China’s AI Ambitions Face Entrenched Chip Manufacturing Moats
Baillie Gifford investment manager Paulina McPadden argues that China remains a formidable AI competitor, but its capital and innovation capacity will not quickly overcome the manufacturing advantages held by TSMC, SK Hynix and ASML. She says those advantages rest on specialized equipment and decades of accumulated process knowledge, while export controls further constrain Chinese efforts to catch up. For investors, she argues, the more durable AI opportunities may also lie in company-specific systems, citing Shopify’s product catalog and MercadoLibre’s software-development workflow.
China can compete, but leading-edge chips and memory cannot be bought quickly
Paulina McPadden said she would not bet against China. She described the country as a longstanding engine of growth, disruption, and innovation, and pointed to Baillie Gifford holdings including BYD, Pinduoduo, and Tencent as “tremendous companies.”
But she drew a distinction between China’s broad innovative capacity and the positions occupied by TSMC, SK Hynix, and ASML, which she described as “almost in a league of their own.” In leading-edge logic chips and high-bandwidth memory, capital is necessary, but it does not by itself create a competitive manufacturing position.
TSMC in particular has created effectively a monopoly at the leading edge of chips, and that's going to be very difficult to disrupt.
TSMC’s advantage, in McPadden’s account, rests on both the cost of building a leading-node fab and manufacturing process knowledge accumulated over decades. That knowledge is difficult to replicate, she argued. China’s restricted access to equipment makes the challenge harder still: Chinese companies do not have access to much of the equipment they would need because of export controls.
She sees a comparable dynamic in high-bandwidth memory, or HBM, where SK Hynix is deriving a growing share of revenue and earnings. HBM requires complicated processes and equipment, McPadden said, while CXMT—the Chinese memory maker Ed Ludlow raised for comparison—does not yet have access to much of that equipment. Chinese firms may eventually reproduce those capabilities, she allowed, but “that will take time.”
Capital follows businesses that can sustain an advantage
Capital matters in AI because the industry is capital-intensive, Paulina McPadden said. But for a long-term investor, the more useful task is to identify exceptional businesses. Those businesses will naturally have better access to capital because they are fundamentally stronger.
Her test is deliberately less responsive to the pace of AI news. Investors should ask what industry a company is trying to disrupt, whether it is creating a new industry, and whether it is investing in innovation over a period long enough for structural change to emerge. In her formulation, “long term” means five or 10 years; true change takes decades, not quarters.
That frame is also a response to market structure. McPadden said 60% of US investing is conducted through passive vehicles, 75% of trading volume comes from quant funds, and retail participation is rising. The result, she argued, is a market that has become faster and noisier—and one more inclined to conflate what is new with what is important.
The practical implication is to treat AI less as a sequence of news events than as a search for durable company attributes. For McPadden, those attributes include the ability to create or disrupt an industry, sustain investment in innovation, and build an advantage that can endure beyond a quarterly cycle.
Platforms can turn specific systems into AI advantages
The non-US AI opportunity is not confined to the semiconductor supply chain, McPadden argued. She highlighted Shopify and MercadoLibre as two businesses where AI is affecting operations and economics in different ways.
Paulina McPadden described Shopify as a company that mediates complexity between merchants and consumers. AI agents add to that complexity: as autonomous systems increasingly make purchase decisions for consumers, they need reliable sources of ground truth about products.
Shopify’s Catalog—its record of merchant items and their associated metadata—serves that role. McPadden said that when AI agents are involved, orders using Catalog data show twice the conversion of orders based on simply scraping website data. For a company that monetizes based on gross merchandise volume, she presented that difference as consequential: better product information for agents can support more successful purchases.
MercadoLibre illustrates a different use of AI. McPadden characterized the company’s current posture as familiar: it is investing substantially and growing users because it sees a critical market inflection point. It is seeing strong GMV and user growth, she said, on the back of lower shipping requirements and lower merchant take rates.
AI enters her MercadoLibre case primarily through internal development. The company is using it to speed the cadence of software development, McPadden said, reporting 75% higher code deployment while the number of rollbacks has fallen. The point is not merely that MercadoLibre is shipping faster, but that it is doing so at higher quality.
| Company | AI-related mechanism | Reported effect |
|---|---|---|
| Shopify | Catalog data provides product ground truth for AI shopping agents | 2× higher conversion than website-data scraping when agents are involved |
| MercadoLibre | AI used to accelerate internal software development | 75% higher code deployment, with fewer rollbacks |
The comparison places AI advantage in concrete systems rather than in chip access alone. TSMC and SK Hynix illustrate the difficulty of reproducing accumulated manufacturing capability. Shopify’s case rests on the product catalog and metadata that agents can use as reliable information; MercadoLibre’s rests on AI’s reported effect on its internal development workflow. Those are distinct mechanisms, and McPadden presented each as part of the longer-term attributes investors should look for.