TSMC Founder Morris Chang

以下是内容的中文翻译： 在这期特别的Acquired节目中，主持人Ben Gilbert和David Rosenthal在台北与台积电（TSMC）93岁的创始人张忠谋博士进行了一次对话。在多年研究半导体行业后，他们成功采访了张忠谋，重点讨论了他自传中的关键故事，特别是那些涉及苹果、英伟达（NVIDIA）以及无晶圆厂（fabulous）产业诞生的故事。 对话从张忠谋回忆他与英伟达首席执行官黄仁勋（Jensen Huang）的关系开始。1997年，黄仁勋给他发了一封信，表达了在圣何塞办事处没有得到回应后，英伟达希望与台积电合作的愿望。张忠谋对此很感兴趣，于是会见了黄仁勋。黄仁勋的口才和乐观精神给他留下了深刻的印象，尽管当时英伟达的财务状况不佳。黄仁勋预测他们的新芯片将拯救公司，并使英伟达成为台积电的主要客户，这对于当时台积电的规模来说是一个大胆的声明。他的预测成为了现实，英伟达很快成为台积电最大的客户之一。 讨论转向了2009年台积电面临挑战的时期，当时40纳米制程的制造和质量出现问题。重新担任首席执行官的张忠谋发现，前任首席执行官依赖有缺陷的质量评估，没有为延误和质量问题向英伟达提供任何赔偿。更糟糕的是，公司的定价策略也跟不上制造成本。在重新担任首席执行官后，张忠谋立即致电黄仁勋。 张忠谋还谈到了一场雇佣危机，前任首席执行官根据绩效评估解雇了数百名员工，这与台积电的理念背道而驰。这些裁员引发了抗议，甚至蔓延到张忠谋的住所，促使他迅速采取行动。张忠谋还分享了他的妻子张淑芬如何向抗议者分发早餐，希望能够和平解决。 为了解决与英伟达的纠纷，张忠谋联系了黄仁勋，提出了一项超过1亿美元的重大和解方案。他提出了一个48小时的最后期限，强调其公平性并避免仲裁。黄仁勋接受了该提议，巩固了两家公司之间牢固的合作关系。 对话转向了台积电在40纳米问题后决定积极推进28纳米制程。他提到自己推动大幅增加研发预算，将其设定为营收的8%，这一举动赋能了研发团队进行创新，并提出了将28纳米制程作为“最佳点”的建议。尽管董事会对增加资本支出表示担忧，但张忠谋做出了完全致力于28纳米的战略决策，从而在智能手机领域取得了成功，开启了智能手机时代。 张忠谋回忆了2010年在富士康的郭台铭促成的一次与苹果的会面。苹果的Jeff Williams提议台积电成为代工厂，但针对的是20纳米制程，这是对计划中的16纳米制程的半步偏离。尽管这是一个偏离，张忠谋同意了，但限制了资本支出。 张忠谋分享了一个关于与高盛接洽以帮助为该项目融资的故事。他们最终决定高盛和台积电将发行公司债券，但台积电只会完成苹果订单的一半。 对话发生转折，苹果考虑英特尔作为代工厂，导致讨论暂停。张忠谋毫不畏惧，拜访了苹果，蒂姆·库克向他保证，强调英特尔缺乏代工专业知识以及客户信任的重要性。最终，台积电获得了苹果的业务，但三星设法在他们之前成为了16纳米供应商。 张忠谋博士接着谈到了高通和IBM的故事。当时，高通与IBM有一份合同，台积电对此有所了解。他知道他们是台积电客户的主要候选人。但随着时间的推移，他发现高通的主要代工厂现在是台积电的代工厂，所以他认为IBM现在有麻烦了。此后不久，IBM提出与台积电共同开发工艺技术。但台积电拒绝了，IBM转向了台积电的竞争对手联电（UMC），张忠谋笑着说联电会后悔这个提议。 张忠谋解释了学习曲线理论，强调了产量和经验在降低成本和获得竞争优势方面的重要性。他将自己在德州仪器的经历归功于他对这一理论的理解和应用。 最后，张忠谋反思了台积电出人意料的成功，承认虽然其规模和重要性超出了他的最初预期，但他始终相信其潜力。他驳斥了台积电将在Fab 2停下来的说法，理由是他相信学习曲线。他看到了ARM等无晶圆厂公司的诞生，市场通过摩尔定律要求更高的计算能力，所有这些都为台积电的成功做出了贡献。 在会后讨论中，Ben和David强调了台积电纯代工模式的精妙之处，其在赋能无晶圆厂公司方面的整体作用，新竹科学园区的地理重要性以及学习曲线的持续相关性。

In this special episode of Acquired, hosts Ben Gilbert and David Rosenthal present a conversation with Dr. Morris Chang, the 93-year-old founder of TSMC, recorded in Taipei. After years of studying the semiconductor industry, they secured an interview with Chang, focusing on key stories from his autobiography, particularly those involving Apple, NVIDIA, and the birth of the fabulous industry. The conversation begins with Chang recounting his relationship with Jensen Huang, the CEO of NVIDIA. It started in 1997 when Huang sent him a letter expressing NVIDIA's desire to partner with TSMC after their San Jose office was unresponsive. Intrigued, Chang met Huang, who impressed him with his articulateness and optimism, despite NVIDIA's financial difficulties at the time. Huang predicted that their new chip would save the company and make NVIDIA a major TSMC customer, a bold statement given TSMC's size. His prediction came true and NVIDIA would soon become one of TSMC's biggest customers. The discussion shifts to a challenging period in 2009 when TSMC experienced manufacturing and quality issues with the 40-nanometer node. Chang, who had stepped back into the CEO role, discovered that the previous CEO, relying on flawed quality assessments, had not offered NVIDIA any compensation for the delays and quality problems. Compounding the issue, the company was also struggling with a pricing strategy that wasn't keeping pace with manufacturing costs. Upon reassuming the CEO position Chang called Jensen immediately Chang also addressed an employment crisis where the previous CEO had laid off hundreds of employees based on performance reviews, a practice against TSMC's ethos. These layoffs sparked protests, even reaching Chang's home, prompting him to take swift action. Chang then shares how Sophie Chang's wife distributed breakfast to the protesters in hopes of having a peaceful resolution. To resolve the NVIDIA dispute, Chang contacted Jensen, offering a significant settlement exceeding $100 million. He presented the offer with a 48-hour deadline, emphasizing its fairness and avoiding arbitration. Jensen accepted the offer, solidifying the strong partnership between the two companies. The conversation moves to TSMC's decision to aggressively pursue the 28-nanometer node after the 40-nanometer problems. He mentions his push to significantly increase the R&D budget, setting it at 8% of revenue, a move that empowered the R&D team to innovate and propose the 28-nanometer node as the "sweet spot." Despite board concerns about increased capital expenditure, Chang made the strategic decision to fully commit to 28-nanometer, leading to success with smartphones which was the smartphone era. Chang recounts a meeting with Apple in 2010 facilitated by Terry Gou of Foxconn. Jeff Williams from Apple proposed TSMC as a foundry, but for 20-nanometer, a half-step detour from the planned 16-nanometer progression. Although it was a detour, Chang agreed but limited the capital expenditure. Chang shares a story about approaching Goldman Sachs to help finance the venture. They eventually decided that Goldman Sachs and TSMC would issue corporate bonds, but TSMC would only fill half of Apple's order. The conversation takes a turn when Apple considered Intel as a foundry, leading to a pause in discussions. Chang, unfazed, visited Apple, where Tim Cook reassured him, emphasizing Intel's lack of foundry expertise and the importance of customer trust. Ultimately, TSMC secured Apple's business, but Samsung managed to become a 16-nanometer supplier ahead of them. Dr. Chang then talks about the Qualcomm and IBM story. Qualcomm at the time had a contract that TSMC knew about with IBM. He knew they were a prime candidate to be TSMC's customer. But over time he found out that Qualcomm's main foundry was now TSMC's foundry, so he figured that IBM was now in trouble. Shortly afterwards, IBM would come in with an offer for TSMC to co-develop process technology. But TSMC passed and IBM went to TSMC's rival, UMC, in which Dr. Chang would laugh about how UMC would regret this offer. Chang explains the learning curve theory, emphasizing the importance of volume and experience in reducing costs and gaining a competitive advantage. He credits his time at Texas Instruments for refining his understanding and application of this theory. Wrapping up, Chang reflects on TSMC's unlikely success, acknowledging that while its scale and importance exceeded his initial expectations, he always believed in its potential. He dismisses the notion that TSMC would stop at Fab 2, citing his belief in the learning curve. He saw the birth of the fabulous companies such as ARM and the market demanded more computing power through Moore's Law, all of which has contributed to TSMC's success. In a post-game discussion, Ben and David emphasize the genius of TSMC's pure-play foundry model, its integral role in enabling fabulous companies, the geographic importance of Hsinchu Science Park, and the continued relevance of the learning curve.