Samsung Receives 4nm AI Chip Orders

According to reports, Samsung Foundry has made significant improvements in its 4nm process yield rate and has secured an AI chip foundry order from a major data center client. However, the semiconductor sector’s recovery is slower than expected, and the overall demand momentum and economic visibility for the second half of 2023 are unclear.

While TSMC has been able to secure high-end GPU-related AI chip orders from Nvidia and AMD, Samsung Foundry was previously unable to compete for these orders. However, recent reports suggest that Samsung’s 4nm yield rate has improved to 75%, enhancing its capability to compete with TSMC. It is also mentioned that Samsung has secured an AI chip foundry order from a data center client, with AMD being mentioned as a possible client.

Despite these developments, it is unlikely that AMD will transfer orders from TSMC to Samsung in the short term due to AMD’s reliance on TSMC’s advanced process technology for its CPUs. However, sources believe that AMD may consider transferring orders to Samsung as a way to affect Nvidia, as TSMC is currently prioritizing production expansion to fulfill Nvidia’s orders.

It is uncertain whether Samsung Foundry can hinder Nvidia with AMD’s support, as Samsung’s HBM (High Bandwidth Memory) capability may not guarantee it the contract to manufacture AI chips for companies like AMD or Google. The capacity bottleneck for high-end AI GPUs and HPC (High-Performance Computing) chips lies in the severe shortage of back-end advanced packaging. TSMC’s advanced packaging technology, known as CoWoS, is highly regarded in terms of yield rate and customer satisfaction, making it the preferred choice for customers.

The success of Samsung Foundry in securing high-end AI GPU chip manufacturing will depend on its ability to present competitive 2.5D or 3D packaging solutions or establish partnerships with high-quality OSAT (Outsourced Semiconductor Assembly and Test) companies. Only by securing advanced packaging capacity will Samsung’s 4nm process have a chance to excel. This factor is likely to be crucial for AI HPC clients when making their decisions.

Source:Digitimes