China vs USA: The AI Arms Race and What It Means for the Global Economy
In January 2025, a Chinese AI startup called DeepSeek released a model that matched the performance of OpenAI’s best systems — at a fraction of the cost, built with chips that were subject to US export controls. The reaction in financial markets was instantaneous: Nvidia lost nearly $600 billion in market capitalisation in a single day. The reaction in policy circles was equally stark: if China could build frontier AI under sanctions, the entire US export control strategy needed rethinking. The US-China AI race is the defining technological competition of the 21st century — and it connects directly to the economic and geopolitical themes throughout our AI and economy series.
- → The US leads in frontier AI model capability and AI chip design; China leads in AI deployment scale, surveillance applications, and manufacturing integration
- → US export controls on advanced chips have slowed but not stopped Chinese AI development — DeepSeek demonstrated that China can innovate around compute constraints
- → Both governments treat AI supremacy as a national security imperative — the economic and military applications are inseparable
- → The competition is bifurcating the global technology ecosystem into two incompatible spheres, with major implications for multinational businesses and investors
The State of Play: Who Is Winning?
As of 2026, the United States maintains a clear lead in frontier AI model development. OpenAI, Anthropic, Google DeepMind, and Meta AI are producing the world’s most capable large language models. US companies dominate the AI chip design market through Nvidia. The underlying research ecosystem — universities, talent pipelines, venture capital — remains unmatched globally.
China, however, is not simply behind and falling further back. It has several structural advantages: an enormous domestic market for AI deployment, a government willing to mandate AI adoption across industries, vast proprietary datasets from its 1.4 billion population, and a manufacturing ecosystem that is integrating AI into physical production at a scale the US cannot match. China leads in AI deployment in surveillance, smart cities, autonomous vehicles (with companies like BYD and CATL embedding AI into industrial systems), and manufacturing automation.
The Chip War: Export Controls and Their Limits
The Biden administration’s October 2022 export controls — restricting the sale of advanced AI chips (Nvidia A100s and H100s) to China — represented the most aggressive US technology policy intervention since the Cold War. The controls were designed to prevent China from accessing the compute needed to train frontier AI models. They were subsequently tightened multiple times, extending to a growing list of chips and chip-manufacturing equipment.
“Export controls can slow a competitor. They cannot stop a determined nation-state with sufficient talent, capital, and strategic intent. DeepSeek proved that efficiency innovation can partially compensate for compute constraints.”
The DeepSeek episode of January 2025 was a direct challenge to the export control logic. By using more efficient training techniques, distillation from existing models, and architectural innovations, DeepSeek’s team built a frontier-competitive model using chips available to them despite export restrictions. This does not mean export controls are useless — they do impose real costs and delays — but it demonstrated that they are not a definitive barrier to Chinese AI progress.
The deepest vulnerability in China’s AI ambitions is semiconductor fabrication. The world’s most advanced chips — at 3nm and 2nm process nodes — can only be manufactured by Taiwan’s TSMC and South Korea’s Samsung, using equipment from Dutch firm ASML that is also subject to export controls. China’s domestic champion SMIC is currently limited to 7nm processes — competitive for many applications, but behind the frontier. Closing this gap is Beijing’s top technology priority and will take years at minimum.
Economic Bifurcation: Two Technology Ecosystems
The US-China AI competition is producing a fundamental bifurcation of the global technology ecosystem. Companies and governments increasingly face a binary choice: build on US-designed AI infrastructure (Nvidia chips, Microsoft Azure, AWS, Google Cloud) or build on Chinese alternatives (Huawei chips, Alibaba Cloud, Baidu AI). These ecosystems are increasingly incompatible — using one set of tools makes integration with the other more difficult.
For multinational companies operating in both markets — consumer goods firms, automotive manufacturers, financial institutions — this bifurcation creates significant operational complexity. Products designed with US AI tools may need to be rebuilt for the Chinese market using Chinese alternatives. Data governance requirements in both jurisdictions increasingly conflict. This is one dimension of the broader deglobalisation dynamic covered in our analysis of de-dollarisation and global fragmentation.
What the AI Race Means for Investors
| Implication | Investment Angle |
|---|---|
| US chip export controls accelerating | ASML, TSMC, domestic US semiconductor equipment exposure |
| China accelerating domestic chip capability | SMIC, Huawei ecosystem beneficiaries (limited Western access) |
| Global tech bifurcation | Caution on multinationals with heavy China AI dependency |
| DeepSeek-style efficiency breakthroughs | Compresses compute moat; potential Nvidia valuation headwind |
| AI military applications accelerating | Defence tech companies with AI integration (Palantir, Anduril) |
The US-China AI race is not a temporary trade dispute — it is a structural, decade-long competition for technological supremacy that will shape the geopolitical and economic order of the 21st century. The US currently leads on frontier capability; China leads on deployment scale and manufacturing integration. Export controls have slowed but not stopped Chinese AI development, as DeepSeek demonstrated. For investors, the key implications are: favour the infrastructure layer that benefits regardless of which AI applications win, remain cautious about companies with unhedged China exposure as bifurcation accelerates, and track semiconductor fabrication capability as the deepest long-term constraint on the entire competition.
Responses