Nvidia’s Upside Case No Longer Depends on China Access
Baillie Gifford investment manager Paulina McPadden argues that Nvidia’s long-term case does not depend on renewed access to China, where domestic high-power chips still trail Nvidia’s leading products by a wide margin. Speaking to Bloomberg’s Ed Ludlow, she said the more important question is whether China can recreate the complex semiconductor supply chain behind AI hardware, while identifying TSMC, SK hynix and ASML as non-US companies with durable roles in that ecosystem.
Nvidia's China access is not central to McPadden's upside case
Paulina McPadden is not relying on Nvidia regaining meaningful China access to justify the upside case she described. China access has not been built into “our upside cases for Nvidia for a while,” she said, because doing so would mean trying to predict future political decisions. Ed Ludlow described the current base-case assumption for Nvidia as “zero access to China” and asked how much that matters to the investment view.
McPadden moved the issue from near-term export permission to the longer-term durability of the supply chain. She said H800 chips had been approved for export to China since last year, but that Chinese government decisions not to approve purchases by Chinese companies had prevented those sales from going through for Nvidia.
The more consequential question, in her view, is whether China can build a domestic chip supply chain that eventually threatens Nvidia. She allowed that this may be possible over the long term, but gave two reasons she is comfortable over at least the next five years.
First, she said existing high-power chips in China remain “a fraction of the performance” of Nvidia's leading chips. That gap, in her view, gives non-Chinese companies and AI researchers a durable advantage. Second, she argued that a scaled semiconductor supply chain is much harder to build than other strategic manufacturing sectors where China has used capital investment to build strong positions, including EVs and batteries.
It's not as easy as just putting capital into it as China has done in the past with things like EVs.
Semiconductors, McPadden said, depend on “extremely complex bits of equipment” and cooperation across countries. She was skeptical that China could reproduce that complexity year after year if it were relying only on domestic demand and unable to sell internationally.
A Bloomberg graphic listed Nvidia, TSMC, ASML, Tencent, and SK hynix as stocks in McPadden's strategy. Her Nvidia answer was therefore also about the broader AI hardware ecosystem: not only who designs chips, but who manufactures them, supplies the memory, and builds the machines that make production possible.
Memory may be less cyclical, but that remains a hypothesis
The question around SK hynix is whether memory demand has become more durable than the boom-and-bust cycle that has historically defined the sector. Ed Ludlow said Jensen Huang had told memory makers, including SK hynix, three years earlier what he thought AI infrastructure demand would require, and had tried to persuade them to invest in permanent capacity.
Paulina McPadden cautioned that a bottleneck is not automatically the same thing as investable upside. Markets can identify bottlenecks before the fundamentals arrive and price them in early. For long-term investors, she said, the task is not just to find constraints in the system, but to identify structural opportunities: companies that become better as they scale and deepen their competitive position.
That led her to compare SK hynix with TSMC. In SK hynix, she sees the possibility of a company becoming “substantially better” over the long term. The change from history, she said, is that SK hynix is now able to extract long-term purchase agreements from customers. The memory industry has consolidated and is less irrational than it used to be.
She also pointed to high-bandwidth memory, and especially custom HBM, as a potential source of customer lock-in. If memory products are increasingly customized for AI systems, they may no longer behave like the commoditized products memory has often been in the past. But McPadden was careful not to state that as proven.
There is a case to be made that with the rise of HBM and in particular custom HBM, that creates a degree of lock-in with customers that means that these are no longer commoditized products as they have been in the past. But that's just a hypothesis.
A Bloomberg market graphic showed SK hynix up 784.16% over one year. McPadden's point was that SK hynix may be changing structurally: longer-term customer commitments, a more consolidated industry, and custom HBM could make parts of memory less commoditized than in the past.
By contrast, she described TSMC as a more established example of the kind of company Baillie Gifford wants to own. TSMC is the largest position in the portfolio she was discussing. Her argument was that TSMC has already demonstrated “extreme control and dominance” over its supply chain and an ability to coordinate many actors in a more complex semiconductor ecosystem. As chip production becomes more complicated, she said, TSMC “just keeps getting better.”
AI is designed in America, but much of the investable infrastructure is international
Paulina McPadden described the AI supply chain as geographically split: chips and AI are “designed in America,” but manufactured internationally. That is why she sees a “rich hunting ground” for AI-related companies outside the US.
The companies she named each occupy a different layer of the global AI supply chain. TSMC manufactures the chips. SK hynix supplies memory inside the systems. ASML, based in the Netherlands, builds the EUV lithography machines used to produce chips in TSMC fabs. Nvidia remains central to the American design side of the ecosystem. Tencent was shown in Bloomberg's list of stocks in McPadden's strategy but was not discussed in the exchange.
Ed Ludlow put the question in the context of chip-sector volatility, saying the sector had been close to correction territory from Friday through Tuesday's close. McPadden's response was that the short-term market move did not change the long-term fundamentals she was focused on. She described TSMC, SK hynix, and ASML as dominant in their particular industries, making them attractive for a concentrated growth strategy.
She also pushed back on one short-term market interpretation. The lack of news on chip exports after the referenced China summit had been treated negatively by the market, she said, because markets often price the absence of a catalyst as bad news. From her long-term perspective, that did not affect the fundamentals of the companies under discussion.
The stronger signal, for McPadden, was TSMC's own outlook. She said TSMC had raised its guidance for AI growth over the next five years “to 56%.” The number was encouraging to her because of where it came from: TSMC has insight into the entire semiconductor supply chain.
The signal, in her account, was not just that one company was more optimistic; it was that a central manufacturer in the AI hardware ecosystem had enough visibility to raise its outlook with confidence. Bloomberg also showed intraday gains for both TSMC ADR and Nvidia, but McPadden's emphasis was on the durability of the companies' positions and the visibility they have into AI demand, not on a single session's price action.