Micro-network news, Intel revealed for the first time yesterday (August 8th) that the company sold artificial intelligence (AI) chips in 2017 amounting to 1 billion US dollars. At the same time, Intel expects that the department's turnover will increase by 10 times in 2022. To 10 billion US dollars.
Undoubtedly, the AI wave is sweeping the world and has become the market that domestic and foreign manufacturers are vying for. The US market research company ReportLinker has published a research report on artificial intelligence chips. In the next five years, global artificial intelligence (AI) chips will enter a stage of rapid development. By 2023, the market size will reach 10.8 billion US dollars, with a compound annual growth rate of 53.6%.
Currently in the AI chip competition, graphics chip maker Nvidia is in a leading position, and Intel has said that its AI chip's global market share is only 20%, but Intel is currently investing a lot of resources to catch up. Even Some analysts believe that Intel's public AI revenue can be seen, Intel is confident in capturing the AI chip market.
Intel invested heavily in the layout of AI
As the absolute overlord of the PC era, Intel has missed the mobile Internet era, and in the AI era that has arrived, the company has actively deployed and has invested heavily in the acquisition of Altera, Mobileye, Movidius, Nervana, Saffron and other companies.
It is understood that Intel's cloud AI chip project named 'Intel Nervana Neural Network Processors' (NNP), the main machine learning training. In the previous first AI developer conference AIDC 2018, Intel released The latest cloud AI chip, the Nervana NNP-L1000, code-named 'Spring Crest', has improved 3-4 times performance over the previous generation 'Lake Crest' and is expected to be available to users in the second half of 2019. About Spring Crest Process node, Intel has not announced any information.
It is worth mentioning that the previous generation 'Lake Crest' used Intel's own 14nm process. In addition, Carey Kloss, vice president of hardware for Intel AI product group, said that the new Spring Crest chip is based on Google's recently released TPU 3.0. The previous generation of Lake Crest chip benchmark is Google TPU 2.0.
Last year, AI revenue may be more than $1 billion.
With PC sales stagnating, Intel is increasingly relying on data center services—selling chips for data centers that support mobile and network applications. These applications rely on AI chips for photo and speech recognition.
Recently, Intel held a meeting in Santa Clara, Calif., and Intel Data Center Director Navin Shenoy said at the meeting that in the past few years, the company has been able to improve its central processing unit and increase their processing power in AI training by more than 200 times. The Intel Xeon processor achieved sales of $1 billion in 2017, while the company’s total revenue was $62.8 billion.
At the same time, Navin Shenoy said that the step-by-step improvement in processor performance has brought considerable business impact to Intel.
Naveen Rao, head of Intel's artificial intelligence products division, said Intel estimates that it sold $1 billion worth of AI processor chips last year, from the customer telling Intel that they are looking to buy AI chips, and how many such data centers are needed from computing customers. Calculation of work.
In an interview, Naveen Rao even said that it may be more than $1 billion, which may be much higher.
Why is the $1 billion figure so important? Because Intel’s earnings last month were lower than Wall Street’s expectations for its data center business, which also caused Intel’s share price to fall. Intel also said that its latest generation of chips will be available Postponed until 2020, this makes analysts worry that Intel's data center market share will be replaced by AMD.