GUC: ASIC Shipments to Surpass GPU Growth; 2nm Projects Maintain Healthy Momentum
IC design service provider Global Unichip Corp. (GUC) predicts that as AI computing shifts from model training to end-device applications, Application-Specific Integrated Circuits (ASICs) will surpass Graphics Processing Units (GPUs) in shipment growth due to their higher power efficiency in inference. GUC expects its operations to benefit, with advanced process projects like 2nm and 3nm maintaining healthy momentum.
📋 Article Processing Timeline
- 📰 Published: May 1, 2026 at 14:27
- 🔍 Collected: May 1, 2026 at 14:31 (4 min after Published)
- 🤖 AI Analyzed: May 1, 2026 at 17:11 (2h 39m after Collected)
(Hsinchu, May 1st, Central News Agency reporter Chang Chien-chung) IC design service provider Global Unichip Corp. (GUC) stated that as AI computing shifts massively from model training to end-device applications, Application-Specific Integrated Circuits (ASICs) possess a higher power efficiency ratio in the inference stage. Therefore, their shipment growth rate will surpass that of Graphics Processing Units (GPUs), and GUC's operations are expected to benefit, with advanced process projects such as 2nm and 3nm maintaining healthy momentum.
GUC is scheduled to hold its annual shareholders' meeting on May 21st. Its business report indicates that after Google's Gemini 3 is released at the end of 2025, it will surpass competitors in various benchmark tests. The core key lies in Google's self-developed Tensor Processing Units (TPUs), which have high power efficiency and cost advantages.
GUC believes it can benefit from this wave of self-designed chips, providing CoWoS and design support to customers to help them develop AI acceleration chips. It anticipates that advanced process projects such as 2nm and 3nm will maintain healthy momentum.
Amid the trend of major cloud service providers accelerating the development of self-designed chips, GUC stated that to shorten time-to-market, customers highly rely on mature and quickly deployable silicon intellectual property (IP).
GUC pointed out that with the rapid increase in demand for data throughput in AI and high-performance computing, high-speed connections within and between chips have become critical, driving the adoption and integration demand for high-speed transmission IPs like HBM4E physical layer (PHY) and controllers.
GUC stated that it will continue to work closely with wafer foundries and back-end packaging factories. By enhancing advanced process verification, IP expansion, and design platform strengthening, it will improve the efficiency of customer project development and enhance overall competitiveness.
GUC pointed out that future ASIC design services are no longer just simple design and wafer capacity provision. They also require incorporating heterogeneous packaging integration, system architecture testing, improving computing power efficiency, and resource integration. The ASIC camp, leveraging the advantages of customization, power saving, and fast time-to-market, will usher in a period of rapid growth in customized computing power this year.
GUC expects that revenue from design services and integrated manufacturing services for chips will continue to grow this year. Projects related to cloud services, optical communication transmission, and automotive electronics will contribute to revenue. (Edited by Lin Shu-yuan) 1150501
GUC is scheduled to hold its annual shareholders' meeting on May 21st. Its business report indicates that after Google's Gemini 3 is released at the end of 2025, it will surpass competitors in various benchmark tests. The core key lies in Google's self-developed Tensor Processing Units (TPUs), which have high power efficiency and cost advantages.
GUC believes it can benefit from this wave of self-designed chips, providing CoWoS and design support to customers to help them develop AI acceleration chips. It anticipates that advanced process projects such as 2nm and 3nm will maintain healthy momentum.
Amid the trend of major cloud service providers accelerating the development of self-designed chips, GUC stated that to shorten time-to-market, customers highly rely on mature and quickly deployable silicon intellectual property (IP).
GUC pointed out that with the rapid increase in demand for data throughput in AI and high-performance computing, high-speed connections within and between chips have become critical, driving the adoption and integration demand for high-speed transmission IPs like HBM4E physical layer (PHY) and controllers.
GUC stated that it will continue to work closely with wafer foundries and back-end packaging factories. By enhancing advanced process verification, IP expansion, and design platform strengthening, it will improve the efficiency of customer project development and enhance overall competitiveness.
GUC pointed out that future ASIC design services are no longer just simple design and wafer capacity provision. They also require incorporating heterogeneous packaging integration, system architecture testing, improving computing power efficiency, and resource integration. The ASIC camp, leveraging the advantages of customization, power saving, and fast time-to-market, will usher in a period of rapid growth in customized computing power this year.
GUC expects that revenue from design services and integrated manufacturing services for chips will continue to grow this year. Projects related to cloud services, optical communication transmission, and automotive electronics will contribute to revenue. (Edited by Lin Shu-yuan) 1150501