The AI arms race is getting more heated, and fears over the AI economic bubble are growing. Massive investments are going into AI companies through circular financing, but profitability seems like a distant dream.
We asked ChatGPT in deep-research mode to analyze the AI arms race between the US companies, Chinese companies, and Mistral (Europe) with a focus on economics.
The AI Arms Race: Can the Economics Keep Up?
The global AI race is no longer just a competition over who builds the most powerful model. It is becoming a competition over who can turn enormous investments in chips, data centers, energy, software, and AI talent into lasting economic returns.
The United States currently dominates private AI investment, while China has rapidly closed the frontier-model performance gap and Europe is attempting to build its own position through companies such as Mistral. The important economic question is therefore not simply who wins the AI race. It is who captures the value created by it—and whether the returns will justify the capital being invested.
The AI Race Requires Extraordinary Capital
Global corporate AI investment more than doubled in 2025. Private investment grew 127.5%, while generative AI investment grew by more than 200%. U.S. private AI investment reached $285.9 billion, compared with $12.4 billion in China.
That does not mean the United States spends 23 times more on AI overall. Stanford notes that private-investment figures understate China’s total AI spending because of substantial government-backed funding.
The infrastructure requirements are enormous as well. AI companies need increasingly expensive computing capacity, while major technology companies continue expanding data centers and other infrastructure.
Stanford’s 2026 AI Index reports that AI-company revenue is rising rapidly at the same time that compute costs and infrastructure spending are reaching record levels.
The Real Economic Problem: Revenue vs. Capital
The central economic question is simple: can AI revenue and productivity gains grow faster than the cost of building and operating AI?
There are encouraging signs. Stanford estimates that U.S. consumer surplus from generative AI reached approximately $172 billion annually by early 2026. Studies also report productivity gains in areas such as customer support, software development, and marketing.
But AI infrastructure is capital-intensive. Google alone reported more than $150 billion in annual capital expenditure in 2025.
The result is an unusual situation: AI can create enormous economic value while individual companies may still struggle to earn returns large enough to justify the amount of capital being deployed.
Open-Weight AI Could Change the Economics
One of the biggest variables is the rise of increasingly capable open-weight models. If businesses can obtain powerful models without relying entirely on a single proprietary provider, competition can push the price of AI services downward.
That would not necessarily destroy the AI industry. Instead, it could change where the money is made.
As model intelligence becomes more competitive and potentially more commoditized, economic value could shift toward chips, cloud infrastructure, electricity, data centers, proprietary data, specialized applications, and distribution.
In other words, open-weight AI could be extremely beneficial to consumers and businesses while simultaneously making it harder for individual model providers to maintain exceptionally high margins.
U.S., China and Mistral: Different Economic Positions
The United States has major advantages in private capital and computing infrastructure. Stanford reports that the U.S. hosts 5,427 data centers—more than ten times the number in any other country.
China’s advantages are different. It combines industrial capacity, government-backed investment, large-scale deployment, and a rapidly developing AI ecosystem. The frontier-model performance gap has also become extremely small: as of March 2026, Stanford measured the leading U.S. model at only a 2.7% advantage over the leading Chinese model.
Europe is pursuing another strategy. French AI company Mistral raised €3 billion in September 2026 at a valuation of approximately €21 billion, according to Reuters.
The competition therefore involves different economic models: U.S. private capital and infrastructure, China’s combination of industrial scale and state support, and Europe’s attempt to establish strategically important AI companies of its own.
Is the AI Boom a Bubble?
It is too early to say that AI as a whole is a bubble. AI is already producing measurable productivity gains and substantial consumer value.
However, real technology does not guarantee that every investment will be profitable.
The Bank for International Settlements has warned about financial-stability risks from the AI boom, particularly because enormous infrastructure investments are being accompanied by debt and other forms of financing.
The more accurate description is therefore not “AI is a bubble,” but parts of the AI investment boom could become overvalued or overbuilt if expected returns fail to materialize.
The Biggest Economic Risk May Be Successful AI
The most interesting economic risk is not necessarily that AI fails.
It is that AI succeeds so well that intelligence becomes cheap.
If increasingly capable models become widely available through intense competition and open-weight alternatives, the price of AI services could fall faster than investors expect.
Consumers and businesses would benefit from cheaper AI, but companies that spent enormous amounts developing the underlying technology could struggle to earn returns matching their valuations.
AI could therefore be enormously beneficial to the economy while still producing disappointing returns for some investors.
What Could Cause an AI Investment Correction?
- AI revenue fails to grow quickly enough to justify infrastructure spending.
- Model prices fall faster than expected because of competition and open-weight alternatives.
- Data centers become underutilized after companies build more capacity than demand requires.
- Higher financing costs reduce the profitability of infrastructure projects.
- Investors begin demanding conventional returns rather than simply betting on future growth.
- AI productivity gains arrive more slowly than current valuations assume.
None of these outcomes is inevitable. But as investment grows, utilization, cash flow, and return on investment become increasingly important.
Why an AI Correction Could Matter to the U.S. Economy
An AI correction would not necessarily mean the collapse of the technology itself. AI infrastructure is becoming connected to construction, energy, semiconductors, cloud computing, finance, and employment.
If AI generates enough productivity growth, those gains could help justify today’s investment. If expected returns fall sharply, however, companies could reduce capital expenditure at the same time that technology valuations decline.
That is the financial risk highlighted by the BIS warning: high valuations, concentrated investment, and debt financing could amplify the effects of an eventual correction.
How the United States Could Reduce the Economic Risk
- Focus on productivity: AI investment should produce measurable improvements in businesses and government.
- Avoid excessive leverage: Infrastructure expansion should not depend on permanently rising valuations or cheap credit.
- Maintain semiconductor leadership: Chips remain one of the most strategically important parts of the AI value chain.
- Build flexible infrastructure: Data centers and power infrastructure should remain useful even if individual AI companies or models lose market share.
- Support competition: A competitive AI ecosystem can reduce costs and prevent excessive dependence on a small number of providers.
- Prepare workers for transition: Productivity gains are economically valuable only if workers and businesses can adapt to changing job requirements.
Conclusion: The AI Arms Race Is Really an Economic Race
The AI arms race is often described as a contest to build the most powerful model. Economically, that is only part of the story.
The larger competition is over who owns the infrastructure, who captures the productivity gains, who controls the supply chain, and who earns a return on the enormous capital being invested.
The United States has enormous advantages in private investment and computing infrastructure. China is a serious competitor with industrial scale, state-backed investment, and rapidly improving models. Open-weight models could further change the economics by making AI intelligence cheaper and more widely available.
The biggest mistake would be assuming that technological leadership automatically guarantees economic profits. The AI industry can transform the global economy and still experience a major investment correction if capital spending gets ahead of sustainable returns.
The goal should not be to prevent AI from becoming cheap. It should be to make sure the United States captures enough of the productivity and economic value created when it does.
