Driven by AI demand, TSMC is undertaking a historic expansion: equipment procurement needs have doubled.

The explosive growth in demand for AI computing power is pushing the global semiconductor supply chain into an unprecedented period of pressure.

On September 3, it was reported that TSMC executives disclosed that the company's equipment demand had surged 1.9 times from the baseline level in less than eight months, from the end of last year to July of this year, representing a cumulative increase of approximately 90%. Even with TSMC expanding production on an unprecedented scale, with nearly 20 wafer fabs under construction simultaneously worldwide, capacity expansion is still struggling to keep up with demand.

At the same time, the demand for AI is also spreading to other segments such as substrates, further exacerbating the supply-demand imbalance in the supply chain. Statements from executives at TSMC and Unimicron Technology indicate that the pressure from the expansion of AI computing power has extended from advanced processes to advanced packaging, substrates, and related materials, and the semiconductor industry chain is undergoing a profound restructuring.

Equipment demand has nearly doubled in six months, but TSMC's capacity expansion is still struggling to meet demand.

Cliff Hou, Senior Vice President and Co-Chief Operating Officer of TSMC, stated at a recent event that he has never seen demand grow at such a rapid pace and frequency in his 30 years in the industry. Data shows that TSMC's equipment demand rose from 100% at the end of last year to 1.9 times in July of this year, an increase of approximately 90% in just over six months.

To cope with demand pressures, TSMC is expanding its production capacity on an unprecedented scale. Currently, the company is simultaneously building 13 wafer fabs in Taiwan, with another five to six under construction in other parts of the world, totaling approximately 20, far exceeding the four to five fabs built during the same period previously. However, even with this level of expansion, Hou admits that "it is still insufficient to meet demand."

Hou pointed out that AI's continuous pursuit of computing power and energy efficiency is accelerating the upgrading of advanced processes, advanced packaging, and system-level performance. Unlike the relatively linear production model of the past, the industry chain now not only needs to ensure that products can smoothly enter the next manufacturing stage, but also needs to consider the performance matching, manufacturability, and yield between upstream and downstream, which places higher demands on supply chain collaboration.

The supply-demand imbalance in substrates is worsening, with upstream core suppliers becoming a constraint.

The surge in AI demand is also impacting the substrate manufacturing process. Chien Shan-chieh, Chairman of Unimicron Technology, stated that the supply-demand imbalance in the substrate market is further exacerbating. Larger sizes, higher layer counts, and more complex substrate designs are continuously increasing manufacturing difficulty, including ever-increasing requirements for lower coefficients of thermal expansion (CTE) and dielectric constants.

Chien Shan-chieh pointed out that the substrate supply chain is at a critical turning point, with core production tools and materials such as ABF materials and TGV-related technologies heavily reliant on small and medium-sized Japanese suppliers. Over the past 18 months, Unimicron has held more than 10 in-depth meetings with customers and Japanese suppliers to coordinate responses to the increasing supply chain pressures.

It's worth noting that the demand driven by AI isn't limited to advanced manufacturing processes. Jian Shanjie believes that advanced and mature technologies are highly complementary within the AI ecosystem, and emerging AI applications will create new market opportunities for mature nodes, especially in customized peripherals and sensing applications.

He specifically mentioned high-density bonding, specialty memories, logic-based dies, silicon interposers with deep trench capacitors, and silicon photonics, believing that these areas all have the potential to reach mature nodes. In other words, the expansion of AI computing power is extending comprehensively from advanced processes to advanced packaging, substrates, materials, and mature nodes.

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.