Semiconductor leaders warn of capital costs and technical gaps in AI growth

Semiconductor executives are warning that high capital costs and a lack of software integration could hinder the long-term expansion of artificial intelligence. At the DBS Asian Insights Conference in Singapore, Broadcom CEO Hock Tan noted that while supply chain issues like memory shortages and power constraints may ease by 2028 or 2029, rising borrowing costs and the need for clear returns on investment (ROI) pose significant risks to infrastructure expansion.

Industry leaders suggest that if AI models fail to demonstrate commercial viability or if performance gains plateau, the current surge in capital expenditure could face a major correction. While demand for computing power is expected to remain tight through 2030, the transition from hardware competition to software and system integration is becoming critical.

In a related industry warning, Etron Technology Chairman Lu Chao-qun cautioned that Taiwan's semiconductor dominance, currently led by TSMC's manufacturing prowess, could be threatened if the region fails to strengthen its capabilities in IC design, system software, and advanced architectures. He highlighted that companies like NVIDIA maintain their edge through software ecosystems like CUDA, and emphasized that the growing market for Application-Specific Integrated Circuits (ASICs) will require significant financial endurance and talent to navigate successfully.