China and the US Are Running Different AI Races
Shaped by a different economic environment, China’s AI startups are optimizing for different customers than their US counterparts — and seeing faster industrial adoption.
Feb 13, 2026
Last month, as three Chinese AI startups went public within days of each other, Hong Kong briefly became a scoreboard for emerging companies in the industry. On January 2, AI chip designer Shanghai Biren Technology listed in Hong Kong and raised $5.58 billion Hong Kong dollars ($717 million). About a week later, model developers Zhipu AI and MiniMax followed, raising HK$4.35 billion ($558 million) and HK$4.8 billion ($619 million), respectively.
Those listings matter less as a market spectacle than as a strategy signal, and the strategy differs noticeably from that of companies across the Pacific. US startups build around abundance: raise huge capital, buy time, push the frontier. OpenAI’s Stargate plan, for example, aims to invest $500 billion over four years in AI infrastructure. Meanwhile, Chinese startups adapt to different constraints: frontier training infrastructure is scarcer, so momentum comes from efficiency, targeted deployment, and market selection.
As AI moves from demos to production, the binding question shifts: what does it cost to deliver useful work reliably, and who will pay? Under different economic pressures, Chinese and US companies are opting for different go-to-market strategies.
US AI startups attract more private investment than those in China. The divergence starts with money. In 2024, US AI startups received approximately $109.1 billion in private investment, while Chinese AI startups received roughly $9.3 billion — a ratio of 12 to 1. Chinese AI funding fell 38% year-over-year in 2023. By 2025, foreign investors provided only 10% of Chinese tech startup funding, with pure-play AI companies seeing foreign participation below 12%.
Chinese government funding only partly makes up for a lack of private investment. One might think the Chinese government would fill the gap, and that is partially true. China’s Big Fund III has registered capital of 344 billion yuan ($47.5 billion), and local governments have introduced “computing vouchers” that subsidize up to 80% of cloud computing costs. But government capital operates differently. Georgetown’s Center for Security and Emerging Technology (CSET) finds Chinese guidance funds have a historical disbursement rate of roughly 50%, meaning that much of the committed capital remains idle. Government programs mitigate but do not eliminate the constraint.
Chinese tech giants are pursuing scale, but smaller players can carve out niches. A second reason to believe that lower capital will have limited effects is the idea that China’s AI development will simply concentrate in large corporations that can afford scale. With ByteDance investing an estimated $21 billion in AI infrastructure in 2025 and Alibaba announcing a 380 billion yuan ($53 billion) three-year plan in February 2025, it is clear that these giants aim to dominate.
/inline-pitch-cta
But China’s market is vast and stratified: demand varies dramatically across regions and industries, from coastal manufacturing hubs to inland services, and from enterprise automation to consumer entertainment. This fragmentation creates defensible niches that no single player can fill. The independent Hong Kong listings of Zhipu and MiniMax reflect a market structure where scale alone does not guarantee capture — startups can win specific segments even as giants pursue the whole.
US startups can earn revenue by selling access to AI models. In addition to the capital gap, the question of who is paying for AI reveals sharper comparisons. In the US, startups sell “capability as product”: subscriptions with clear price points. ChatGPT Plus launched at $20 per month. GitHub Copilot starts at $10 per month. In high-salary markets, subscribers justify small monthly fees by arguing that these tools save time.
In China, consumers expect AI model access to be provided for free. Chinese consumer AI operates on a fundamentally different model. Following DeepSeek’s free-tier disruption, Baidu made Ernie 4.0 completely free for consumers in April 2025, abandoning its subscription experiment. ByteDance’s Doubao has been free since its launch, monetizing through API revenue and ecosystem integration rather than consumer subscriptions. The price anchor in China is not $7 or $5 — it is $0.
Chinese startups provide free AI access and monetize by other means. This reflects strategic logic, not charity. For Baidu and ByteDance, AI functions as a traffic platform, the next-generation search engine or super-app, where user acquisition matters more than subscription revenue. Monetization happens elsewhere: enterprise API calls, cloud service bundling, and advertising conversion. Lower per-capita income and the absence of established software subscription habits make this approach more viable than transplanting US-style pricing.
Chinese firms spend far less on software than US firms do. On an economy-wide basis, the World Intellectual Property Organization WIPO (drawing on S&P Global Market Intelligence) estimates 2024 software spending of $368.52 billion in the US compared with $61.8 billion in China (a ratio of roughly 6 to 1). Normalizing those totals by official employment figures, the implied spending is roughly $2,284 per employed person in the US compared with roughly $84 per employed person in China. These gaps help explain why many Chinese AI startups lean into institutional procurement tied to operational outcomes: the “product” often bundles deployment, customization, and accountability — not just model access.
For Chinese startups, constraints shape strategy in three major ways.
/odw-inline-subscribe-cta
Chinese AI developers design models to minimize usage costs. First, developers aim for efficiency. DeepSeek-V3 is a Mixture-of-Experts model with 671 billion parameters, but only 37 billion activated per token. While MoE techniques are not unique to Chinese models, cost per token and cost per task are business variables; when capital is scarce and customers are price-sensitive, efficiency expands the addressable market.
Chinese AI startups often target international markets from inception. Second, constraints influence companies’ market positioning. MiniMax served more than 212 million users across 200-plus countries through AI companion and role-play apps like Talkie, with overseas markets contributing over 70% of revenue. This “born-global consumer” strategy works in entertainment categories where spending habits are established and value is experiential. While MiniMax’s scale is exceptional, other Chinese AI products like CapCut (for video editing) and Faceu (for augmented reality filters, among other features) have demonstrated similar overseas-first patterns, suggesting this approach is becoming a viable alternative to competing directly with US incumbents in productivity tools.
Chinese companies tailor hardware for inference, not training. Third, chip companies focus on deployment-oriented infrastructure. Biren’s IPO shows that hardware startups can succeed by optimizing for inference rather than promising to dominate frontier training.
Economic pressures are fueling faster industrial deployment in China. More broadly, capital and monetization constraints also increase pressure on Chinese startups to find valuable use cases faster. Evidence suggests this is happening. In manufacturing, 67% of Chinese industrial firms have deployed AI in production, compared with 34% of analogous US firms — roughly double the adoption rate. Deloitte notes that many US manufacturers remain in “pilot purgatory,” beginning scaled deployment only in 2026. In logistics, China’s JD Logistics has leveraged AI to offer 12-hour delivery in core cities, versus Amazon Prime’s 1 to 2 days. Cainiao’s AI-powered consolidation has cut cross-border delivery times by 50%.
This accelerated deployment is due to three dynamics. First, while US enterprises wait for frontier models, Chinese companies are more pragmatic, deploying open-source models with heavy fine-tuning to solve immediate problems. Second, lower compliance friction in China shortens procurement cycles. Third, Chinese AI companies more often sell end-to-end solutions rather than use-specific tools, incentivizing buyers to adopt deep workflow integrations. When revenue depends on delivering outcomes rather than access, speed to deployment becomes a survival metric.
The US leads in model capability, while China leads in widespread deployment. This difference demonstrates that there is more than one vision of success in the AI industry. If we interpret progress as meaning frontier model capability, then the US keeps its edge — Chinese models trail by approximately 7 months. If, however, progress means economy-wide deployment, then China may be ahead, with its constraint-driven strategies compounding faster in specific layers.
Economic signals are the best markers of progress for China’s AI industry. Over the next 6 to 12 months, the strongest signs of whether Chinese AI startups’ strategies are working will come from economics, rather than model benchmarks. Starting with inference pricing, if efficiency gains are real, cost per useful task should decline steadily, reflected in API discounting and enterprise contract terms. Looking at the mix of revenue streams in public filings will also be informative; shifts between consumer and enterprise segments reveal which go-to-market motion is actually scaling. Additionally, monitoring renewal and expansion rates will distinguish pilots from durable adoptions. And, finally, overseas revenue growth from Chinese consumer AI products will be the measure of whether the born-global path can sustain momentum through distribution alone.
The conclusion is not that one country has the only viable route to AI success. Startups in different regions face different buyers, constraints, and distribution channels, so they optimize accordingly. The next winners in the AI industry will be the companies that treat those realities as product requirements, and then scale what works.
See things differently? AI Frontiers welcomes expert insights, thoughtful critiques, and fresh perspectives. Send us your pitch.
Modern memory architecture is vital for advanced AI systems. While the US leads in both production and innovation, significant gaps in export policy are helping China catch up.
Traditional insurance can’t handle the extreme risks of frontier AI. Catastrophe bonds can cover the gap and compel labs to adopt tougher safety standards.
