For many years, it has been quite common for desktops and laptops to feature Intel chips. Less well-known is that the company is also busy creating a place for itself in the AI processor chips market.
Yesterday, Intel announced that it sold $1bn of AI processor chips last year. This is the first time the second-biggest chip manufacturer on earth revealed revenue from a computing division that has driven sales growth at competitors, such as Nvidia Corp.
With PC sales stagnating, Intel is increasingly relying on purchases by data centres that deliver the largely-unseen computing power for web-based and mobile applications. They often depend on AI for features such as speech and image recognition. Many researchers, though, are convinced that Nvidia’s ‘graphical processors’ are better for training AI computer models than the CPU units that Intel has produced for many years.
Intel data centre chief, Navin Shenoy, however, recently announced that over the last couple of years the firm was able to adapt its CPUs to make them over two hundred times better at AI training. The end result was $1bn in Xeon processor sales and total revenue of $62.8bn last year.
According to the man in charge of Intel’s AI product group, Naveen Rao, the $1bn figure was ‘probably a lot higher’. This is important information, particularly since Intel’s share price dropped last month after earnings for its data centre business failed to meet analysts’ expectations and the announcement that Intel’s latest generation chips would not be ready before 2020.
Ex-chip executive turned analyst, Patrick Moorhead agreed that the $1bn figure is conservative. Intel’s share price subsequently increased by 22c to $49.92.