After a year of successful operation, everyone hopes to get an H100, and NVIDIA is now ready to launch the next generation GPU. The company is introducing the GH200 GPU to selected companies for testing and deployment models. Recently, Bindu Reddy, CEO of Abacus AI, a company focused on building large-scale generative AI applications and agents, announced at X that they will receive the GH200 supercomputer today. She said she hopes to focus on open-source projects using AI supercomputers starting in January. The NVIDIA GH200, as the next generation of H100, is expected to be on the market by the end of this year, and systems based on it are expected to start in the second quarter of 2024. In November, AWS announced that it will be the first customer to use NVIDIA GH200 on its cloud. Previously, Google Cloud, Meta, and Microsoft are also expected to be early adopters, authorized to use DGX GH200 to explore its potential in processing generative AI workloads. Oracle Cloud Infrastructure (OCI) also announced plans to use GH200 for its cloud services. NVIDIA also plans to design DGX GH200 as a model to be shared with cloud service providers and other large-scale computing companies, enabling them to better adapt to their infrastructure. On the other hand, Microsoft and Meta also announced plans to integrate AMD's recently launched Instinct MI300X accelerator for AI workloads. Meta will use MI300X to build its new data center, and AMD directly compared it to H100. In addition, Intel will also announce the launch of its Gaudi3 AI accelerator at the Intel "AI Everywhere" conference on December 14. Intel showcased Gaudi2, which is very similar to NVIDIA's H100 and is cheaper.