The news is that to ease the tight production capacity of TSMC, Nvidia is considering outsourcing some AI graphics card orders to Samsung.

Gamingdeputy news on July 5th, according to Digitimes reports, due to the increasingly tight supply of TSMC capacity, Nvidia is considering outsourcing some AI GPUs to Samsung Electronics for manufacturing. Industry watchers pointed out that if Samsung’s 3nm trial products pass the performance verification and its 2.5D advanced packaging technology meets Nvidia’s requirements, Samsung may get some orders from Nvidia.

Korean media Chosun Biz recently quoted local semiconductor industry sources as saying that Nvidia is negotiating with Samsung on chip foundry matters, and the two parties will conduct performance verification discussions based on the most advanced process.

Most industry insiders in South Korea believe that the possibility of Samsung obtaining large orders from Nvidia is not high.However, due to the risk of relying on TSMC alone to meet the needs of all AI graphics card customers, Nvidia will not rule out the option of using Samsung as a second foundry partner.

With the rise of AI-related services, IT companies are actively purchasing GPUs to build AI data centers.Market Watch predicts that the global GPU market size may grow from US$19.711 billion in 2021 (Gamingdeputy Note: currently about 142.313 billion yuan) to around US$33.463 billion in 2028 (currently about 241.603 billion yuan). The growth rate was 7.85%.

Currently, the AI ​​GPU market is dominated by Nvidia and AMD, but Nvidia has more than 90% market share in AI GPUs that drive large language models (LLMs) such as ChatGPT.

Some industry analysts said that in response to the rapid surge in demand for AI GPUs, Nvidia may work with Samsung to mitigate the risk of supply shortages, although Samsung’s 3nm manufacturing process has not yet achieved stable yield performance. Therefore, in addition to the performance verification of the 3nm test chip, whether Samsung’s advanced packaging technology can meet Nvidia’s requirements is another key factor in determining the potential cooperation between the two parties.

In order to upgrade packaging technology, Samsung just established the multi-chip integration alliance (MDI) in June, seeking to expand the ecosystem of partners, continuously improve stacking technology and actively invest in 2.5D and 3D packaging technology. In contrast, TSMC has been cultivating 2.5D CoWoS packaging technology for many years, successfully using this technology to improve the computing performance of chips, and becoming the preferred foundry partner of AI and HPC chip manufacturers.

In the past, Samsung has also completed graphics card orders for Nvidia, including the production of the GeForce RTX30 series using an 8nm process in 2020, and before that, it also manufactured 14nm mobile GPU chips. However,All of Nvidia’s current flagship A100 and H100 GPUs that support ChatGPT and other AI applications are manufactured by TSMC.