When Microsoft President Brad Smith speaks publicly about geopolitical competition, the technology industry listens. His latest remarks — a candid warning about Chinese government subsidies powering a new wave of artificial intelligence competition — have struck a nerve across Silicon Valley and Washington alike. At a moment when American AI companies are commanding trillion-dollar valuations and racing to build ever-larger models, Smith is urging the industry not to become complacent.
“I do think we always have to think about, maybe even worry a little bit about, Chinese subsidies,” Smith said, according to TechRadar. The Microsoft executive’s comments were pointed and deliberate, framing the competitive threat from China not as a distant possibility but as a present reality that American firms must prepare for now.
The Subsidy Machine Behind China’s AI Ambitions
Smith’s concerns are rooted in a well-documented pattern. The Chinese government has for years directed enormous state resources toward industries it considers strategically important — from solar panels and electric vehicles to semiconductors and, now, artificial intelligence. Beijing’s “New Generation Artificial Intelligence Development Plan,” first published in 2017, set explicit targets for China to become the world leader in AI by 2030. Provincial and municipal governments have since poured billions of yuan into AI research parks, startup incubators, and direct subsidies for companies developing foundation models.
The scale of this investment is difficult for private-sector competitors to match through market forces alone. Chinese AI firms such as Baidu, Alibaba, and newer entrants like DeepSeek have benefited from a combination of direct government funding, subsidized computing infrastructure, favorable regulatory treatment, and access to vast pools of domestic data. Smith’s warning suggests that Microsoft — which has invested more than $13 billion in OpenAI and is spending tens of billions more on data center construction — sees this state-backed competition as a structural challenge, not merely a cyclical one.
DeepSeek and the Shock of January 2025
The urgency behind Smith’s remarks was amplified earlier this year when DeepSeek, a Chinese AI startup, released models that appeared to rival the performance of leading American systems at a fraction of the reported cost. DeepSeek’s R1 model, in particular, sent tremors through financial markets in late January 2025, briefly wiping hundreds of billions of dollars off the market capitalizations of U.S. chip and AI companies. The episode forced a reckoning: the assumption that American dominance in AI was assured by superior capital and talent suddenly looked less certain.
DeepSeek’s apparent efficiency gains raised uncomfortable questions. Were Chinese researchers finding ways to train competitive models with fewer high-end chips — chips that U.S. export controls were supposed to keep out of their hands? Or were the reported training costs misleadingly low? Either way, the market’s reaction revealed deep anxiety about the durability of America’s AI lead. Smith’s comments can be read as a direct response to this anxiety, acknowledging that the competitive threat is real and that U.S. companies cannot afford to rest on their laurels.
Export Controls: A Necessary but Imperfect Shield
The U.S. government has not been idle. Since October 2022, the Biden administration — and now the Trump administration — has imposed increasingly stringent export controls on advanced semiconductors and chip-making equipment destined for China. The restrictions target Nvidia’s most powerful AI accelerators and the lithography machines made by the Netherlands’ ASML, among other technologies. The goal is to deny China access to the computational horsepower needed to train frontier AI models.
But Smith’s remarks implicitly acknowledge the limits of this approach. Export controls can slow a competitor; they cannot stop one indefinitely. Chinese firms have responded by stockpiling chips before restrictions took effect, developing workarounds, and accelerating domestic chip development. Huawei’s Ascend series of AI processors, while not yet matching Nvidia’s top-tier products, has shown steady improvement. Meanwhile, Chinese companies have demonstrated ingenuity in algorithmic efficiency — doing more with less — which is precisely what made DeepSeek’s results so alarming to Western observers.
The Case for American Industrial Policy
Smith’s warning carries an implicit policy argument: if China is subsidizing its AI champions, the United States may need to respond in kind. This is a position that would have been considered heterodox in mainstream American business circles a decade ago, but it has gained significant traction in recent years. The CHIPS and Science Act, signed into law in 2022, committed more than $50 billion in federal subsidies to domestic semiconductor manufacturing — a direct acknowledgment that market forces alone were insufficient to maintain U.S. competitiveness in a strategically vital industry.
Whether similar direct subsidies should flow to AI model development, data center construction, or energy infrastructure to power AI workloads is now an active debate in Washington. Microsoft itself has been a vocal advocate for permitting reforms that would accelerate data center construction and for investments in nuclear and other clean energy sources to meet the staggering power demands of AI training. Smith has previously argued that the U.S. needs to build AI infrastructure faster and at greater scale if it hopes to maintain its position. His latest comments about Chinese subsidies add a competitive dimension to that argument: speed matters not just for commercial reasons but for national security ones as well.
A Two-Front Competition: Technology and Talent
The competition with China extends beyond hardware and capital. It is also a contest for talent. China produces more STEM graduates annually than any other country, and its top AI researchers publish in the same elite journals and conferences as their American counterparts. For years, a significant number of Chinese-born researchers chose to build their careers in the United States, drawn by superior research institutions, higher salaries, and the open culture of American technology companies. But that flow has slowed in recent years, partly due to visa restrictions and partly due to improved opportunities at home.
American technology executives, including Smith, have repeatedly urged policymakers to maintain the country’s attractiveness to foreign-born talent. Immigration policy, in this framing, is AI policy. If the United States makes it harder for the world’s best researchers to work at American companies and universities, it effectively hands an advantage to competitors — including Chinese ones — who are eager to recruit them. This talent dimension is inseparable from the subsidy question: China’s state-backed AI push includes generous compensation packages and research funding designed to lure top scientists back from overseas.
What Microsoft’s Posture Reveals About the Industry’s Mood
It is notable that these warnings are coming from Microsoft’s president rather than from a think tank analyst or a government official. Microsoft is not a company prone to public hand-wringing. Its market capitalization exceeds $3 trillion. It has the deepest partnership in the industry with OpenAI, the company behind ChatGPT. It is spending more on capital expenditures than at any point in its history, with plans to invest more than $80 billion in AI-related infrastructure in fiscal year 2025 alone.
And yet Smith is telling the industry to worry. That signal carries weight precisely because of the source. When the president of the world’s most valuable public company says that Chinese subsidies represent a competitive threat, it reflects a sober internal assessment — not alarmism. It suggests that Microsoft’s own strategic planning accounts for a future in which Chinese AI companies, backed by state resources, compete aggressively not just in China but in global markets across Asia, Africa, Latin America, and even Europe.
The Global Market Is the Real Battleground
This global dimension is often underappreciated in American-centric discussions of the AI race. While the U.S. and China may each dominate their respective domestic markets, the contest for the rest of the world remains wide open. Chinese technology companies have a strong track record of competing effectively in developing markets, offering products at lower price points and with fewer strings attached than their American rivals. If Chinese AI models — subsidized to be cheaper or even free — gain traction in Southeast Asia, the Middle East, or sub-Saharan Africa, they could establish standards and dependencies that would be difficult to dislodge later.
Smith’s comments, viewed in this light, are not just about protecting Microsoft’s bottom line. They are about shaping the terms of a generational competition. The question is whether American policymakers and business leaders will heed the warning before the window of advantage narrows further. The subsidies are flowing. The models are improving. And the race, as Smith has made clear, is far from over.