Nvidia  Now we are renting an artificial intelligence supercomputer independently developed from those who are eager to visit their hardware and software packages and provide GPUs provided in the cloud.
    DGX Cloud service will include its high -performance AI hardware, including the current supplyH100And A100 GPU.Users will be able to rent the system through Nvidia's own cloud infrastructure or oracle service.
Tony Paikeday, director of Nvidia Dgx platform, said: DGX Cloud can be available globally, except for regions controlled by the United States.
    This cloud will be provided in the cloud infrastructure of NVIDIA, including DGX systems in the United States and Britain.DGX Cloud will also be provided through the Oracle cloud infrastructure.
    nvidia announced the DGX cloud service for the first time at the GTC conference in March, and then announced the extensive launch of the service.The announcement issued on Tuesday was issued after a series of cloud artificial intelligence announcements released last week.
    competitors CEREBRAS SYSTEMS is installing an artificial intelligence system in cloud services running in the Middle East cloud provider G42, which will provide 36 Exaflops.Tesla announced the beginning of the production of Dojo supercomputers, which will run its independently developed D1 & Nbsp;chipBy the end of next year, 100 Exaflops can be provided.The benchmark varies from data type.
    Tesla Executive Elon Musk talked about the Nvidia GPU shortage of existing artificial intelligence hardware last week, Tesla is waiting for supply.Users can lock in NVIDIA hardware and software on DGX Cloud, but pay a high price.
    DGX Cloud lease includes cloud computers accessing NVIDIA. Each cloud computer is equipped with H100 or A100 GPU and 640GB GPU & Nbsp;MemoryThe company can run AI applications on it.Nvidia's goal is to run its artificial intelligence infrastructure -input data as raw materials like a factory, and output is an available information that the company can put in.Customers do not have to worry about the middle software and hardware.
    Paikeday also mentioned that DGX Cloud meets a key requirement: a special calculation for multi -node training for large -scale complicated AI models (such as large language models).Paikeday continues to say that companies will also gain deep technical knowledge to deploy and operate the environment to support such workloads.What's more?
    DGX cloud starting price is $ 36,999 per example per month.
    This is about twice the price of Microsoft Azure ND96ASR instance. This instance is equipped with 8 NVIDIA A100 GPUs, 96  CPU  kernel and 900GB RAM, the monthly cost is $ 19,854.The basic price of NVIDIA includes AI Enterprise software, which provides access to large language models and tools to develop AI applications.
    leased products include a software interface called Base Command Platform, so that the company can manage and monitor DGX Cloud training workloads.Oracle Cloud has a cluster of up to 512 NVIDIA GPUs, a RDMA network of 200 GB per second, and supporting multiple file systems (such as Lustre).
    All major cloud providers have their own NVIDIA H100 and A100 GPU deployment, which is different from DGX Cloud.
    Google released an A3 super computer equipped with 26,000 NVIDIA H100 Hopper GPU earlier this year. Its settings are similar to NVIDIA DGX Superpod with 127 DGX nodes. Each node is equipped with 8 H100 GPUs.Amazon AWS EC2 Ultraclusters with a P5 instance will be based on H100.
    We hope that DGX Cloud can attract new generation AI customers and workloads into the cloud of our partners, Paikeday said.
    locking also follows -Nvidia is trying to allow customers to use proprietary AI hardware and software technology based on CUDA programming models.In the long run, this may bring high costs to the company because they need to pay the software license and the cost of GPU time.Nvidia said that investment in artificial intelligence will benefit the company in the form of long -term operation.
    The artificial intelligence community is promoting open source models and opposing proprietary models and tools, but Nvidia occupies a dominant position in the artificial intelligence hardware market.NVIDIA is one of the few that can provide hardware and software stacks and services, making the actual implementation of machine learning as one of the possible companies.
    people are interested in Nvidia's artificial intelligence hardware because people are eager to use the prospects of generating artificial intelligence.OpenAI's ChatGPT shows the function of artificial intelligence in the form of a chatbot, but it is now a new model of vertical markets including medical care, insurance and finance.