US vs. China in AI: Where America Leads and Where It's Falling Behind
The New York Times · September 23, 2026
Key takeaways
- The US leads in frontier AI model performance and controls the advanced chip supply chain that powers cutting-edge training.
- China is closing the gap through highly efficient open-weight models and faster real-world AI deployment across manufacturing and consumer tech.
- Energy capacity is emerging as a critical battleground, with China building out electricity generation faster than the US to fuel AI infrastructure.
The US-China AI rivalry isn't a simple story of American dominance anymore. It's a split-screen: the US is way out front in some areas, and quietly getting outpaced in others. Here's the real scoreboard.
Where the US Is Winning
America still holds the crown on frontier AI models — think the most advanced systems from labs like OpenAI, Anthropic, and Google DeepMind. These models generally outperform Chinese counterparts on complex reasoning, coding, and multimodal tasks. The US also dominates the chip supply chain that makes all of this possible. Companies like Nvidia design the GPUs that train these massive models, and export controls have made it harder for China to get its hands on the most advanced silicon. On top of that, US cloud infrastructure — the data centers and computing power needed to run AI at scale — remains unmatched in scope and reliability.
Where China Is Closing the Gap
Here's the part that should get more attention: China is not far behind, and in some areas it's arguably ahead. Chinese labs like DeepSeek and Alibaba have released open-weight models that rival top US systems in performance, often at a fraction of the training cost. That efficiency matters — it means China can compete with less access to cutting-edge chips by squeezing more out of what it has. China is also moving faster on real-world deployment. AI is getting baked into manufacturing, robotics, and everyday consumer apps at a pace that's hard for the US to match, partly because of scale and partly because of looser deployment friction. And then there's energy: building AI infrastructure requires massive amounts of power, and China has been adding electricity generation capacity far faster than the US, which could become a decisive advantage as AI models get bigger and hungrier for compute.
The Bigger Picture
This isn't a race that ends with one country "winning" outright. It's more like two different strategies colliding — the US betting on cutting-edge capability and chip supremacy, China betting on efficiency, speed of deployment, and raw infrastructure buildout. The open-weight model strategy from Chinese labs is especially notable because it changes who gets to use advanced AI. If Chinese open models become the global default for startups and developers outside the US, that's a form of influence that doesn't show up on a benchmark leaderboard.
What to Watch
Keep an eye on three things going forward: whether US export controls actually slow China down or just force more innovation, how fast each country can build out energy and data center capacity, and whether open-weight Chinese models start showing up as the backbone of apps and products used by people who've never thought about where their AI comes from. The headline numbers get attention, but the infrastructure race underneath is where this actually gets decided.
Why it matters
The AI race between the US and China will shape which technologies, products, and even values end up embedded in the tools people use every day, from your phone's assistant to factory robots. Understanding who leads where helps you see past the hype and grasp what's actually driving global tech competition.
Want deals on what you love?
Val finds local offers matched to your interests — free to start.
Meet Val