Taiwan’s chip testing providers invest as AI raises demand

Taiwan’s semiconductor testing providers are reporting stronger sales and profits as AI investment increases both chip production and the work needed to verify each device. ASE, KYEC, Powertech Technology, Sigurd, and ChipMOS are responding with new facilities, equipment and capabilities, ranging from capacity for leading AI processors to services for memory and optical communications.
These companies test chips for designers and manufacturers. Wafer probing identifies defects before individual chips are separated and packaged. Final testing evaluates completed devices, burn-in exposes potential early failures through electrical and thermal stress, and system-level testing assesses performance under conditions closer to actual use.
AI makes this work more demanding. Accelerators combine processing chips and high-bandwidth memory in expensive packages, so one defective component can waste other valuable parts. Faster signals and higher power consumption require more sophisticated testing equipment and temperature control. The commercial opportunity grows with both shipment volumes and the complexity of each device.
ASE Technology Holding combines testing with a broad packaging business that includes Siliconware Precision Industries, or SPIL. Third-quarter 2026 group revenue rose 44.8% year on year to NT$244.16 billion (approximately US$7.44 billion), while net profit increased 148% to NT$27 billion (US$823 million). Those figures include electronics manufacturing as well as packaging and testing.
The group’s strategy is to handle more of the production sequence for advanced chips. Its leading-edge assembly and testing business, known as LEAP, is forecast to generate more than US$4 billion in 2026 and double in 2027. ASE raised its annual capital expenditure budget to US$11.5 billion, with the additional US$1 billion divided between facilities and equipment. Factory space remains a constraint, and management is prioritizing wafer probing within its testing investments.
KYEC is far more concentrated, with testing accounting for approximately 98% of its revenue across high-performance computing, communications, and automotive electronics. Third-quarter revenue rose 30.7% to NT$12.14 billion (US$370 million), while net profit attributable to shareholders increased 14.9% to NT$2.64 billion (US$81 million).
Its expansion also shows how testing capacity is becoming a strategic concern for customers. Nvidia announced a US$1.1 billion capacity-reservation advance to support KYEC’s new US testing facilities and equipment, with high-volume production targeted for April 2027. The arrangement helps finance KYEC’s investment while securing future capacity for Nvidia. Across its suppliers, Nvidia expects the testing capacity available for its own chips to double between 2026 and 2027 as sites expand in the United States, Taiwan, and Singapore.
Powertech Technology benefits from AI through memory and storage, packaging, and testing products used in servers and enterprise solid-state drives. Third-quarter revenue increased 26.9% to NT$25.33 billion (US$772 million), net profit attributable to shareholders rose 69.8% to NT$2.61 billion (US$80 million), and gross margin reached 24% on stronger demand and an improving product mix.
The company expects strong DRAM and enterprise storage demand to support the fourth quarter. It is also developing fan-out panel-level packaging, which processes chips on rectangular panels, with customer qualification progressing toward production in 2027. That would extend its role memory services into more complex computing products.
Sigurd Microelectronics is pursuing growth in both AI computing and the connections between computing systems. Third-quarter revenue rose more than 27% to NT$6.17 billion (US$188 million), and net profit attributable to shareholders increased more than 68% to NT$1.35 billion (US$41 million) on demand for testing CPUs, GPUs, custom ASICs, AI accelerators, optical communications chips, and memory.
As data centers adopt more light-based connections, manufacturers must assess optical performance alongside electrical behavior. Sigurd is developing testing capabilities for electronic chips, photonic integrated circuits, and optical engines, while preparing customer products for volume production. It expects AI-related applications to exceed 30% of its business as these programs advance.
ChipMOS brings a different mix of memory and display-driver packaging and testing. Third-quarter revenue increased 36.7% to NT$8.40 billion (US$256 million), while net profit attributable to shareholders reached NT$1.32 billion (US$40 million), compared with NT$352 million a year earlier. Gross margin improved to 21.7% from 12.4%. The company attributed its record quarterly sales to sustained AI demand and a recovering industrial market, and is expanding capacity against customer forecasts and long-term supply agreements.
Nvidia processors require more testing capacity, cloud operators’ custom chips create additional programs, and memory and optical components broaden the work available. Providers with different customer and product portfolios can therefore benefit from the same expansion in AI infrastructure.
Additional machines must atill translate into productive operating hours. Equipment, electricity, cooling, and skilled staff all add costs before new capacity reaches full utilization, which is why commitments such as Nvidia’s reservation agreement with KYEC matter. Taiwan’s providers are increasingly competing on their ability to develop a testing process, qualify it with customers, and deliver enough capacity when a new chip enters production.
US$ conversions use an illustrative exchange rate of NT$32.8 per US$1.
