Increased demand for artificial intelligence (AI) chips continues to strain the global semiconductor supply chain, with TSMC's advanced packaging capacity a key bottleneck.
Illustration by IMF Alpha editorial · Reviewed by IMF Alpharoom AI
The global semiconductor industry faces persistent supply constraints, primarily driven by surging demand for chips critical to artificial intelligence applications. Taiwan Semiconductor Manufacturing Company (TSMC), a dominant force in advanced chip manufacturing, reported a 50% increase in revenue attributed to high-performance computing (HPC) in its Q4 2023 earnings call, largely reflecting this AI-driven demand.
Analysts at Susquehanna Financial Group indicate that lead times for some advanced logic components have extended in the first quarter of 2024, signaling continued tightness in the supply chain. This is particularly pronounced in AI accelerators, where companies like AMD and Nvidia rely heavily on TSMC's manufacturing prowess.
A significant bottleneck has emerged in advanced packaging, specifically TSMC's CoWoS (Chip-on-Wafer-on-Substrate) technology. This specialized packaging process is essential for integrating high-bandwidth memory (HBM) with advanced processors, a cornerstone for high-performance AI chips. TSMC’s CoWoS capacity was reported to be fully booked through 2025 by company executives during their most recent earnings briefing.
Companies such as Advanced Micro Devices (AMD) are actively working to secure additional packaging capacity for their MI300 series AI accelerators. AMD CEO Lisa Su stated in a recent interview that while wafer supply is generally sufficient, the constraint lies in the availability of advanced packaging. Similarly, Broadcom (AVGO) and Nvidia (NVDA) are also vying for limited CoWoS resources.
Investment by TSMC and its partners continues in an effort to alleviate these bottlenecks. TSMC plans to increase its capital expenditure to between $28 billion and $32 billion in 2024, with a significant portion allocated to expanding advanced process technologies and packaging. Additionally, companies like ASML Holding N.V. (ASML), a key supplier of lithography equipment, are also expanding production to support long-term industry growth.
Despite these efforts, industry projections suggest that the demand for AI chips will continue to outpace supply for the foreseeable future. Data from Gartner estimates that the global AI chip market will grow from $30.8 billion in 2022 to $119.4 billion by 2027, underscoring the sustained pressure on the semiconductor supply chain and advanced manufacturing capacity.
Big banks are trimming yields. Short-term Treasuries, ultra-short ETFs and I Bonds offer alternatives — here’s a practical plan to protect liquidity and returns.
From FICO to machine learning: fintechs promise smarter lending, but consumers and regulators are pushing back. What the shift means for credit, risk and markets.
As money floods AI-focused funds, one chipmaker dominates holdings. That concentration changes the risk profile of a supposedly diversified bet on artificial intelligence.