The United States is betting the house on winning the artificial intelligence race with China.

Consider the U.S. economy, which is increasingly becoming a highly leveraged bet on the promise of this new technology. AI is the primary driver of investments, stock market gains, and growth in the United States. As The Wall Street Journal recently noted, AI-related investment is driving most of the growth in U.S. GDP. Spending on AI data centers accounts for half of all business investment in the country. AI companies made up more than 40 percent of the value of U.S. equity markets beforeSpaceX went public in June. If OpenAI and Anthropic follow suit later this summer with expected trillion-dollar-plus IPOs, AI companies will account for more than half of the value of the stock market. Indeed, of all the S&P 500’s gains in 2026, 85 percent have come from AI companies.

Or consider the growing national security-AI industrial complex. The U.S. government is becoming a major customer of leading AI companies, deepening entanglements with the Defense Department, CIA, National Security Agency, and others. An odd alliance that includes U.S. President Donald Trump, socialist independent Sen. Bernie Sanders, and OpenAI CEO Sam Altman is seeking ways for the U.S. government to become a major stakeholder of these companies.

As David Sacks, who recently stepped down from his position as the Trump administration’s AI czar, pointed out in late July, the United States’ champions—OpenAI and Anthropic—are now resorting to a “regulatory capture strategy” focused on persuading the U.S. government to adopt regulations and laws that will handicap their competition, both foreign and domestic.

Three powerful forces are fueling this race for the brightest shiny object that current observers have ever seen.

First, AI companies are chasing what they believe will be the biggest-ever pot of gold at the end of the rainbow. For the first time in history, companies and individuals are talking not in billions but trillions. Their leaders stand to gain not only heretofore unimagined fortunes but also fame and influence. Note, for example, the June G-7 meeting in France, at which AI CEOs took seats around the same table as leaders of the world’s advanced economies to discuss the road ahead. Is there anything they would not say or do to win?

Second, U.S. Treasury Secretary Scott Bessent and other economic policymakers are acutely aware of the huge systemic risk posed by the nation’s unsustainable annual fiscal deficits (now at nearly 6 percent of GDP) and government debt (the annual interest payments on which now exceed the defense budget). An AI miracle that produces unprecedented increases in productivity and growth offers their best hope for avoiding an economic catastrophe.

Finally, the national security strategist community has declared that AI is existential, in that whoever wins the AI race will rule the world. They have accepted the technology titans’ proposition that whichever nation is first to build a super-intelligent “AGI,” artificial general intelligence, will have a decisive, irreversible advantage—since an agent smarter than all the Nobel laureates in the world will then begin teaching itself. In former U.S. Secretary of State Condoleezza Rice’s summary: This is a race that “we absolutely have to win.”

Confronting this blizzard of claims and counterclaims about a bedazzling technology, four questions demand our attention.

First, is the course that the United States has chosen prudent? Another way to think about this is, would a financial investor with fiduciary responsibility place this large a bet on a single sector? Second, what could go wrong? Could the eye-popping increases in investment and stock prices be another bubble—analogous to the financial wizardry that created the real-estate bubble that burst in the Great Recession of 2008 or the dot-com bubble in 2000? Third, why is the only other AI superpower—China—making such a radically different bet?

And finally, if the miracle is delayed or never happens, what will the likely consequences be—not just for the U.S. economy but also for national security?

---

However attractive a stock such as Elon Musk’s SpaceX or a sector such as AI, successful investment firms know better than to place most of their eggs in one basket. As Bridgewater founder Ray Dalio put it, piling into “one new sector that is highly volatile and risky” is “unsophisticated.”

Nonetheless, in the first half of 2026, more than 80 percent of global venture capital funding has flowed into AI start-ups. Google, Microsoft, Meta, and Amazon have announced plans to cumulatively spend more than $1 trillion on AI in 2025 and 2026—relying on debt markets to finance their ambitions. Moreover, in the drive for more computing power, or compute, a growing number of circular transactions have frontier AI companies contracting with chipmakers that invest the funds from their purchases back into their customers, further inflating share prices.

How could this go wrong? Ask your AI assistant, and it will provide a list of a dozen ways. In his last article published before he died in 2023, Henry Kissinger and I warned about the possibility of an AI Chernobyl or AI 9/11, in which a powerful AI assists a malign actor in producing a bioweapon or similar device that could kill hundreds of thousands of people, disrupt the financial markets, or shut down the electricity grid.

In an open letter published in June, all the AI titans—Altman, Anthropic’s Dario Amodei, Google DeepMind’s Demis Hassabis, Microsoft’s Mustafa Suleyman, and Meta’s Alexandr Wang—highlighted the risk of an AI-enabled killer pathogen and called for serious safeguards. Beyond that, Amodei has offered his view that there is a “25 percent chance that AI could lead to catastrophe for the human race.”

The more immediate risk is a collapse of what increasingly looks to many analysts like a classic bubble. In its monthly global fund manager survey for July, Bank of America ranked a bursting of the AI bubble as the key risk to financial markets. Financial institutions such as the Bank for International Settlements are now warning that valuations for AI companies exceed earnings by an amount not seen since 1929.

Drawing lessons from his reading of the history, Michael Burry, one of the investors portrayed in The Big Short, has cautioned: “We are getting into that rare air, so extreme that the consequences will be unavoidable, no matter where one hides.”

Recent market corrections reflect concern that the hyperscalers are building more compute than they require. For example, Meta, SpaceX, and others are actively exploring renting capacity to other companies. As Forbesconcluded from a recent analysis by Bain and Company, “tech companies will need to find $2 trillion in new AI revenue to make a profit on the cost of deployment.” Last week, DeepSeek released a powerful upgrade to its V4 model that can complete complex coding tasks at a level comparable to Anthropic’s top-tier Claude Opus 4.8. But as Axios notes: “DeepSeek charges about 28 cents for the same amount of output that costs $25 on Opus 4.8—a 99% discount.” Coming on top of OpenAI’s announcement that it was cutting the price of its fastest model for high volume tasks by 80 percent only three weeks after its launch, the release is igniting debate about whether AI products are in effect becoming “commoditized”—eroding the frontier labs’ pricing power.

At the same time, the companies are facing a rising tide of resistance to their deployment plans. The governor of New York state has imposed a one-year ban on new AI data centers. Polls now show that 7 in 10 Americans oppose the construction of data centers in their neighborhoods. And 80 percent of Americans are concerned about AI. Their fears include the possibility that AI will take their jobs, harm their children’s education, and make truth inseparable from fiction online.

---

The fact that the United States’s only credible competitor in AI—China—is taking such a strikingly different path should make Americans pause and think. Beijing shares with Washington the view that AI is of critical national importance. On the day in 2016 when DeepMind’s AI beat the world’s Go champion, President Xi Jinping declared that this was a technology in which China could and should achieve global supremacy.

But in contrast to the frenzy in Silicon Valley to install evermore compute in pursuit of a world-changing AGI, Beijing has been more relaxed. Why is a good question. Money is certainly not the obstacle—since China has a sovereign wealth fund of $3.6 trillion and recorded a trade surplus of $1.2 trillion last year. U.S. constraints have limited Chinese access to the most advanced semiconductors and the equipment to produce them—though Chinese companies have demonstrated remarkable skill at finding holes in these fences and other work arounds.

While one can never know what one doesn’t know, there is no evidence of a hidden, state-directed Manhattan Project for AI in China. China’s latest five-year plan identifies AI as a top priority. But its “AI Plus” program focuses primarily on strengthening the integration of AI across society to increase productivity, multiply scientific breakthroughs, streamline governance, and improve everyday life.

Indeed, China’s strategy for becoming the world leader in AI reflects its basic strategy for achieving dominance in the production of everything. In essence, as we have seen in its domination of solar panels, fast trains, 5G, consumer electronics, and now increasingly electric cars, China begins by offering a product almost as good as the alternative but at a far more affordable price.

Unlike their U.S. counterparts, China’s large corporate champions in the AI race—Baidu, Alibaba, and Tencent—are not aggressively building massive data centers. In 2025, the three Chinese champions spent roughly one-tenth of what the U.S. hyperscalers spent on AI. Yet they are releasing models that rival the performance of those produced by the United States’ AI leaders.

Moreover, an array of small start-ups is producing startling results. This month’s launch of Kimi K3 by Moonshot produced what some in Silicon Valley called a “moon quake”—upending the frequently stated conviction that the United States leads China by at least six to 12 months. Kimi’s results on AI benchmark tests are comparable to those of Anthropic’s and OpenAI’s most advanced models.

Rather than offering hundred-million-dollar signing bonuses to attract “superintelligent” AI stars, DeepSeek demonstrated that with just 200 engineers earning around $150,000 a year, it could produce a model that could perform almost as well as U.S. competitors—and one that it offers to users around the world for free. Moonshot, Zhipu, and China’s other “AI tigers” are following a similar playbook. Nvidia CEO Jensen Huang has estimated that half of all the AI super talent in the world is in China. So, as one Chinese AI leader put it, in the rivalry between brain cells and GPUs, China is betting on the former.

The United States’ frontier labs keep their model weights and training data proprietary in order to sell their services for revenue. China’s AI companies have instead chosen to open-source their models, making them available to anyone in the world essentially for free. China’s approach is analogous to Google’s in its competition with Apple to develop the leading smartphone operating system. While Apple’s iOS is a premium product available only to those who buy iPhones, Android can be installed on many devices free of charge, encouraging adoption and enabling app development within the ecosystem. Today, Android is used in 70 percent of smartphones in the world and offers twice as many apps as Apple’s iOS.

Similarly, users around the world are increasingly selecting Chinese AI models over U.S. models. Chinese models now handle almost two-thirds of the world’s AI use, measured in tokens. On the leaderboard of AI models that process the most tokens, the five most popular are Chinese. U.S. companies such as Airbnb that employ Chinese models rather than American ones report that their AI costs are roughly one-tenth that of the U.S. alternative.

---

In the year ahead, if the United States’s AI miracle fails to materialize, the economic consequences are likely to exceed those of the real-estate bubble of 2008 and the dot-com bubble of 2000. Indeed, we could see another 1929. After the 2008 crash, U.S. GDP fell by 4.3 percent. During the Great Depression, it dropped nearly 30 percent. The downturn that began in 2000 erased $5 trillion of Americans’ wealth as the Nasdaq lost almost 80 percent of its value—and it took 15 years to recover to pre-recession levels.

Beyond economics, the national security consequences of a U.S. AI bubble that bursts are even more disturbing to contemplate. The collapse of wealth, fall in government revenues, and spike in demand for social programs would upend all plans for national security and defense investments.

The impact of this event—the visible failure of the world’s technology leader in what it had declared was the most consequential technology race of all time—on confidence in the United States both at home and abroad would be even graver. Remember the crash of 1929. It led to the Great Depression of the 1930s and created the conditions in which fascists took power in Germany and Italy. It ended in World War II.

In 2008, the failure of mortgage-backed securities created by Wall Street financial engineers caused the global Great Recession, which was on the precipice of becoming another Great Depression. As Hank Paulson, who served at that time as secretary of the U.S. Treasury and the designated lead for averting another depression, has explained, catastrophe was avoided only by a coordinated stimulus in which the U.S. and Chinese governments acted together. But as Paulson also noted, this became, for Xi and his colleagues, a decisive learning moment. Previously, they had assumed that Wall Street’s supposed masters of the universe knew what they were doing and could be relied on to manage the global financial system prudently. After seeing their reckless risk-taking fail, Xi and his team chose a different course.

How would China respond to the chaos in the United States and the world that would accompany an AI collapse?