A recent report that Doosan Tesna will test AI chips for NVIDIA marks a pivotal moment for the company's transformation.
The story begins with the explosive demand for AI. At its GTC 2026 conference, NVIDIA signaled a demand pipeline potentially worth a trillion dollars for AI hardware. This unprecedented demand for its new chips, like the Blackwell and Vera Rubin platforms, created immense pressure on its entire supply chain, especially in the stages after the silicon wafer is made.
This leads us to the core problem: a bottleneck in advanced packaging. Specifically, TSMC's cutting-edge CoWoS (Chip-on-Wafer-on-Substrate) technology, which is essential for assembling high-performance AI chips, couldn't keep up. With reports suggesting NVIDIA had already claimed about 70% of TSMC's 2025 CoWoS capacity and demand still exceeding supply, NVIDIA had to find alternative partners for crucial final steps like testing.
This is where Korea enters the picture. The Korean government and major corporations, in partnership with NVIDIA, announced the 'K-AI Infrastructure' initiative. This strategic alignment encouraged NVIDIA to look more closely at the Korean semiconductor ecosystem—renowned for memory, but also home to strong back-end or OSAT (Outsourced Semiconductor Assembly and Test) companies. Diversifying to Korea became a logical step to de-risk its supply chain.
Doosan Tesna was perfectly positioned to capitalize on this opportunity. For years, the company had been strategically preparing to shift away from its heavy reliance on the volatile mobile and image sensor (CIS) markets. It made a massive ₩171.4 billion investment in new test equipment and announced a new factory in Pyeongtaek. This forward-thinking capital expenditure was precisely for handling the complex, high-performance SoCs that power AI. In essence, Doosan Tesna built the capacity before the client knocked on the door.
Therefore, this deal isn't a stroke of luck. It's the result of a powerful convergence: surging AI demand, a critical manufacturing bottleneck, and Doosan Tesna's timely strategic investments. By securing NVIDIA, the company not only diversifies its revenue but also solidifies its position as a key player in the global AI supply chain.
- Glossary
- CoWoS (Chip-on-Wafer-on-Substrate): An advanced 2.5D packaging technology used to integrate multiple chips on a single substrate, crucial for high-performance computing and AI chips.
- OSAT (Outsourced Semiconductor Assembly and Test): A company that provides third-party services for semiconductor packaging and testing.
- AI Accelerator: Specialized hardware designed to speed up artificial intelligence and machine learning computations, far more efficient than general-purpose CPUs for these tasks.