The Winner Take All AI Race Between The US and China
Capital Flows
· Capital Flows
· July 15, 2026 at 04:03
· ⏱ 3 min read
| Read on Substack ↗
Summary
The US and China are locked in a winner-take-all AI race that drives equity market liquidity and valuations, with AI IPOs functioning as liquidity clearing mechanisms. China's refusal to revalue its currency creates excess dollar liquidity that fuels record foreign buying of US equities, while the AI trade has fragmented into four separate sub-trades (chipmakers, hyperscalers, power, and the China cycle). New CME derivatives (single stock futures and compute futures) will reshape hedging flows and risk pricing across the AI complex.
•July FOMC now pricing 82% probability of a hold, bonds caught a bid, equity pullbacks driven by rate risk remain short-term.
•AI IPOs (Cerebras, SK Hynix) are acting as FOMC-like events; Cerebras set the top in NVIDIA and SK Hynix set the low almost to the day, then rallied 24% on 61 million shares.
•Equity valuations are the highest in human history across S&P, Russell, Nikkei, and Europe, driven not just by AI but also by China's currency policy creating excess dollar liquidity.
•Chinese open source models have captured ~46% of US enterprise usage, and the top 10 S&P 500 names now account for a record 41% of the index.
•The AI trade has split into four distinct trades (chipmakers, hyperscalers, power, and the China cycle), and the binding constraint migrates along the chain from wafers to memory to power.
•CME plans to launch single stock futures on 55 names with a 100x multiplier, with compute futures following, which will reprice hedging and geopolitical risk across the AI complex.
Author explicitly notes SK Hynix 'set the low almost to the day' and then 'traded up 24 percent on 61 million shares' — strong price momentum and volume confirmation of a liquidity-driven rally.
Author explicitly notes SK Hynix 'set the low almost to the day' and then 'traded up 24 percent on 61 million shares' — strong price momentum and volume confirmation of a liquidity-driven rally.
Risk: Memory cycle volatility; SK Hynix is highly cyclical and exposed to HBM pricing.
Author states 'Cerebras set the top in NVIDIA' — the AI IPO cycle may have marked a local peak for NVIDIA, suggesting near-term downside risk or rotation out of the leader.
Author states 'Cerebras set the top in NVIDIA' — the AI IPO cycle may have marked a local peak for NVIDIA, suggesting near-term downside risk or rotation out of the leader.
Risk: NVIDIA's dominance could reassert if AI capex accelerates further; the 'top' may be temporary.
Author notes the binding constraint migrates 'from wafers to memory to power' — wafers directly implicate TSMC as the dominant foundry, implying sustained demand and pricing power.
Author notes the binding constraint migrates 'from wafers to memory to power' — wafers directly implicate TSMC as the dominant foundry, implying sustained demand and pricing power.
Risk: Geopolitical risk (Taiwan) and cyclical capex shifts.
Author identifies 'power' as one of the four separate AI trades and states the binding constraint moves to power — Vertiv is a leading provider of power and cooling infrastructure for AI data centers,
Author identifies 'power' as one of the four separate AI trades and states the binding constraint moves to power — Vertiv is a leading provider of power and cooling infrastructure for AI data centers, directly benefiting.
Risk: Competition from Trane (TT) and others; execution on large data center projects.
This newsletter, published July 15, 2026,
features Capital Flows
discussing 000660.KS, NVDA, TSM, VRT.
4 trade ideas extracted by AI with direction and confidence scoring.