The American giant of the semiconductor industry Intel Corp introduced its new processor on the Chinese market for deep training in artificial intelligence (AI). This processor does not fall under the operation of American export restrictions, in contrast to the products of the competitor of Intel – NVIDIA Corp, whose graphic processors (GPU) became the smuggling object in China due to huge demand for them.
At a press conference in Beijing on Tuesday, Intel representatives demonstrated their Gaudi2 processor, which, according to them, is an alternative to the premium GPU A100 from NVIDIA, widely used for teaching AI systems.
The last initiative of Intel – a company that received 27% of its total income in 2022 in China, according to its last annual report, emphasizes the continued importance of a huge Chinese market for American suppliers of semiconductor technologies, despite Washington’s export restrictions.
She should promote modified versions of its flagship GPU A100 and H100 in China in China in order to comply with American restrictions and maintain supplies for its customers in mainland China, where new projects for the development of AI, creating services similarly Openai’s Chatgpt.
In August last year, the US Department of Trade banned NVIDIA and Advanced Micro Devices (AMD) to sell some advanced chips to China. NVIDIA received an indication of stopping the sale of its GPU A100 and H100, and AMD – export your MI250 chips to the mainland.
This forced the AI Chinese sector to fight for the acquisition of limited reserves of advanced chips from NVIDIA, which has almost a monopoly on GPU used for training AI systems.
Strong demand even created a fast -growing market for smuggled GPU, such as A100 and H100 devices from NVIDIA.
GPU A100 Tensor Core, according to the NVIDIA website, ensures the operation of the data in the world of data centers used for AI applications, data analysis and high-performance calculations.
For Intel, demand in China strengthens its obligation to provide customers with a wide selection of hardware solutions, said Sandra Rivera Executive Vice-President of the Intel General for working with Data Centers and AI at the Beijing press conference in Beijing .
She said that Gaudi2 was developed to reduce the threshold of entering and increasing the ability of its Chinese clients to “introduce AI using cloud and intellectual technologies on the edge of the network, helping to create the future of AI in China”.
Intel announced that he cooperates with the Inspur Group – the second largest in the world by the manufacturer of AI servers based in the province of Shandun in east of China – to build new machines based on Gaudi2 for the mainland market.
Habana Labs, the Intel team in working with Data Centers, specializing in processors for deep training technologies AI, initially launched Gaudi2 in May last year in the United States, where she stated that the learning processor was twice as higher than than The GPU A100 from NVIDIA 80 gigabytes for the computer model of the Resnet-50 vision and the model of the natural language processing Bert.