Can Nvidia’s 75% gross margin be sustained? The key lies in TSMC’s bargaining power.
```
Nvidia’s ultra-high profit margin is not invulnerable—the most overlooked risk by the market actually comes from its closest partner.
Nvidia currently boasts a gross margin of 75%, far surpassing the historical level of around 50% typical of traditional chip manufacturers, and closer to the profit structure of luxury brands or software companies. According to Financial Times columnist June Yoon, this margin is the core variable underpinning Nvidia’s current valuation; whether it can be sustained largely depends on whether TSMC will assert a greater claim to the profits.
June Yoon points out that the market is highly attentive to whether demand for Nvidia’s AI offerings will persist, yet discussion on the risk posed by TSMC’s increasing bargaining power is clearly lacking. Based on current revenue scale, every percentage change in Nvidia’s gross margin translates to roughly $2 billion in annual gross profit—a significant factor that can materially affect earnings forecasts and thus market valuation.
Nvidia stock currently trades at about 21 times forward earnings. June Yoon believes this valuation appears restrained, but it rests on two assumptions: that demand for AI infrastructure remains robust, and that TSMC does not seek to expand its share of profits from each chip—both conditions must be met, and must persist for years.
Nvidia’s Profit Model: Division of Labor Between Design and Manufacturing
Nvidia is often described as a chip design company that captures value from its proprietary software and developer platform, but June Yoon believes this description is incomplete. Nvidia’s revenue comes from selling physical chips, which must be manufactured in fabs it does not own.
Nvidia’s most profitable products—including H200, Blackwell series, and next-generation Rubin architecture—are produced using TSMC’s advanced 4nm and 3nm processes. Currently, no other manufacturer can produce these designs at the same scale, performance, and yield.
This division of labor creates rich returns for both sides: Nvidia focuses on chip architecture and software ecosystem, without bearing the more than $40 billion in annual capital expenditure for fabs; TSMC keeps its most advanced production lines running at full capacity. June Yoon describes this relationship as symbiotic—but it is also Nvidia’s key vulnerability.
TSMC’s Real Leverage: Capacity Scarcity As a Bargaining Chip
In the chip industry, the one who controls advanced manufacturing holds the most sway. TSMC decides pricing for each wafer and how much advanced capacity each client receives. Nvidia is among TSMC’s largest clients, but it is competing with Google, Microsoft, and other global tech giants for the same scarce capacity—these companies are also designing their own AI chips and rely on TSMC’s advanced lines.
The global boom in AI infrastructure construction has left TSMC unable to keep up with demand. June Yoon notes that in this scenario, TSMC has every reason to support Nvidia’s growth, but no obligation to protect its profit margin.
Geopolitical factors further amplify this risk. Any supply disruption in advanced chip manufacturing would directly impact output. While TSMC’s expansion plans in Arizona provide some geographic diversification, the factories still belong to the same company and the power to allocate capacity is unchanged. June Yoon notes that any disruption would make advanced capacity even scarcer, further strengthening TSMC’s bargaining position.
The Boundary of CUDA’s Moat: The Nature of Software Advantage vs. Manufacturing Control
Nvidia’s competitive edge lies not only in chip design but also in the software ecosystem built around its processors. The CUDA platform is the foundational tool for developers programming Nvidia GPUs and is deeply embedded in today’s core AI development processes, making migration costly.
However, June Yoon stresses, this is not the same as controlling advanced manufacturing. Leadership in chip design must be continuously proven with each product cycle; control over core manufacturing infrastructure is far harder to challenge.
She also notes that the entire chip industry revolves around two bottlenecks: ASML controls advanced lithography equipment, and TSMC controls advanced chip manufacturing. Both lack credible substitutes and have extremely high barriers to entry.
Two Preconditions for Valuation—The Market Watches Only One
June Yoon's central insight: Nvidia’s long-term software-like margins rely on a rare industry structure—TSMC consistently chooses to partner rather than maximize profits.
Nvidia’s profits are massive enough that even a modest slide in its margin would have a significant impact on company value. The market remains intensely alert to the sustainability of AI demand, but pays little attention to whether, or when, TSMC will reprice its partnership.
From her perspective, power in this industry follows capacity. And today, capacity belongs to TSMC.
Risk Warning and DisclaimerThe market carries risks; investors should be cautious. This article does not constitute personal investment advice, nor does it take into account any individual user’s specific investment objectives, financial situation, or needs. Users should consider whether any opinion, viewpoint, or conclusion herein is suitable to their specific circumstances. Invest at your own risk. ```