Guotai Haitong Securities has released a research report indicating that with the rapid development of artificial intelligence, the firm is optimistic about the swift growth in demand for related testing equipment driven by AI. This includes: ① Rapid growth in the global AI computing testing equipment market, reaching $2.3 billion in 2024; ② Increased complexity of HBM products creating new HBM testing demands; ③ Development of super-node technology leading to growing demand for server motherboard testing equipment.

Artificial intelligence represents a new industrial revolution. According to NVIDIA founder Jensen Huang's projections, global artificial intelligence infrastructure spending is expected to reach $3-4 trillion by 2030. The firm believes that AI development will drive rapid growth in demand for testing equipment for SoCs, HBM, power management chips, and server motherboard-level testing equipment, presenting new growth opportunities for related companies.

The global AI computing testing equipment market is experiencing rapid growth, reaching $2.3 billion in 2024. Integrated circuit production requires WAT testing, CP testing, and FT testing, where WAT and CP testing primarily target wafer testing before packaging processes, while FT testing focuses on post-packaging chip testing. As chip complexity and integration levels continue to rise, the importance of testing becomes increasingly prominent. According to SEMI data, the global integrated circuit testing equipment market size was $75.4 billion in 2024 and is expected to reach $97.7 billion by 2026, representing a 29.58% year-over-year growth. Based on content from Teradyne's Q2 2025 earnings call, the global testing equipment leader estimates the 2024 global AI computing testing equipment market (including VIP ASIC, Merchant GPU, etc.) at $2.3 billion, with continued growth expected.

HBM product iterations are bringing continuously increasing testing demands. Benefiting from strong HBM demand from AI chip customers like NVIDIA, SK Hynix holds a leading position in the HBM market. On September 12, 2025, Hynix announced successful completion of HBM4 development, a new ultra-high-performance memory product for AI, and established the world's first mass production system. With HBM product iterations, HBM memory stacks are evolving from 8-layer DRAM chip compositions to 12 layers, plus one base die. Previously, each DRAM chip and base die required wafer-level testing, as well as post-stack wafer testing after DRAM chip and base die stacking. However, due to HBM product quality issues that often cause problems when HBM products are packaged and assembled with AI acceleration chips, affecting final yields, some companies are considering adding an additional testing step after HBM stack wafers are diced into individual HBM stacks to improve quality.

Server testing equipment demand is growing rapidly. The rapid growth in AI large model parameter scales brings demands for massive computing power and memory resources, giving rise to super-node technology. Using NVIDIA's NVLink-based NVL72 solution as an example, the entire system consists of 18 Compute Trays and 9 Switch Trays, where each Compute Tray contains 2 GB200 superchips, with each GB200 superchip having 2 Blackwell series B200 GPUs. Each Switch Tray contains 2 NVLINK Switch chips, and the entire cabinet's rear section uses cables to interconnect Compute Trays and Switch Trays. As cabinets become increasingly complex, Guotai Haitong Securities believes that to ensure proper cabinet operation, servers must undergo a series of tests including but not limited to ICT, FCT, burn-in, SIT, performance, and compatibility testing. The importance of related testing equipment suppliers is gradually becoming prominent, with rapidly growing demand.