020 (+33%).
While manufacturing construction already took off with Biden’s infrastructure stimulus (CHIPS Act and Infrastructure Investment and Jobs Act), it remains approximately three times higher than in 2020. A significant share of that spending touches AI: semiconductor fabrication plants account for a massive chunk, followed by battery plants and clean tech (both of which are needed to support the increased electricity demand coming from AI data centers).
Corporate data reveals that the big American tech companies alone are planning to spend more than $600 billion in capital expenditures compared to $150 billion just a few years ago.
The investment share of GDP dedicated to information processing equipment (like computers and servers), software, and research and development for intangibles is already exceeding levels not seen since Dotcom.
Economic data thus supports the story of a substantial investment boom driven by AI and AI-related economic activities.
Fact 3: Worker productivity picks up, though with a lag
While an investment boom creates higher output quite immediately, the long-run benefits from the new technology and infrastructure spread through the economy more slowly. Companies need to adapt to the changing economic environment, production processes must be rethought and reorganized, products development takes time, workers need to reskill and learn how to use the new technology, etc.
It therefore typically takes a few years until the benefits of a GPT are transformative enough to make workers more productive. While there are early signs that productivity growth is indeed picking up, it is too early to say to what extent AI is a contributing factor.
Fact 4: The occupational mix and the demand for skills shift
Another reason GPTs are transformative is because they create millions of jobs in occupations that previously did not exist. They also replace jobs that rely on and service outdated technologies and infrastructure. This is the “creative destruction” economies need to become more prosperous.
For example, the number of telephone operators increased from just over 150,000 in the 1920 to more than 300,000 in 1950, reaching almost 0.6% of total employment. In the following two decades, more than 250,000 jobs were shed as automation made the occupation redundant.
A faster pace of innovation therefore typically invites worker churn as both job creation and job destruction accelerate. So far, AI hasn’t appreciably driven its own wave of job switching — at least, not yet. For now, other macroeconomic forces have led to a significant slowdown in hiring demand. Global uncertainty, tariffs, and changes to immigration policies have created a low-churn labor market with little worker movement.
Nevertheless, hiring data indicates that the demand for skills in the economy is shifting rapidly with the deployment of AI. In countries like the U.S., the U.K., and Ireland, the number of job postings mentioning AI skills has more than doubled since 2023. Meanwhile, the demand for jobs that are potentially vulnerable to AI replacement — routine white-collar workers and graduates — has fallen.
The AI boom is real - what does that mean for recruiters?
Equity markets are booming, the investment share of GDP is rising, and large U.S. tech companies are spending hundreds of billions of dollars in capex on AI and AI-related infrastructure. All of this speaks to AI being the next big thing transforming our economies — a new GPT.
While one key ingredient is missing – accelerated job creation (and destruction) – AI is changing the demand for skills. Hiring for routine white-collar workers has declined whereas AI knowledge workers are benefiting from the deployment of the technology.
AI is also affecting local labor markets within the U.S. Some metropolitan areas could become more exposed as mid-tier jobs in consulting, tech, and banking are more vulnerable to AI replacement. Meanwhile, the infrastructure and data center boom is creating a hiring bonanza for skilled blue-collar workers – engineers, electricians, technicians, construction workers - in more remote areas of the country: think Virginia, Illinois, Texas, Ohio and Georgia.
While total hires remain low, the demand for skills is shifting, and so is the geography of jobs — which will be the topic of our upcoming follow-up piece.