Democrats' Midterm Win Reshapes the US–China AI Race

Democrats have taken control of both the Senate and the House in the US midterm elections, and analysts say the contest with China for artificial intelligence supremacy is entering a new phase. The immediate question is whether Democratic regulation of AI model development and data centers will cool America's infrastructure investment boom — and whether Beijing can use the opening to close the gap.

Not everyone thinks China benefits. Some observers argue a Democratic Congress would tighten controls on Chinese companies' access to advanced US-made semiconductors and overseas computing resources, while the United States keeps a firm lead in both the performance and profitability of frontier AI models. The stakes are straightforward: whoever builds the better models and the computing power behind them shapes the next generation of the global economy.

Why some analysts see an opening for China

Speaking to Britain's Financial Times on November 9, Meng Qinji, a global governance researcher at China's Tsinghua University, said the Democrats' midterm victory could offer an opportunity to narrow the AI gap between the two countries. The logic is mechanical: if Democrats, who now hold majorities in both chambers, impose stronger rules on data center construction and AI companies, the US infrastructure investment boom loses steam.

Gary Ng, a senior economist at French investment bank Natixis, framed it the same way. "China will see a slowdown in the development of the US AI industry as an opportunity to catch up with the United States," he said, adding: "China has favorable conditions, including limited resistance to data center construction, abundant electricity supplies and a well-developed industrial ecosystem."

Legislative groundwork is already moving. Democratic lawmakers plan to introduce a series of bills when the new Congress convenes in January, covering safety testing for AI models and regulation of data center development. The contrast with the outgoing Republican-led Congress is sharp: Republicans prioritized technological development and infrastructure expansion and refrained from directly restricting business activity, while Democrats emphasize consumer protection and developer accountability.

The specific priorities under discussion include:

• Easing the burden data centers place on household electricity bills
• Mandating independent safety evaluations of advanced AI models
• Clarifying developers' legal liability for harm caused by AI systems

The takeaway: for the catch-up camp, regulation is the variable that could slow US capital spending first.

What about tighter export controls?

Other observers read the election result as bad news for Beijing. Dan Wang, an analyst at political risk consultancy Eurasia Group, told the FT: "The Democrats' approach to China differs considerably from that of US President Donald Trump," and "the shift in the balance of power in Congress could actually deal a bigger blow to China's AI industry."

The mechanism is export controls. Democrats could impose stricter limits on Chinese companies' access to advanced US-made semiconductors and AI technology on national security grounds. Major Chinese tech companies — ByteDance, Alibaba and Tencent in particular — are named as potential targets. These companies are known to have used computing resources at data centers in Southeast Asia and elsewhere to circumvent US semiconductor export controls, exploiting a regulatory gap: US restrictions focus on physical equipment entering China, so remotely using US-made products installed in third-country data centers falls outside the rules.

There is an unused enforcement lever. According to the Carnegie Endowment for International Peace, the US Commerce Department's Bureau of Industry and Security (BIS) has the authority to restrict such access through conditions attached to export licenses for the equipment — but has yet to exercise that authority actively. In plain terms: the rules already allow a crackdown; nobody has pulled the trigger.

How big is the capability gap, really?

Market share data suggests China's reach is wide but shallow. Chinese companies have expanded their influence in the global AI market through low prices and open models. According to the "2026 State of Enterprise AI" report released November 4 by US venture capital firm Menlo Ventures, Chinese AI models accounted for around 89% of US companies' spending on open AI models. The catch: open models represented just 4% of total enterprise AI spending. In the paid enterprise market, Anthropic and OpenAI held an overwhelming lead, with shares of 55% and 31%, respectively.

The profitability gap is equally clear. Last month, AI research organization Epoch AI estimated that the combined AI-related revenue of six major Chinese AI companies, including DeepSeek and Zhipu AI, amounted to only about 10% of the combined revenue of OpenAI and Anthropic. Open models — the strategy Chinese companies emphasize — can be downloaded and run on companies' own servers or accessed through third-party services, which limits the revenue captured by original developers. Epoch AI noted that unless Chinese companies overcome this limitation, they may struggle to recoup their enormous research and development costs, weakening their capacity to invest in next-generation AI models.

The pattern in one line: China dominates a market slice that is barely monetized, while US labs collect most of the actual enterprise money.

Where do the benchmarks and the compute stand?

China has also been slow to catch up in measured model performance. Under the new "Intelligence Index v5" evaluation metric released October 27 by AI benchmarking organization Artificial Analysis, Moonshot AI's "Kimi K3.1 Pro" scored 46 points — well below the 58 points earned by Anthropic's "Claude Opus 5.5." The gap between the best-performing US model and open models, which had narrowed to 3 points under the previous framework, widened to 12 points after the methodology was revised. DeepSeek's "V4.1 Flash" and Zhipu AI's "GLM-5.3," both with publicly available model weights, scored 32 points each.

The computing infrastructure numbers are even more lopsided. According to Epoch AI's analysis, 72% of the AI computing resources newly added worldwide from 2022 through June of this year were concentrated in the United States, while China's share was only about 10% — including domestically produced semiconductors. A large proportion of the AI semiconductors China has secured are either reduced-performance products designed to comply with US export controls or domestic chips, leaving China at a pronounced disadvantage in actual computing performance relative to raw chip counts.

A summary of the constraints on China's catch-up, per Menlo Ventures, Epoch AI and Artificial Analysis:

• Market dominance: 89% of open-model spending, but open models are just 4% of total enterprise AI spending
• Profitability: Six major Chinese AI companies' combined revenue is about 10% of OpenAI's and Anthropic's combined revenue
• Technological capability: The score gap between China's best model and the leading US model widened after Artificial Analysis updated its evaluation metrics
• Computing infrastructure: The US accounts for 72% of newly added AI computing resources; China, about 10%

Will Big Tech money soften the blow?

Some analysts predict Democrats will adjust policy toward stronger safety testing and consumer protection obligations rather than impose stringent regulations that depress AI capital expenditure across the board. The reason is money. An analysis by global consultancy D&A Advisory of 63 fundraising organizations registered with the US Federal Election Commission found that 59.4% of the USD 1.16 million in contributions from the "Magnificent Seven" — Nvidia, Google, Microsoft, Amazon, Meta, Apple and Tesla — went to Democrats. Google, Microsoft, Amazon and Meta are pursuing massive AI investments, including their own AI models, data center construction and advanced semiconductor procurement, and could suffer significant damage from tighter regulation.

Regulation of data center construction may also shift investment rather than stop it. Opposition to data center construction is spreading in large US cities and affluent suburbs, where Democratic support is relatively strong, while small towns in Republican-leaning areas compete to attract such facilities. According to The Washington Post, residents of Harwood, North Dakota, rejected a ballot measure on November 3 that would have imposed additional review procedures on large industrial facilities including data centers — 105 votes in favor, 325 against. Harwood, population 845, expects its annual property tax revenue to rise to USD 1.65 million after the data center is completed, roughly six times its 2024 level.

That points to a middle path: by steering data centers toward small towns, Democrats could address core supporters' concerns about electricity bills and the environment while keeping the AI industry's growth momentum intact. Whether Beijing can exploit the next two years may depend less on US regulation than on whether China's open-model strategy ever starts paying for itself.