TIVP024: TSMC (TSM): The Most Important Business in the World? w/ Daniel Mahncke & Shawn O’Malley

这期播客节目深入探讨了台湾积体电路制造股份有限公司（简称台积电，TSMC），强调了它在现代世界中的关键作用，并审视了支撑其主导地位的“多重优势护城河”（Lollapalooza moat），以及它所面临的重大地缘政治风险。 播客主持人首先介绍了台积电由创始人张忠谋于1987年开创的革命性商业模式。张忠谋反其道而行，没有遵循当时在公司内部设计、制造和封装芯片的垂直整合常态，而是设想了一个“纯晶圆代工”模式——即*只*为其他公司制造芯片。这一概念最初因涉及宝贵知识产权的外包而被认为是“疯狂的”，但随着芯片制造在摩尔定律的推动下变得指数级昂贵和专业化，它蓬勃发展起来。台积电将制造作为一种服务提供，承担了高资本密集度的负担，并实现了无与伦比的规模和专业知识。 台积电的历史显示出政府支持以及来自飞利浦的早期投资，后者提供了资金和技术诀窍。张忠谋在五十多岁时创立台积电，展现了惊人的信念，他选择用自己的薪水而不是初始股权来认购股份，最终成为了亿万富翁。 该公司的竞争优势是多方面的。如今，台积电在全球尖端芯片生产中占据90%以上的份额，并在整体晶圆代工市场中占据约60%的份额。这个“赢家通吃”的行业以巨大的成本为特征：建造一座最先进的晶圆厂可能超过200亿美元，而每台ASML的极紫外（EUV）光刻系统成本高达1.5亿美元以上。台积电每年超过300亿美元的资本支出（capex）对于维持其技术领先地位至关重要；一旦停止投资，它就会衰落。尽管如此，台积电仍拥有高投资资本回报率（ROIIC），通常高达30%以上，这彰显了其高效运营。 一个关键的差异化因素是台积电卓越的制造良率，其高端节点良率超过90%，而三星仅为60%。这种效率推动了“飞轮效应”：来自较旧、良率较高节点的利润资助了新节点的开发，台积电比竞争对手更快地提升新节点的产能并实现更高的良率，从而吸引更多客户并进一步扩大规模。这种工艺技术领先地位使得台积电引领了向7纳米、5纳米、3纳米以及即将到来的2纳米节点的过渡。较旧的节点虽然不那么尖端，但仍然有利可图，对于汽车芯片等要求不高的应用仍然很重要。 竞争对手难以匹敌台积电。三星的晶圆代工部门在亏损运营，面临较低的良率和产能利用率不足。英特尔尽管承诺向新晶圆厂投入超过1000亿美元，但仍落后数年，其晶圆代工部门蒙受巨额亏损。中国最大的晶圆代工厂中芯国际（SMIC）受到美国制裁的严重限制，无法获得EUV设备，这使得它不可能赶上台积电的先进能力。英伟达（NVIDIA）首席执行官黄仁勋曾盛赞台积电是“世界上最好的……遥遥领先”。 台积电的增长机会强劲。高性能计算（HPC），包括AI加速器，目前占其收入的50%以上，与AI相关的芯片收入在2024年将增长两倍，预计2025年将翻一番。即使是汽车等行业，也越来越需要用于电动汽车（EV）和自动驾驶的先进芯片。这种多元化降低了半导体行业的历史周期性。 最重要的担忧是地缘政治风险。台积电几乎所有的生产都位于台湾，这里是中美之间的一个潜在冲突点。沃伦·巴菲特（Warren Buffett）曾因担忧其“地理位置”而出售了他在台积电的股份。台积电正在全球范围内进行多元化布局，在亚利桑那州（投资超过400亿美元）、德国和日本建设晶圆厂，但复制其复杂的台湾生态系统面临挑战。“硅盾”理论认为，台湾的战略重要性足以阻止军事行动，因为一场冲突将摧毁全球经济，包括中国的经济。 在考虑投资时，主持人指出台积电的市盈率（P/E）较低（19倍），与其22%的收入复合年增长率（CAGR）形成对比，而其依赖者如英伟达（P/E 47倍）和苹果（P/E 32倍）的估值更高，这表明市场对其施加了显著的地缘政治风险溢价。丹尼尔（Daniel）的估值模型，由于台积电的高回报资本支出而采用每股收益（EPS），预测了强劲的收入增长。他采用情景分析来考虑地缘政治风险，估计未来五年每年有5%的完全损失（total loss）概率才能解释台积电目前的股价。 主持人讨论了这种“押注”的恰当性。肖恩（Sean）表示，当第三次世界大战的可能性是一个需要认真考虑的问题时，他对投资感到不安，并指出在快速变化的行业中存在的“创新者困境”风险。然而，他区分了台湾风险和中国风险，认为美国不太可能对台湾实施制裁。两人都同意，台积电卓越的业务质量使得在投资组合中占据小部分（2-3%）可能具有合理性，同时需要密切关注地缘政治发展。

The podcast episode delves deep into Taiwan Semiconductor Manufacturing Company (TSMC), highlighting its crucial role in the modern world and examining the "Lollapalooza moat" that underpins its dominance, alongside the significant geopolitical risks it faces. The hosts begin by introducing TSMC's revolutionary business model, pioneered by founder Morris Chang in 1987. Going against the vertically integrated norm of designing, manufacturing, and packaging chips in-house, Chang envisioned a "pure-play foundry" that would *only* manufacture chips for other companies. This concept, initially deemed "insane" due to the outsourcing of valuable intellectual property, flourished as chip manufacturing became exponentially expensive and specialized, driven by Moore's Law. TSMC offered manufacturing as a service, taking on the burden of high capital intensity and achieving unparalleled scale and expertise. TSMC's history shows government backing and early investment from Philips, which provided capital and know-how. Chang, who was in his 50s when he founded TSMC, demonstrated incredible conviction, acquiring shares with his salary rather than initial equity, eventually becoming a billionaire. The company's competitive advantages are manifold. Today, TSMC commands over 90% of the world's leading-edge chip production and about 60% of the overall foundry market. This "winner-take-all" industry is characterized by immense costs: building a state-of-the-art fab can exceed $20 billion, and each ASML extreme ultraviolet (EUV) lithography system costs upwards of $150 million. TSMC's annual capital expenditure (capex) of over $30 billion is crucial for maintaining its technological lead; if it stops investing, it declines. Despite this, TSMC boasts high returns on invested capital (ROIIC), often in the high thirties, demonstrating its efficiency. A key differentiator is TSMC's superior manufacturing yield, achieving over 90% for high-end nodes, compared to Samsung's 60%. This efficiency fuels a "flywheel effect": profits from older, higher-yield nodes fund the development of newer ones, which TSMC ramps up faster and with better yields than competitors, attracting more customers and further boosting scale. This process technology leadership has seen TSMC lead transitions to 7nm, 5nm, 3nm, and soon 2nm nodes. Older nodes, while less cutting-edge, remain profitable and important for less demanding applications like automotive chips. Competitors struggle to match TSMC. Samsung's foundry division operates at a loss, facing lower yields and underutilization. Intel, despite committing over $100 billion to new fabs, remains years behind, with its foundry division incurring significant losses. SMIC, China's largest foundry, is heavily constrained by US sanctions, unable to acquire EUV equipment, making it impossible to catch up to TSMC's advanced capabilities. NVIDIA CEO Jensen Huang famously praised TSMC for being "the world's best...by a mile." Growth opportunities for TSMC are robust. High-performance computing (HPC), including AI accelerators, now accounts for over 50% of revenue, with AI-related chip revenue tripling in 2024 and expected to double in 2025. Even sectors like automotive are increasingly demanding advanced chips for EVs and autonomous driving. This diversification reduces the historical cyclicality of the semiconductor industry. The paramount concern is geopolitical risk. Almost all of TSMC's production is in Taiwan, a flashpoint between China and the US. Warren Buffett famously sold his stake in TSMC due to concerns over its "location." TSMC is diversifying globally, building fabs in Arizona ($40 billion+ investment), Germany, and Japan, but replicating its complex Taiwanese ecosystem is challenging. The "silicon shield" theory suggests Taiwan's strategic importance deters military action, as a conflict would devastate the global economy, including China's. When considering investment, the hosts note TSMC's low P/E ratio (19) compared to its growth (22% revenue CAGR) and the higher valuations of its dependents like Nvidia (P/E 47) and Apple (P/E 32), suggesting the market applies a significant geopolitical risk premium. Daniel's valuation model, which uses EPS due to TSMC's high-returning capex, projects strong revenue growth. He employs scenario analysis to factor in geopolitical risk, estimating that an annual 5% probability of a total loss for the next five years would justify TSMC's current price. The hosts debate the appropriateness of such a bet. Sean expresses discomfort with investing when the potential for World War III is a serious consideration, and notes the "Innovator's Dilemma" risk in a fast-changing industry. However, he differentiates Taiwan risk from China risk, believing US sanctions are less likely against Taiwan. Both agree that TSMC's exceptional business quality makes a small portfolio position (2-3%) potentially justifiable, while closely monitoring geopolitical developments.