Intel has sold 220 million Xenon processors since the company was launched 20 years ago, generating $130 billion in revenue. Now Intel has received $1 billion revenue just through the sale of Artificial Intelligence(AI) chips in 2017. In Intel’s market assessment, Intel revealed to have only 20% of market share in key categories.
Five years ago Intel’s revenue was from data-centric applications, which is now almost close to half of its regular business, so the executive president of Intel, Navin Shenoy told at an Intel event that Intel’s AI chip business is strategically important, as it is a shift-data-centric company. Shenoy also expressed his concern about the company giving a lot of its attention to only AI chips now, as the Cascade Xenon processor coming later this year will be 11 times better than AI image recognition than the previous silver lake Xenon processor.
Shenoy also told that for AI chips, in particular, the market opportunity will grow at 30% every year, from $1 billion in 2017 to $10 billion by 2020.
Researchers believe that NVIDIA’s so-called graphic processor is better suited to train artificial intelligence computer models than for the central processor units. Naveen Rao, head of Intel’s artificial intelligence product group, said that the $1 billion estimate was derived from customers that told they were buying chips of AI from calculations of how much customer’s data centre is dedicated to such work. Rao also told the revenue is actually a lot more than the estimated value.
Intel still has a huge task ahead as they move towards the world of self driving cars, and creating maps for those cars requires a camera that can capture and deliver data in a given real time. Self driving cars are expected to generate traffic of 4 terabytes of data per hour, which has to be sent to the cloud, analysed in the data centre, and then sent back to the respective cars. Intel estimated that it will require to use data on an average from 2 million cars on the road. That means AI will have to infuse everything from vehicle chips to edge devices to cloud computing.
Moorhead, the analyst, said that Intel does not have a lot of growth potential though they have a 98% market share in the data centre. In order to solve this problem, Intel hired Jim Keller, a chip expert who architected chips to everyone from Tesla to Apple. He sad that he came on board of Intel to change the world faster, he also said that the next 25 years will be bigger than the last 25 years, the AI revolution is really getting big as it impacts the client, data centre, and graphics.
Analyst Moorhead also told that it is hard to predict how well Intel competes with NVIDIA on AI training, as there isn’t any information provided on that.
Lastly, Intel’s Ice lake is a future Intel Xenon scalable processor based on 10 nanometer manufacturing technology, and is planned to be released in 2020. As Ice lake is a couple of years late, Moore said that it will be a tough choice to whether to wait for Ice lake or adopt cooper lake earlier.
Rao also told that the diversity of processing needs to vary from client to data centre, and Intel recognises that 1 size cannot fit for all, so they provide a different solution of using works from acquizitions like Movidius, Mobileye, and Nervana. Nervana will have 3 to 4 times the training perfoormance compared to the first generation chip.