The global semiconductor supply chain is experiencing significant pressure, driven by increasing AI demand and ongoing capacity limitations at leading foundries like TSMC.
Illustration by IMF Alpha editorial · Reviewed by IMF Alpharoom AI
The semiconductor industry is grappling with persistent supply chain constraints, primarily stemming from the surging demand for artificial intelligence (AI) chips. This has placed considerable strain on leading chip manufacturers, particularly Taiwan Semiconductor Manufacturing Co. (TSMC), which is a critical supplier for many AI-focused companies.
TSMC, a dominant player in advanced chip manufacturing, reported a 16.5% year-over-year increase in revenue for April 2024, reaching approximately $7.2 billion USD. Despite this growth, the company's Q1 2024 earnings call indicated that while demand for AI chips is robust, overall semiconductor market growth remains moderate excluding memory.
The capacity crunch is most acutely felt in advanced process technologies. TSMC's 3-nanometer and 5-nanometer nodes, crucial for high-performance AI accelerators, are operating at high utilization rates. Industry analysts project that new capacity additions, including TSMC's planned fabs in Arizona and Japan, will not significantly alleviate the immediate supply deficit until late 2025 or early 2026.
Companies like Advanced Micro Devices (AMD) and Broadcom (AVGO) rely heavily on TSMC for their AI offerings. AMD's MI300X accelerators, foundational for AI supercomputing, compete directly with Nvidia's offerings and require substantial manufacturing allocation. Broadcom's custom AI accelerators and networking chips also depend on TSMC's advanced process technology, contributing to the demand pressure.
Equipment manufacturers such as ASML Holdings NV (ASML) are critical bottlenecks. ASML, the sole producer of extreme ultraviolet (EUV) lithography machines essential for cutting-edge chip production, faces a backlog extending into 2025. The pace of new fab construction and equipment installation directly impacts the industry's ability to scale up production to meet future AI-driven demand.
The long-term outlook suggests continued investment in capacity expansion across the industry. However, the lead time for constructing new fabs and installing complex equipment means that short-to-medium term supply-demand imbalances are likely to persist, potentially influencing pricing and market dynamics for AI hardware.
As privacy rules tighten and labeling costs skyrocket, companies are betting on synthetic datasets to train models. Here’s who stands to gain — and who might lose.
Smartphones are running larger models locally. That shift reshapes app economics, chips, and financial services in ways investors and developers are only starting to price in.
Cybercriminals are using large language models to craft hyper-personalized lures and voice deepfakes. Defenders can fight back, but speed and strategy matter.