Hedged Bets on the U.S.-China AI Race

The United States’ economic growth aspirations are tied to frontier AI’s promise of transformative productivity. To American stakeholders, these gains are expected to be imminent and justify historic levels of capital investment, shaping U.S. technology regulation, frontier lab support, and large-scale infrastructure buildout. China, by contrast, is deploying narrower investments across manufacturing and consumer-facing verticals, believing that targeted investments and fierce industry competition will diffuse productivity gains throughout the economy and avoid bubble risk.  

At the national level, China has adopted a three-pronged structure of sectoral investment in core industries, reliable open-source platforms, and in-depth integration of AI across the economy. The underlying theory is that AI models tailored to specific commercial demands will deliver meaningful economic benefits sooner than frontier models. Similarly, the lower cost of Chinese models will enable entry into developing economies, securing future ecosystems and revenue streams. By embedding AI into its leading industrial sectors, China could extract substantial productivity gains even if its frontier models and hardware lag. Moreover, reaping productivity gains sooner may provide early advantages, enabling further advances in model performance and AI infrastructure. 

China has emphasized this approach through state initiatives like the “AI+” initiative, while the White House’s AI Action Plan describes an artificial intelligence inflection point at the intersection of an “industrial revolution, an information revolution, and a renaissance.”  

Analysts have therefore described the United States and China as pursuing distinct theories of value in artificial intelligence. CEO of Relativity Space Eric Schmidt and Selina Xu argue that American tech firms are distracted by artificial general intelligence, while China prioritizes near-term applications. Lt. Gen. Jack Shanahan and Lawfare Senior Editor Kevin Frazier contend the United States ought to reorient its approach to more closely mirror China’s and instrumentalize America’s “formidable innovative spirit into practical applications and evidence-based impact.” Council on Foreign Relations President Mike Froman similarly questions if the United States is “racing toward the wrong finish line,” with Chinese initiatives strategically integrating AI as a “default layer of the industrial economy.” 

This framing captures a real difference in direction but obscures more fundamental questions about diffusion capacity, industrial policy design, and capital market depth. Both countries face interlinked vulnerabilities in their AI strategies. China’s industrial policy builds real productive capacity while suppressing firm-level valuations, a constraint partially offset by continued access to U.S. capital markets. American capital, meanwhile, remains heavily indexed to expectations of AI-driven productivity gains. By pursuing their respective AI aspirations, the United States and China mutually reinforce their systemic advantages and risks, complicating efforts on both sides to fully indigenize the AI stack and establish long-term economic leadership. 

Diffusion Advantages 

Despite China’s emphasis on diffusion and many hailing its success thus far, it is unclear that China actually has a diffusion advantage. Depending on the survey and the criteria applied, comparative data on AI adoption in the United States and China produce contradictory findings. 

• July 2024: A survey by SAS and Coleman Parkes Research found that Chinese organizations outpaced the United States, with 83 percent of organizations “using GenAI” versus 65 percent of American enterprises. The United States is marginally ahead in “fully implementing” AI into commercial practices.
• March 2025: A McKinsey survey indicates that AI use is similarly widespread in Greater China and North America overall (56 percent versus 57 percent). Respondents in Greater China are more likely to use AI regularly for work (25 percent versus 15 percent), respondents in North America are more likely to use AI for both professional and personal purposes (42 percent versus 31 percent).
• November 2025: According to Microsoft's AI Economy Institute, 15.4 percent of China's working-age population uses AI, versus 26.3 percent of the American population. Using this estimate as a baseline, that would imply 57.9 million working-age Americans are using AI and 150.4 million Chinese, a per-capita lead by the United States, but a 2.6:1 advantage in overall population to China.  

Where all three surveys align is that adoption is rapidly accelerating. Microsoft claims that AI’s diffusion is the fastest general-purpose technology “in history”, and much faster than adoption of the smartphone, radio, and internet. SAS and Coleman Parkes Research note an incline in AI utilization, with 86 percent of companies investing in AI between 2024 and 2025. The McKinsey survey, meanwhile, finds that the rate of commercial application has risen from 19 to 56 percent of surveyed organizations use AI in ‘Greater China’ and from 16 to 57 percent in ‘North America.’ 

There are domains where China’s advance is sizable and probable. While the surveys account for organizational and individual application of artificial intelligence, China’s deployment of industrial robotics is significantly greater than that of the United States. In 2024, China had more than five times as many factory robots in operation (2,027,200 to 393,700). On access to electricity, in recent years, China has generated more electricity and begun construction on more nuclear power plants than the rest of the world combined. In 2024, China generated an additional 249.95 terawatt-hours; the year prior, China produced 9.5 million gigawatt-hours in total — 31.7 percent of the global output and more than double the United States’ output. China’s 1.1 billion internet users also means that China’s applications and platforms have significant traffic, but that advantage may be diminished by the share of internet data — AI model fuel — that is in English.  

Where the United States stands out is in information technology. America has superiority in cloud computing and cloud adoption, broadband and fiber internet subscriptions. Google, Microsoft, and Amazon alone account for more than 50 percent of the global cloud infrastructure market share. China's “relatively low rate of cloud adoption” complicates efforts to streamline and scale upgrades and deployments. Together with advantages in semiconductor hardware and America’s deep capital markets, these cojoined features allow for significant and patient investments in data center construction; however, these investment levers also expose investments to being misallocated and companies to fluctuating valuations.  

Various measures of digital capabilities introduce contradictory theories of what diffusion is and how it materializes. The United States’ digital infrastructure may allow for more broad-based access to artificial intelligence, but that does not necessarily mean that access will translate to applications. China, on the other hand, may have more concentrated uses of training data in areas such as consumer platforms. Moreover, as a growing number of American companies rely on China’s open-weight AI models to meet business needs, direct comparisons of diffusion outcomes mischaracterize underlying diffusion capacity – attributing American diffusion success partly to adoption of Chinese models.  

One of the findings of McKinsey’s survey was that while 'Greater China' and 'North America' were at parity for commercial applications of artificial intelligence, respondents in 'Greater China' were using artificial intelligence at greater rates than their peers for non-commercial applications. These non-commercial uses also play into data accumulation that could translate into commercial capabilities. At scale – and as diffusion of artificial intelligence continues to surge across geographies – such a dynamic has the potential to yield AI advances by volume in narrow disciplines alone.  

Industrial Policy and Capital Markets 

The divergence between American capital markets and Chinese state-led financing creates distinct economic dynamics.  

Leaning into the frontier of American innovation has yielded dividends in the past, but the risk of betting on AGI is significant. It means the United States is increasingly optimized for a particular pathway — one that depends on high-end chips, hyperscale data centers, and continued model scaling — rather than an application-first growth model. 

In the United States, capital markets allow for companies like OpenAI to attain high valuations despite limited revenue, thereby facilitating the continued growth of the U.S. economy in the data buildout. However, capital can just as easily flow in the opposite direction. Firms can be overvalued, market corrections can cascade, and contagion effects can leave sectors spiraling.  

China is renowned for making targeted industrial-policy bets in critical industries, including electric vehicles, batteries, photovoltaics, and robotics. China's industrial development has led China to compete with “advanced economies’” export goods in third markets and reduce reliance on foreign markets for frontier technologies and critical supply chain inputs. While the Chinese government has acknowledged and begun investigations of deflationary competition in "excessively subsid[ized]” sectors (according to the State Council), driven down costs entrench China as a supplier of choice in emerging markets – impacting China’s ability to accelerate the adoption of technologies.  

Importantly, China’s industrial policy often targets entire sectors rather than individual firms for state subsidies, thereby promoting intra-industry competition and enhancing the global competitiveness of Chinese industries. This contest for government support – 80 percent of which occurs at the local level – reduces the profitability of firms. In the case of electric vehicles, from 2017 to 2024, total profits dropped by 33 percent. China’s state subsidies also create investor uncertainty, as the duration of sectoral support varies dramatically – 34 percent of industries are supported for one year, whereas 24.5 percent are supported for over a decade. This uncertainty increases the cost of capital and has resulted in investors undervaluing stocks supported by government subsidies.  

Overlaying the market distortions of China’s industrial policy is state support for the stock exchange. In anticipation of China’s September 2025 stimulus package, the Hong Kong, Shenzhen, and Shanghai composites all grew in double-digits from September 16-30. This support reduced the reserve requirement ratio, lowered mortgage rates, and established a swap program for securities, funds, and insurance companies and a re-lending facility for listed companies to purchase by-backs. Since then, all three exchanges have seen stable growth in spite of trade uncertainty.  

In addition to direct state support of financial markets, China’s stock market has benefited from four compounding factors. First, the erosion of alternative assets like real estate has consolidated stocks as a stable asset class. Second, investments in technology and in manufacturing are being used to offset real estate’s downturn. Third, the People's Bank of China has injected liquidity into the market. And forth, a flurry of IPOs, including 114 in Hong Kong raising $37.22 billion and high-profile listings in Shanghai like Moore Threads and in Shenzhen with CATL, have increased interest in Chinese stocks.  

Despite this growth and state support in 2025, America’s premier technology companies maintain higher valuations and greater price-earnings ratios. Comparing select companies in the “Magnificent Seven” (Nvidia, Apple, Microsoft, Meta, Alphabet, and Amazon) to China's BATX (Baidu, Alibaba, Tencent, and Xiaomi), the Chinese firms have grown significantly faster from Q3 2024 to Q3 2025 – nearly 3.5 times higher growth in their market shares compared to their American counterparts during this period.  

The contrast further extends when examining market capitalization differences between American and Chinese tech companies. As of January 2, 2026, the Magnificent Seven company with the lowest valuation is Tesla at $1.46 trillion. The BATX companies put together have a valuation of $1.22 trillion, a lower combined valuation than Tesla, 30.3 percent of the valuation of Apple alone, and 26.9 percent of NVIDIA.

Net Assessment 

The United States and China face different vulnerabilities borne out of their unique approaches to AI. The question is whether either country’s approaches will achieve sustained and compounding economic effects.  

America’s hope that AGI’s capabilities benefit broad swaths of the American economy poses serious structural risks. This carries across the American economy – discounting information processing equipment & software (data centers), GDP growth in the first half of 2025 was 0.1 percent. Overvalued firms could cause abrupt market shifts and downturns, reducing investor confidence.  

To this end, the United States’ continued global AI leadership status is contingent on capital markets retaining their resiliency, growing American diffusion, and frontier model advantages. Moreover, the United States’ industrial policy moves, such as the U.S. government stake in Intel, indicate an affinity for picking winning firms as opposed to winning sectors, which could limit competition and open the door to capital misallocation.  

China faces its own set of challenges. If China’s expected industries of the future materialize too slowly and IT constraints drag on broad-based diffusion, a misallocation of capital could further constrain tight budgets. However, investments in the infrastructure that bolsters diffusion would amplify existing access to data and existing AI integration in consumer-based applications. 

Ultimately, both countries are pursuing strategies attuned to their unique domestic systems, but these strategies are incomplete. Chinese technology firms remain deeply integrated into U.S. financial and technological ecosystems, benefiting from access to American capital markets and listings. According to the U.S.-China Commission, 286 Chinese companies are listed on U.S. stock exchanges, with 48 IPOs in the United States since January 2024. Of the BATX firms, Alibaba is sold on the New York Stock Exchange and Baidu on NASDAQ, whereas Tencent and Xiaomi can be purchased over the counter by American investors.  

Beyond financing, American companies – and companies from third countries – benefit from the cost-competitive nature of Chinese AI large language models and their open-weight flexibility. Airbnb, for instance, relies on the accessibility of Alibaba’s Qwen model for its customer service interface. America losing access to China’s models and China losing America’s systems of finance and critical IT infrastructure would imperil the sustainability of these complementary stacks.  

Without incorporating elements from the other's model, the United States and China will face significant shortcomings in capital, technology, and adoption as they attempt to vertically integrate AI. Whether open versus closed source or state versus private capital, the United States and China play distinct roles in building and setting the standards for global innovation. The question is not who wins the race, but whether either can finish without the other's capabilities. 

Richard Gray is a program coordinator and research assistant with the Economics Program and Scholl Chair in International Business at the Center for Strategic and International Studies (CSIS) in Washington, D.C. Michael H. Gary is a former intern with the Economics Program and Scholl Chair in International Business at CSIS. 

For more analysis on how states interact in a transitioning global economy, check out our blog series, Charting Geoeconomics.