Is it the prologue of the industrial revolution or the eve of a huge bubble? The future of AI, fueled by trillions of dollars, remains to be seen.
God Translation Bureau is a compilation team under 36Kr, focusing on fields such as technology, business, the workplace, and life, and mainly introducing new technologies, new ideas, and new trends from abroad.
Editor's note: One side is fire, the other side is seawater. Only by understanding the two - extremes of the US economy can we see the real impact of AI. This article is from a compilation.
The US economy can be summarized in one sentence: it is a tale of two extremes.
On one side is the booming artificial intelligence economy, and on the other side is the sluggish consumer economy.
This is evident in the economic data. In the last quarter, the growth of spending in the artificial intelligence field exceeded the growth of consumer spending. Without artificial intelligence, the economic growth of the United States would be negligible.
This is also reflected in the stock market. In the past two years, about 60% of the growth in the stock market has come from companies related to artificial intelligence, such as Microsoft, NVIDIA, and Meta. Without this wave of artificial intelligence boom, the stock market returns would be dismal.
This is also obvious in business data. According to Stripe's data, companies that call themselves "artificial intelligence companies" dominate the revenue growth on the platform, with a growth rate far exceeding that of any other type of company.
No one can be sure whether this wave of artificial intelligence boom is the precursor of the next industrial revolution or the next huge bubble. We only know that it is happening. Discussions about questions like "What will happen when artificial intelligence dominates the economy one day in the future?" can stop. The artificial intelligence economy has already arrived, right now. Whether it is good or bad, we are all in it.
So, what exactly is this wave of artificial intelligence fever? How did it happen? Where does the huge amount of money for building artificial intelligence come from? Who is using this technology? Has it improved people's productivity? This article will try to provide an intuitive guide to this question in the form of chart - based Q&A:
How large is the scale of the artificial intelligence boom?
The components of artificial intelligence are simple: computer chips, server racks in data centers, a huge amount of electricity, and the network and cooling systems that ensure all devices run continuously without overheating.
The cost of these hardware is extremely high. In the past six months, the four companies that have invested the most in the artificial intelligence field - Meta, Google, Microsoft, and Amazon - have spent between $100 billion and $200 billion on chips, data centers, etc. Christopher Mims of The Wall Street Journal wrote: "The most valuable technology companies are purchasing and building at an unprecedented pace."
The Wall Street Journal
Can we put these numbers in a historical context?
This is the largest technological infrastructure project since the 1960s (the beginning of the computer era) or since the 1880s (the heyday of the railway era).
In January this year, Michael Cembalest of JPMorgan Chase calculated that NVIDIA, the leading artificial intelligence chip manufacturer, is expected to account for the highest share of market capital expenditure since IBM's revenue peaked in 1969. Economic writer Paul Kedrosky also didn't lag behind. He calculated that the proportion of artificial intelligence capital expenditure in GDP has exceeded that during the Internet bubble period and is now approaching levels never seen since the railway construction in the "Gilded Age".
JPMorgan Chase
Where does all this money come from?
The booming development of today's artificial intelligence infrastructure benefits from the extraordinary and unprecedented profits of top - tier technology companies. As Cembalest explained, today's top - tier technology companies have made huge profits in the past few years. The share of their total "free cash flow" (that is, revenue minus operating expenses and infrastructure expenditure) dwarfs any period since the end of World War II. The existing business models of these companies - whether it's Meta's advertising or Google's search advertising - are sufficient to generate huge amounts of money to invest in the next - generation technology. Cembalest told me: "They are generating unprecedented free cash flow. They are making a fortune, which is why they are able to invest hundreds of billions of dollars in annual capital expenditure in research and development and infrastructure related to artificial intelligence."
JPMorgan Chase
So, is this the reason why the stock market is performing so strangely?
I think so. There is an interesting debate in the financial circle about why the stock market seems to ignore Trump's tariff policies and the slowdown in economic growth. I think the most obvious answer is a combination of the following points: (a) Some investors still think that Trump will back down on the tariff issue; (b) They think that the ultimate impact of tariffs will not be significant; and (c) Tariffs have little impact on the digital economy, while artificial - intelligence - related stocks dominate the market returns, and the rest of the market is stagnant.
As shown in this chart from Société Générale, the ten largest companies in the S&P 500 index have absolutely dominated the net income growth in the past six years. So, it may be more appropriate to view the market as the "S&P 10 index" versus the "S&P 490 index" now. If you are a portfolio manager who has invested in the other 490 stocks, the equity returns in the past six years will not impress you, because the overall profits of these companies have not increased.
Société Générale
Okay, companies are investing an unprecedented amount of money. Have they made it back?
Not yet. As Greg Ip of The Wall Street Journal wrote, the "disturbing" aspect of the artificial intelligence boom is that all this spending on chips and data centers is "draining the cash of US companies." OpenAI and Anthropic have suffered large - scale losses, and the largest technology companies still rely on their old business models to create the highest profit margins. If these companies' spending far exceeds the costs they may recover in the future, it means that we are in a historic infrastructure bubble.
From an optimistic perspective: Payment company Stripe has found evidence that the revenue growth rate of artificial intelligence startups exceeds that of any previous generation of technology. The company announced in a recent report: "Artificial intelligence companies are reaching revenue milestones faster than previous generations of startups." "Among the top 100 artificial intelligence companies on the Stripe platform, the median time to achieve an annualized revenue of $1 million is only 11.5 months - four months faster than the fastest - growing SaaS companies."
Stripe
Who is using this technology?
There is a saying that the popularity of generative artificial intelligence tools like ChatGPT and Gemini is spreading faster than almost any technology for which we have reliable data. The Federal Reserve Bank of St. Louis estimates that the spread of generative artificial intelligence is about twice as fast as that of the Internet.
Hartley et al.
In a recent large - scale survey on generative artificial intelligence - the 2025 paper "The Labor Market Effects of Generative Artificial Intelligence" - economists estimated that more than 50% of employees in the information service industry (that is, software companies) and management positions are already using this technology in their work. In contrast, in companies in traditional economic fields such as mining or fishing, few people use this technology. In addition, the popularity of artificial intelligence among college graduates is also much higher than that among people without a college education.
Do employees think that artificial intelligence has improved their productivity?
Yes. The landmark research proving that new artificial intelligence models can improve productivity comes from companies with highly repetitive work, such as call centers. But we are receiving more and more self - reported reports from employees saying that artificial intelligence has helped them save a lot of time. A surprising example is teaching. According to a recent Gallup survey, about 60% of primary school teachers said they had used artificial intelligence to prepare lessons, review teaching materials, create practice questions, or handle administrative work. Most teachers who use artificial intelligence said that this technology has improved their work; and teachers who use it frequently said that they can save 6 hours per week - which is equivalent to saving six weeks per school year. From a very optimistic perspective, it's like saying that artificial intelligence gives primary school teachers a month and a half of paid vacation every year.
Gallup
Perhaps the most optimistic indicator that artificial intelligence will help people improve work efficiency comes from the discovery of the non - profit artificial intelligence research institution METR: The duration of tasks that artificial intelligence agents can complete doubles every 7 months. In 2021, artificial intelligence could automatically complete a simple Google search in 10 seconds. Two years later, ChatGPT could search for facts on the Internet, a task that an ordinary person would take about 4 minutes to complete. Now, some models are performing coding tasks that a typical developer would take 50 minutes to complete. The researchers said: "Based on this trend, in less than a decade, we will see that artificial intelligence agents will be able to independently complete most software tasks that currently take humans days or weeks to complete."
So, is the conclusion that artificial intelligence will create a productivity boom that sweeps across the entire economy?
Not so fast. In fact, many employees may seriously overestimate the productivity improvement brought by artificial intelligence.
METR also conducted an in - depth study, asking experienced developers to use a popular artificial intelligence assistant for coding. After completing the task, the developers claimed that using artificial intelligence had increased their productivity by 20%. But the independent assessors in the study came to the opposite conclusion: Using artificial intelligence actually increased the task - completion time by about 20%. I don't want to speculate too much about the long - term significance of this study. But for now, I think this is a necessary warning for those bold claims that ChatGPT is about to replace tens of millions of entry - level white - collar jobs.
Tell me one more interesting thing about how artificial intelligence is changing the world
A new paper in the journal Science found that since the rise of large language models, there has been a huge shift in the field of academic writing. Data shows that in 2024, the frequency of the word "delves" was 2700% higher than the historical average. Analysis shows that in 2024, about one - seventh of the paper abstracts were processed by artificial intelligence.
