The global technology landscape is experiencing a massive, capital-intensive restructuring driven by the rapid expansion of artificial intelligence infrastructure. As asset managers around the world race to allocate capital to the semiconductor supply chain, a clear division is emerging between the companies that dominate the global commercial market and those focused on domestic self-sufficiency. In August 2026, Paulina McPadden, an investment manager who directs the international concentrated growth strategy at prominent Scottish asset management firm Baillie Gifford, strongly backed TSMC and SK Hynix over Chinese chipmakers, describing the Taiwanese and South Korean giants as indispensable, near-monopoly leaders of the global AI hardware ecosystem.
McPadden’s high-conviction thesis comes at a time of heightened volatility and geopolitical tension in the technology sector. While the Chinese government is pumping billions of dollars of state subsidies into domestic semiconductor companies to build a self-reliant tech stack, these local firms are facing severe technical bottlenecks and operational inefficiencies. For global asset managers looking to maximize long-term capital appreciation, Chinese chip foundries remain structurally unattractive investments, as their state-backed operations prioritize national volume over corporate profitability.
In sharp contrast, TSMC and SK Hynix control the vital physical gateways of global computation. By holding exclusive positions in advanced foundry lithography and high-bandwidth memory (HBM) packaging, these two Asian giants possess robust economic moats and highly secure commercial margins. According to Baillie Gifford, investors seeking exposure to the multi-trillion-dollar AI revolution should prioritize these established industry leaders, as their near-monopoly status makes them the primary beneficiaries of the global hardware buildout.
The Near-Monopoly Moats of TSMC and SK Hynix
To understand why Baillie Gifford is backing TSMC and SK Hynix so aggressively, it is necessary to examine the physical and capital-intensive barriers that protect their market leadership. These are not simple software companies that can be easily replicated; they are massive industrial giants whose manufacturing facilities require billions of dollars in continuous upfront investments.
TSMC’s Capital-Intensive Two Nanometer Dominance
Taiwan Semiconductor Manufacturing Company, commonly known as TSMC, is the undisputed king of advanced contract chip manufacturing, fabricating virtually every high-end artificial intelligence processor in the world, including those designed by Nvidia, AMD, and Apple. The scale of the company’s operations is historic. TSMC expects its revenue for the second quarter of 2026 to reach $40.2 billion, with third-quarter revenue guidance set between $44.6 billion and $45.8 billion.
The primary competitive barrier protecting TSMC is its rapid transition to advanced manufacturing nodes. While competitors are still struggling to stabilize their 5-nanometer and 3-nanometer yields, TSMC is already preparing for high-volume production of its next-generation 2-nanometer chips.
To maintain this absolute technological lead, TSMC has announced a massive capital expenditure budget of $60 billion to $64 billion for the year. This level of capital spending is virtually unmatchable for any other foundry in the world, ensuring that TSMC will remain the sole provider of the advanced processors required to run the next generation of frontier AI models.
SK Hynix and the High-Bandwidth Memory Bottleneck
The second major bottleneck in the artificial intelligence supply chain is high-bandwidth memory. An advanced AI processor cannot operate at peak speed without highly specialized, high-density memory chips capable of transferring massive amounts of data to the processor instantly.
South Korea’s SK Hynix has established an unbreakable grip on this high-end memory market, securing an exclusive position as the primary supplier of HBM3e and next-generation HBM4 chips to Nvidia.
This dominant market position has allowed SK Hynix’s valuation to soar, with its market capitalization recently surpassing $1 trillion. While traditional memory chips were historically treated as a low-margin, cyclical commodity, the extreme demand for HBM has turned the sector into a high-margin, custom hardware space.
Because manufacturing HBM requires highly complex advanced packaging and strict quality controls, SK Hynix’s yield rates are significantly higher than those of its competitors, giving the company an exceptional competitive advantage that is currently driving record-breaking corporate profits.
Decoding China’s Chip Gap: The Yield and Lithography Hurdles
The massive demand for artificial intelligence hardware has forced the Chinese government to accelerate its own domestic semiconductor development, but local firms are encountering severe physical and technological limitations.
The Costly Complexity of DUV Multi-Patterning at SMIC
China’s leading semiconductor foundry, Semiconductor Manufacturing International Corporation (SMIC), has made notable progress, successfully producing 7-nanometer-class processors for domestic giants like Huawei. However, SMIC remains generations behind TSMC, which is already entering 2-nanometer production.
The primary obstacle preventing Chinese foundries from scaling up advanced production is their lack of access to Extreme Ultraviolet (EUV) lithography systems. Under United States-led export controls, Dutch equipment giant ASML is legally barred from selling its advanced EUV systems to Chinese firms.
Consequently, SMIC must rely on older, less-advanced Deep Ultraviolet (DUV) lithography machines to manufacture smaller circuits.
To print 7-nanometer patterns using DUV machines, engineers must rely on highly complex “multi-patterning” techniques, passing the silicon wafers through the machine multiple times.
This multi-patterning process dramatically increases manufacturing costs, lowers production speeds, and results in extremely low yield rates compared to TSMC’s efficient, EUV-based process. This operational complexity makes Chinese-manufactured advanced chips economically uncompetitive on the global stage.
CXMT’s Generational Lag in Advanced DRAM Architecture
The domestic Chinese memory sector is facing similar, highly restrictive hardware bottlenecks. ChangXin Memory Technologies (CXMT), China’s primary manufacturer of dynamic random-access memory (DRAM), has experienced massive domestic support, with its shares surging by over 460% during its highly publicized Shanghai STAR Market debut in July 2026.
However, from a technical perspective, CXMT remains at least one to two generations behind South Korean market leaders Samsung Electronics and SK Hynix.
The company’s early attempts to develop high-bandwidth memory are severely restricted by lithography licensing bans and advanced packaging constraints.
Without the specialized machinery needed to stack and connect memory dies vertically, Chinese-made HBM cannot achieve the data transfer speeds and thermal efficiencies required to power frontier-scale AI training servers, establishing a clear technological ceiling on the capabilities of China’s domestic hardware stack.
The Divergence of Investment Economics: State Priorities vs. Corporate Profits
For global asset managers like Baillie Gifford, the choice between investing in Western-aligned semiconductor leaders and Chinese foundries is not just a technical debate; it is a fundamental question of investment economics.
Why Beijing Prioritizes Volume Over Profitability
The primary reason why global investors are avoiding Chinese foundries is a misalignment of corporate goals. In a standard market economy, a public corporation’s primary objective is to maximize capital efficiency, generate high profit margins, and deliver strong returns to its shareholders.
In China’s state-directed economy, however, the government’s primary objective is national security, technology self-reliance, and volume production.
Beijing has pumped billions of dollars of state subsidies into domestic foundries like SMIC and CXMT, instructing them to produce as many chips as possible to insulate the domestic economy from foreign export controls.
Because these companies are backed by massive state capital, they do not need to operate profitably to survive. They can afford to run inefficient, high-cost manufacturing lines with extremely low yields, prioritizing national self-sufficiency over corporate margins.
For a global investor seeking capital appreciation, backing a company that does not prioritize profitability is a highly risky strategy, as the firm’s massive capital spending will rarely translate into strong cash flows or dividend payouts for minority shareholders.
The Parallel “Good Enough” Domestic AI Stack in China
Despite these financial inefficiencies, analysts warn that China’s domestic semiconductor progress should not be underestimated. While Chinese-made chips like Huawei’s Ascend series remain technically inferior to Nvidia’s latest Blackwell hardware, they are proving commercially “good enough” to support a parallel, highly functional domestic AI stack.
Major Chinese internet companies, including Alibaba, ByteDance, and Tencent, are actively optimizing their software applications to run on local silicon, reducing their dependency on Western hardware by design.
For instance, ByteDance recently signed a massive five-year agreement with memory manufacturer CXMT worth over $7 billion, ensuring a stable, domestic supply of memory chips for its local data centers.
While this domestic replacement campaign ensures that China’s digital economy can continue to function and innovate behind trade barriers, it exists as a closed, highly subsidized ecosystem that offers very little commercial value or growth potential to international portfolio investors.
Navigating Market Volatility: The July 28 Tech Rout and Strategic Realignment
The global semiconductor market has experienced intense volatility over the past several weeks, as investors grapple with shifting geopolitical landscapes, high valuations, and the massive capital expenditures required to fund the AI transition.
Buying the Valuation Disconnect After the Seoul Market Crash
The volatility reached a dramatic peak on July 28, 2026, when a massive, regional tech selloff wiped out billions of dollars in market value across Asian stock exchanges. Shares in memory giants Samsung and SK Hynix plummeted by 13.4% and 14.7% respectively in a single session, dragging South Korea’s benchmark KOSPI index down by 10.8% and triggering a sharp pullback in global semiconductor valuations.
For short-term traders, this rapid selloff was a terrifying event. For long-term, active managers like Baillie Gifford, however, the sharp decline represented a classic valuation disconnect and a highly attractive buying opportunity.
Paulina McPadden noted that the fundamental demand for advanced computing hardware remains completely unchanged. Data center construction is accelerating globally, and Nvidia’s partners continue to face severe supply shortages for advanced chips and HBM.
By purchasing shares of TSMC and SK Hynix during these panic-driven market corrections, active managers can accumulate high-quality, near-monopoly assets at a significant discount, capitalizing on short-term market anxiety to drive long-term portfolio outperformance.
SK Hynix’s Three-Billion-Dollar Chongqing Divestiture
To maintain its technological lead and fund its massive expansion plans back home, SK Hynix is also executing strategic realignments of its international asset portfolio. The South Korean memory giant is currently consulting with financial advisers regarding potential strategic options, including a stake sale, for its massive back-end packaging and testing facility in Chongqing, China.
The Chongqing facility, which serves as a major back-end production base for SK Hynix’s global NAND flash memory lines, carries an estimated valuation of approximately $3 billion.
By bringing in new investors or selling a controlling stake to local Chinese funds, SK Hynix can unlock massive amounts of liquid capital.
The company plans to reallocate these funds directly into constructing its core advanced HBM manufacturing bases in South Korea, ensuring it has the capital required to scale up its next-generation HBM4 production lines and protect its near-monopoly supply agreements with Nvidia, further solidifying its competitive moat against Chinese competitors.
Securing the Future of Global Computation
The clear-eyed analysis presented by Baillie Gifford’s Paulina McPadden provides a vital, much-needed reality check for the global investment community. While the geopolitical trade war has forced China to build a parallel, highly functional domestic semiconductor ecosystem, the technical bottlenecks, high operational costs, and low yield rates of DUV multi-patterning make Chinese foundries structurally unattractive investments for global capital.
By focusing on near-monopoly giants like TSMC and SK Hynix, long-term investors can place their capital in the companies that possess the absolute technological and manufacturing lead in the AI era.
As TSMC scales its 2-nanometer production and SK Hynix reallocates its international capital to expand its advanced HBM4 lines, these two Asian titans will continue to set the pace for global technological innovation, proving that the ultimate winners of the digital age will be the companies that can successfully manufacture, scale, and deliver the physical engines of computation to the global marketplace.
