The semiconductor industry is facing a new logistical challenge driven by the unstoppable rise of artificial intelligence, which has already caused prices for key components to soar. According to reports from the supply chain, TSMC now has its advanced packaging production lines fully booked, forcing the company to change its usual strategy.
To meet the orders of its most important clients, the foundry has begun outsourcing part of this workload to specialized external companies.
Growing Pressure on Advanced Packaging
The bottleneck lies specifically in CoWoS technology (Chip-on-Wafer-on-Substrate), a crucial process for assembling high-performance chips used by companies like NVIDIA, AMD, Google, Amazon, and Apple. This method allows multiple chiplets to be combined, increasing processing power and making it a cornerstone of AI hardware.
However, TSMC’s internal capacity for CoWoS-L and CoWoS-S variants has reached its physical limits, unable to take on more orders in the short term.
Faced with this saturation, TSMC has turned to Taiwanese assembly and testing firms to handle the overflow. Companies such as ASE Technology and SPIL have emerged as the main beneficiaries of this surge in demand.
Analyst reports, including those from Counterpoint, suggest that these companies will see a strong positive impact on their operations by 2026 as they take on the workload TSMC cannot handle itself.
A Strategic Move to Stay Ahead of Competitors
This decision to outsource is not only a short-term logistical fix but also a strategic move in the face of growing competition. With rivals like Intel quickly developing their own advanced packaging capabilities, TSMC must ensure that it meets customer demand without delay. There are even rumors that companies such as Apple and Qualcomm are considering Intel’s alternatives, making it crucial for TSMC to guarantee supply and avoid losing clients.
While TSMC continues investing billions to expand its facilities in both Taiwan and Arizona—the latter still lacking on-site advanced packaging capabilities—building new production lines takes time. The company is exploring the possibility of converting future manufacturing phases, such as the P6 plant, to speed up the process.
Until then, partnering with external firms remains the most practical solution to keep pace with the accelerating AI market.
