China Chips Still Behind US, Nvidia in Performance: Baillie Gifford

Baillie Gifford's Paulina McPadden said Nvidia remains a key part of the global chip ecosystem, while US AI researchers and companies still have a long-term edge over China. She also described a "rich hunting ground" for AI investment outside the US, suggesting broadening opportunities across the sector rather than a single-country trade.

Analysis

The key takeaway is not that Nvidia is the only winner, but that the AI stack is still being economically centralized in the US while the marginal investment opportunity is spreading globally. That favors the highest-quality compute, networking, and software enablers over generic semiconductor exposure: the moat is increasingly in access to capex, power, cloud distribution, and developer ecosystems, not just model quality. In practice, that means NVDA remains the cleanest expression of AI infrastructure demand, but the second-order beneficiaries are likely to be US cloud capex enablers and select non-US application-layer names with local distribution advantages.

The more interesting risk is that China’s AI progress may be constrained less by talent than by access to frontier compute and advanced packaging over the next 12-24 months. That creates a bifurcated ecosystem: US firms can monetize at scale sooner, while non-US markets may offer better valuation entry points only if they can convert AI adoption into earnings faster than the market currently expects. For semis, the real watch item is whether supply-chain bottlenecks ease enough to broaden winners beyond NVDA; if not, revenue concentration stays high and multiple dispersion across AI hardware will widen.

Consensus may be underestimating how durable the US advantage is if regulation, export controls, and hyperscaler capex remain aligned for another 6-8 quarters. But the market may also be overpaying for perfect execution: NVDA is exposed to any pause in enterprise AI inference spending, which can show up with a lag of 1-2 quarters even if headline demand remains strong. The contrarian setup is to own the picks-and-shovels beneficiaries while being selective on the most crowded AI beta names, especially where margin expectations already discount uninterrupted hyperscaler demand.

Key Decisions for Investors

• Stay long NVDA into any 5-8% pullback as the highest-conviction US AI infrastructure expression; use 3-6 month horizon and size for volatility, not linear upside.
• Pair long NVDA / short a basket of weaker AI hardware laggards or over-owned AI beta names to isolate quality-of-demand and moat durability; target a 2:1 reward/risk if AI capex stays strong for another quarter.
• Add exposure to US cloud and networking beneficiaries on weakness over the next 1-2 quarters, as the second-order winner is the infrastructure layer that captures recurring capex, not just the chip vendor.
• Avoid broad China AI exposure as a catch-up trade until there is evidence of compute access and commercialization closing the gap; the risk is a value trap over a 12-24 month window.
• Consider a selective long on non-US AI application/platform names only where valuation already discounts slower growth, because the upside is in local monetization, not frontier model leadership.