Two Systems. Two Economics. One Supercycle.

“America raises the ceiling. China lowers the floor.”

America is building increasingly powerful intelligence. China is building increasingly abundant intelligence.

The U.S.–China AI race is no longer primarily a contest over who has the best model. It is becoming a contest between two fundamentally different economic systems for producing and deploying intelligence.

America is maximizing the value of intelligence. China is maximizing the volume. That distinction changes the investment thesis: America pushes the frontier higher, China pushes the deployment cost lower. One creates scarcity value, the other creates abundance — and together they are making the AI Supercycle larger.

1. 01 America Monetizes Scarcity

   Create intelligence few others can produce, deploy it into high-value tasks, capture premium economics, and reinvest the rents into the next frontier. The American system asks: how valuable can the next unit of intelligence become? signal: Premium economics funds the next frontier.

2. 02 China Industrializes Abundance

   Make intelligence increasingly sufficient, compress its cost, deploy it across enormous volumes, and turn scale into data, learning and lower unit economics. China asks: how many units of intelligence can the economy absorb? signal: Scale turns sufficiency into a flywheel.

3. 03 China No Longer Needs Parity

   China does not need Ascend = Nvidia, CANN = CUDA, or DeepSeek = the best U.S. frontier model. It needs only that Chinese stack economics ≤ economic value of Chinese AI. A weaker accelerator can be compensated by better interconnect, more efficient models, cheaper energy, lower-cost inference and manufacturing scale. The objective shifts from technological parity to economic sufficiency. signal: The hurdle is sufficiency, not parity.

4. 04 The American Risk Is Not Losing The Frontier

   The biggest American risk may not be China building a better model — it may be China making frontier superiority less economically decisive. If one system delivers 100 capability for $100 while another delivers 90 for $10, enormous categories — manufacturing, customer service, logistics, coding, mobility, robotics, industrial automation, edge AI, emerging markets — may choose the second. China does not need to eliminate the gap; it needs to make the gap worth less. signal: The threat is a cheaper good-enough, not a better best.

5. 05 The Chinese Risk Is The Opposite

   If frontier intelligence begins producing non-linear returns, being slightly behind becomes extremely expensive. Better AI could accelerate AI research, drug discovery, chip design, software engineering, robotics and scientific discovery — making the American capability advantage recursive: FRONTIER AI → AI BUILDS BETTER AI → FRONTIER MOVES FASTER. If that mechanism dominates, abundance below the frontier may not compensate for falling behind at the top. signal: Recursive capability could outrun commoditization.

Key takeaways

Two theories of economic power

• Stop treating U.S.–China AI as one race — the systems optimize different economic variables.
• America owns the capability frontier, strongest where marginal intelligence creates disproportionate value.
• China is building the economic frontier — making sufficiently powerful intelligence cheap, deployable and ubiquitous.
• China's objective is not parity but making American superiority less economically decisive.
• America's counterweapon is recursive intelligence: frontier AI accelerating AI research itself.
• Physical AI may become China's strongest battlefield, converting cheap intelligence into industrial output at scale.
• The infrastructure thesis strengthens under either outcome — more capability and lower costs both drive more compute, memory, networking and power.