Jensen Huang’s Five-Layer Cake: Who Leads the Global AI Race
By Joel Wong
Jensen Huang’s Five-Layer Cake: Who Leads the Global AI Race
NVIDIA CEO Jensen Huang argues that AI is not a single technology but a five-layer industrial stack. Leadership in the AI era depends on strength across all layers. His assessment shows a split world: the U.S. leads in frontier innovation, while China leads in scale, infrastructure, and application.
Layer 1 — Energy (Advantage: China)
AI requires enormous electricity for data centers. China now has roughly twice U.S. energy capacity and is expanding rapidly, while U.S. capacity is “flat.” Energy abundance becomes the foundation for AI “factories.”
Layer 2 — Chips (Advantage: United States)
The U.S. remains “generations ahead” in cutting-edge chips (e.g., NVIDIA). But China is closing the gap via massive state subsidies, localized supply chains, and energy discounts of up to 50% for AI-related manufacturing.
Layer 3 — Infrastructure (Advantage: China)
China builds data centers and AI supercomputers with exceptional speed and low cost, giving it an infrastructure “velocity” that challenges the West. This layer is where China’s industrial capacity becomes a strategic advantage.
Layer 4 — Models (Advantage: United States)
The U.S. leads by roughly six months on frontier models (OpenAI, Anthropic, Google). However, China is far ahead in open-source models, which accelerates developer adoption and ecosystem growth inside China.
Layer 5 — Applications (Advantage: China)
China is integrating AI into everyday life and business much faster than the U.S.
Public optimism about AI’s benefits drives broad adoption—from retail to manufacturing to government services. Huang emphasizes that the country that applies AI at scale first will dominate the new industrial revolution.
Huang’s Strategic Warning
Because U.S. export controls restrict American firms from competing in China, Beijing is racing to build a fully domestic, vertically integrated AI stack—energy, chips, infrastructure, models, and applications—which it can later export globally.
In Huang’s view, the global AI competition is no longer about winning a single technology, but about building and deploying an entire AI industrial platform faster than the other side.
