all wars
Active conflict

U.S. vs. China · the Civilizational Race

Two AI stacks, two chip ecosystems, two definitions of victory.

Party 1United States · NVIDIA + frontier labs + cloud trinity
Party 2China · indigenous chips + open labs + national strategy
Kernel

The U.S.-China AI competition is the dominant geopolitical fact of the 2020s technology cycle. By 2026 each side has its own frontier-lab portfolio, its own preferred chip supply chain, its own open-or-closed posture, and its own narrative about why it will win. The substantive technical convergence in 2024–2025 (DeepSeek R1, Qwen-Max, the Mistral-style efficiency frontier) has scrambled the comfortable U.S. assumption of permanent lead.

§ 01

Frontline

Chip exports (the October 2022 controls + 2024 expansion). HBM and EUV equipment. Frontier-model release cadence. Open weights as a vector of soft power. Talent flows (Chinese-trained researchers in U.S. labs, U.S.-trained Chinese researchers returning, talent visa friction).

§ 02

Doctrine — U.S.

Maintain the compute lead through export controls; ride NVIDIA-CUDA-cloud trinity at the frontier; bind allies into one compute zone; accept that AI is the defining national-security technology of the 21st century and act accordingly.

§ 03

Doctrine — China

Substitute via efficiency. Race up the open-weights curve. Build an indigenous chip industry on whatever node generation is reachable. Cultivate a parallel application ecosystem (WeChat-class platforms, sovereign cloud, hardware vertical integration) that is structurally insulated from the U.S. stack.

§ 04

Stakes

Whose stack becomes the default for the Global South in 2027–2030 may be the single most consequential question of the cycle. The U.S. assumption that English-language frontier models will dominate everywhere is increasingly questionable. In Africa, parts of Latin America, and most of Southeast Asia, Chinese open models are already common.

§ 05

Outlook

The most likely 2030 outcome is a bifurcated world: two AI stacks, two chip supply chains, two agent-payment rails, and a long unstable middle. Neither side wins outright; both pay enormous compliance and inefficiency costs for the bifurcation; the global development frontier slows accordingly.