In this episode of the Acquired podcast, the founders discuss TSMC's rise to becoming the world's leading semiconductor foundry. Morris Chang, TSMC's founder, details the company's pure-play foundry model, which focused on building trust with customers by avoiding competing with them. Chang shares TSMC's strategies for leveraging economies of scale through high-volume manufacturing and using "learning curve" pricing models.
The hosts explore how TSMC achieved technological leadership by heavily investing in R&D rivaling Intel's capabilities. Chang highlights TSMC's exceptional manufacturing execution, enabling the company to resolve critical yield issues and advance process nodes for clients like Nvidia. The discussion also covers the benefits of TSMC's location within Taiwan's vibrant semiconductor ecosystem, facilitating collaboration with partners and major customers.
Sign up for Shortform to access the whole episode summary along with additional materials like counterarguments and context.
Chang strategically positioned TSMC as a pure-play foundry to avoid competing with customers like Apple. As Chang notes, this non-competing stance let TSMC build trust while serving all customers' demands without overextending capacity.
Chang implemented the learning curve strategy he refined at TI: as production volume increases, costs decrease. TSMC took orders en masse and spread costs through high volume, allowing investment into new process nodes. This flywheel effect drove TSMC's pricing and scaling strategies.
Chang describes balancing Apple's massive orders with avoiding overextension by only partially fulfilling them. TSMC navigated financial decisions carefully to maintain capacity and stability.
Chang felt TSMC was nearly on par with Intel technologically due to TSMC's R&D investment enabled by high volume manufacturing. TSMC advanced process nodes like 20nm to meet demanding customers like Apple and Nvidia.
Chang indicates TSMC surpasses Intel in manufacturing quality, noting TSMC resolved yield issues for Nvidia on the 40nm node, enabling further advancement to benefit clients.
TSMC's operations are concentrated in government-sponsored science parks in Taiwan, facilitating collaboration with partners like Cadence, ARM, and customers like Qualcomm. This vibrant ecosystem and local talent from universities reinforce TSMC's leadership.
1-Page Summary
Nvidia faced challenges with the production yield of their chips manufactured using the 40nm semiconductor manufacturing process at TSMC. Yield issues in this context typically refer to the percentage of working chips produced per wafer, impacting production costs and efficiency. Resolving these yield issues was crucial for Nvidia to ensure a higher number of functional chips per manufacturing run, optimizing their production output and profitability. TSMC's ability to address and improve yield rates on the 40nm node was essential for Nvidia to continue advancing their product offerings and meeting market demands.
TSMC's Foundry Model and Strategy have transformed the semiconductor industry, highlighting the vision and strategic decisions made by its founder, Morris Chang.
Morris Chang's strategic decision to position TSMC as a pure-play foundry rather than an integrated device manufacturer (IDM) has been a critical aspect of its success.
Chang's realization that not competing directly with TSMC's customers would be advantageous led him to reject the IDM route. Chang reflected on TSMC's role in securing most of Apple's 16-nanometer manufacturing requirements, which was possible due to TSMC's lack of direct competition with customers like Apple. This approach built trust with fabless customers, as TSMC focused solely on manufacturing without entering into IDM conflicts that Texas Instruments divided by product types. He believed it was key that the company serve a collective group of customers without segmenting operations that could lead to internal conflicts or compete with customers.
Chang highlighted that TSMC's success was due to not competing with customers since TSMC does not have its own end product division or chip designs. This non-competing stance allowed TSMC to build trust with customers such as Nvidia, which was small at the time they approached TSMC, and Apple, ensuring their demands were met without overextending capacity. Chang emphasized treating all potential and current customers with respect, regardless of their size, to avoid damaging relationships that could limit TSMC's market reach.
TSMC has utilized the "learning curve" to optimally price services and scale production, expanding its dominance in the semiconductor market.
Morris Chang implemented the learning curve strategy he refined during his tenure at TI, understanding that as production volume increases, costs go down, making the company more efficient. Despite other companies structurally dividing themselves, TSMC remained focused on acting as a unified foundry to serve its diverse customer base. Chang's deep understanding of the learning curve influenced the pricing and scaling strategies that allowed TSMC to grow into the largest volume player in the industry.
By taking orders en masse and managing them strategically, TSMC ensured high volume outputs that allowed for cost spread, making it possible to invest the savings into further process node development. This flywheel effect of ...
TSMC's Foundry Model and Strategy
TSMC’s technological leadership and manufacturing capabilities have made it a dominant force in the semiconductor industry, serving demanding clients like Apple and Nvidia.
Morris Chang discussed TSMC's position and capabilities in technology and manufacturing and felt that they were almost at par with Intel in technology, indicating strong R&D capabilities. TSMC can invest more in both [restricted term] and R&D than other companies due to their high volume, which puts them in a superior position relative to their competition. Chang also talked about a significant decision made regarding Apple's orders that impacted TSMC's R&D focus and advancement in semiconductor technology. They had to choose between advancing to the 20 or 16 nanometer node and ended up delaying the 16 nanometer development to fulfill Apple's needs for the 20 nanometer technology.
Chang mentions TSMC's ability to meet the needs of a demanding client in the high-performance chip market while discussing TSMC's relationship with Nvidia, who became one of their top five customers within a few years due to a specific chip's success.
The hosts discuss TSMC's proprietary packaging technology, CO-OS, hinting at their ability to deliver high-performance manufacturing and yield optimization. Chang talked about Apple's surprise request for a 20-nanometer process node instead of the expected 28-nanometer node, which showcases TSMC's capability to serve high-profile clients like Apple.
Chang indicates that TSMC is superior to Intel in manufacturing quality. A yield problem occurred when trying to manufacture high-quality chips on the 40-nanometer node, particularly for Nvidia. Chang returned as CEO to directly address these priority problems, and resolving the manufacturing issue with the 40-nanometer node became one of his main focuses. This resolution was critical for advancing to the next process node (28 nanometers), which would have significant implications for demanding clients.
TSMC's customer trust is highlighted by Chang, suggesting their reliability and quality service as a significant factor why demanding clients like Apple chose TSMC for their chips.
Ben Gilbert and David Rosenthal discuss the geographic ecosystem within the science parks in Taiwan, focusing on TSMC's success due to these parks. David Rosenthal compares the Hsinchu Science Park to Silicon Valley, suggesting it's like if Silicon Valley were all contained within a government-sponsored industria ...
Tsmc's Technological Leadership and Manufacturing Capabilities
Download the Shortform Chrome extension for your browser