TSMC’s CapEx Delay: Analyst Warns Foundry’s Impact on AI Development Could Result in Billions in Lost Revenue for Hyperscalers

According to insights from Stratechery, TSMC has emerged as the largest “risk”in the artificial intelligence (AI) supply chain. The leading semiconductor manufacturer reportedly underestimated the surge in chip demand earlier this decade, resulting in significant supply constraints.

The Impact of TSMC’s Early Skepticism on the AI Industry

The current landscape of the AI supply chain underscores TSMC’s pivotal role, particularly given its extensive foundry services and its crucial support for the rapid buildout of hyperscaler infrastructure. The surge in high-performance computing (HPC) orders has not only cemented NVIDIA’s position as TSMC’s largest client but has also eclipsed Apple’s longstanding dominance as a major customer. Despite TSMC’s recent capital expenditure (CapEx) increases and ambitions for expansion, Stratechery highlights that the company poses a “risk”to the AI sector, primarily due to supply chain imbalances rather than geopolitical tensions.

That lack of increased investment earlier this decade is why there is a shortage today, and is why TSMC has been a de facto brake on the AI buildout/bubble.

– Stratechery

The analysis indicates that TSMC’s hesitance regarding elevated CapEx has largely contributed to bottlenecks in its production lines. Under the leadership of C. C.Wei, TSMC had exhibited “mixed emotions”toward the hyperscaler growth until more recently. The company announced plans to escalate its spending to $56 billion this year, following confirmation from its CEO regarding the legitimacy of financial commitments surrounding the hyperscaler developments.

A bar chart titled 'TSMC Capital Expenditure (2022-2026)' shows an increase from $36.3B in 2022 to an estimated $54.0B in 2026.

As companies like NVIDIA and AMD experience extended delivery timelines due to TSMC’s supply limitations, other players in the custom silicon arena—such as Microsoft, Google, Meta, and more—struggle to place orders that can ensure timely deliveries. Microsoft’s recent showcase of the Maia 200 AI chip, utilizing TSMC’s N3B fabrication process, illustrates the severe supply shortages impacting production. Stratechery points out that, ultimately, the “risk”will translate into lost revenue for the hyperscalers.

In addition to semiconductor challenges, the advanced packaging sector is also facing critical bottlenecks. TSMC’s CoWoS technology is currently the leading option for chip manufacturers; however, its production capacity pales in comparison to that of semiconductors, resulting in substantial constraints. Given the increasing importance of advanced packaging within the AI supply chain, TSMC needs to ramp up its responsiveness to meet rising demand.

TSMC reportedly introducing price hike for 2nm wafers for four consecutive years.

The analyst also elaborates on the risks associated with diversification among foundry services for high-performance computing (HPC) clients and ASIC manufacturers. TSMC has established a uniquely irreplaceable supply chain and trust. Although alternatives like Intel Foundry and Samsung are garnering interest, AI chip manufacturers perceive outsourcing orders away from TSMC to be a significant gamble. In the long term, companies may face tough decisions on whether the potential loss of “billions in revenue”outweighs their reliance on the Taiwanese giant.

Source & Images