AI isn't just changing internet services, cars, robotics, and healthcare. It's changing the computer chip market too.
This shift was underlined on Monday when Intel said it would pay $15.3 billion to acquire Mobileye, an Israeli company that makes chips and cameras for cars and trucks, including the self-driving variety. The purchase will be Intel's second largest ever, following its $16.7 billion acquisition of chip-maker Altera in 2015. The Altera buy was also driven, in part, by the recent rise of machine learning, where machine can learn discrete tasks on their own.
These are enormous acquisitions in many respects. After acquiring Mobileye, Intel will move its autonomous driving team to the Mobileye's headquarters, not vice versa. In other words, Intel is letting Mobileye take over a growing wing of its company. The world's largest chip maker knows the market is changing rapidly, and it's determined to change with it.
As long as PCs and servers dominated the computing landscape, Intel reigned supreme. But PCs now share the stage with phones, tablets, drones, TVs, smart thermostats, and, yes, cars. For the most part, the smartphone and tablet revolution sidestepped Intel in favor of chips from ARM, and though these gadgets still connect to data centers in the cloud, where Intel chips are still dominant, this hegemony is now being threatened by the rise of chips suited to running neural networks, complex mathematical systems that can learn discrete tasks by analyzing vast amounts of data, including recognizing images and translating from one language to another.
Typically, neural networks analyze data with help from graphics processing units, or GPUs, chip originally built to help power video games in the late 1990s, and most of these are supplied by Intel rival nVidia. But other companies are exploring more specialized chips for machine learning. Last year, Google unveiled its own AI-focused chip called the Tensor Processing Unit, or TPUs, and Microsoft revealed that it uses Altera's customizable chips to power its own AI applications. That's why Intel acquired Altera.
But increasingly, specialized AI chips are also needed in devices, including cars. Think of it this way: companies need chips in the data center to train the neural networks, and as these machine learning models get more complex, they will also need chips in the devices to help execute the models. Once a company teaches a system to recognize detailed images, for instance, it will need a way of running a system on a phone or a car.
With this in mind, Intel has bought-up mobile-focused AI chip company Movidius and now Mobileye. Plus, it has already makes the RealSense 3D camera for drones, cars, and any other gadget that needs artificial eyes, and it has acquired drone-makers Ascending Technologies. You saw some of the fruits of these efforts during the Super Bowl half time show this year.
Intel is determined not to miss the next chip wave the way it missed the switch to mobile. But there are obstacles aplenty. Microsoft recently revealed that it's exploring the use of ARM chips in its data centers, another push against Intel's hegemony in the data center. However the battle of the chips plays out, one thing is certain: the market isn't what it used to be.
