Research firm praises TSMC as "the world's most fair and just company," noting it could have leveraged the AI boom to raise prices and earn more.

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1868 words

According to the latest analysis by the AI research firm Semi Analysis, TSMC's N3 process has become the biggest bottleneck for global AI computing power. NVIDIA, Broadcom, and AMD are all competing for capacity, yet TSMC's pricing remains conservative and restrained. (Context: The Financial Supervisory Commission has introduced the "TSMC Clause," relaxing the single-stock holding limit for Taiwan stock funds to 25%, potentially injecting up to 200 billion TWD into the market.) (Background: We want TSMC-level salaries too! 37,000 Samsung employees are protesting against the "15% chip profit sharing" policy, threatening a general strike.) The world's most advanced chip manufacturing process is in short supply, yet TSMC, which controls this process, has chosen not to raise prices. AI semiconductor research firm Semi Analysis recently published a long-form analysis, offering a series of interconnected market observations ranging from TSMC's pricing strategy and the structure of the AI computing power market to NVIDIA's next steps. In its report, Semi Analysis points out that TSMC's N3 (3nm, currently the most advanced process node) has become the biggest bottleneck in the global AI computing power supply chain. Major accelerator manufacturers such as NVIDIA, Broadcom, Annapurna, MediaTek, and AMD have almost all centered their 2025 and 2026 product roadmaps on the N3 process, leading to intense competition for capacity. However, TSMC's quotes remain relatively stable. Semi Analysis describes TSMC as "the most fair and righteous company in the world"—not as a sarcasm, but as an affirmation tinged with surprise: despite holding such strong bargaining power on the supply side, TSMC chooses not to maximize its pricing advantage. Semi Analysis analyzes that TSMC's long-term strategic logic has been to protect profits through stable pricing during economic downturns, at the cost of compressed growth space during economic expansions. This strategy had its rationale in past semiconductor cycles, but in the current era of rapid AI computing power demand growth, it has created significant opportunity costs. The report further cites a public statement made by NVIDIA CEO Jensen Huang in 2024: he believes TSMC should raise wafer prices. This is uncommon in the industry. When a customer proactively calls for a supplier to raise prices, it indicates that the scarcity of N3 supply has reached a critical point where buyers prioritize "supply chain security" over "cost control." Semi Analysis proposes two paths for TSMC: first, adopt a more aggressive pricing strategy, which customers would accept; second, lock in strategic customers through long-term agreements combined with guaranteed capacity and prepayments, which would also strengthen financial resilience without significant price hikes. Semi Analysis uses one sentence to describe the degree of time compression in the current AI industry: "A day in AI is a year in other industries." The cycles for model releases, software breakthroughs, and hardware iterations have been compressed from years to weeks. Against this backdrop, the value distribution in the computing power market is shifting rapidly. The report notes that over the past few months, Agentic AI (AI systems capable of autonomously executing tasks) has crossed a true turning point, driving a leap in the market value of tokens (the units of text processed by AI). In numerical terms, this shift is already reflected in the hardware market: - Memory prices have risen 6-fold over the past year. - The price of one-year H100 lease contracts has rebounded 40% from its October 2025 low. Semi Analysis believes that AI labs are the biggest beneficiaries in the current market—having captured almost no value last year, they are now rapidly accumulating pricing power and influence. This shift simultaneously drives upstream demand for computing power, making the already tight N3 capacity even more constrained. It is worth noting that in this computing power supply chain, AI computing power buyers like Anthropic are in a rather passive position. Semi Analysis points out that because NVIDIA possesses the asymmetric procurement capability to secure upstream supply (especially TSMC wafers), computing power demanders like Anthropic are effectively "pushed" into the NVIDIA ecosystem rather than choosing it freely. Semi Analysis believes that only TSMC and NVIDIA currently possess true, strong pricing power in the industry, but both have been relatively slow to react. Compared to TSMC, NVIDIA's next move is more worthy of attention. The report points out that NVIDIA's current strongest competitive advantage lies not in technical specifications, but in procurement capability. Being able to obtain a disproportionate share of restricted upstream supply, including TSMC N3 wafers, is key. In a Fabless model, whoever can lock in capacity controls the market. Semi Analysis predicts that NVIDIA's pricing strategy is about to undergo a structural shift: moving from a cost-based pricing model to "value-based pricing"—that is, pricing

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching1 items

2026-05-03

Research firm praises TSMC as "the world's most fair and just company," noting it could have raised prices to profit more from the AI boom.

Similarity 180%關鍵字世界最公平正義公司/研調讚台積電/明明可趁

💡 Currently matching via keywords + symbols (MVP) · Will be upgraded to embedding semantic search later

Raw Information

ID:b8c6f981ab

Source:動區 BlockTempo

Published:2026-05-03 06:33:45

Category:zh_news · Export Category zh

Symbols:Unspecified

Community Votes:+0 / −0 · ⭐ 0 Important · 💬 0 Comments