With Arm Purchase, Nvidia May Dominate AI Edge Computing

Moore’s Law — Intel co-founder Gordon Moore’s prediction that the number of transistors on a chip doubles about every two years — has been the foundation of the semiconductor industry. But, as the industry nears the limits of circuitry and physics of electronics, it’s being replaced by another one: that silicon chips powering AI more than double in power every two years, due to hardware and software. As Moore’s Law was the foundation for improvements in computers, this new law will power the Internet of Things. With its $40 billion acquisition of Arm Holdings, Nvidia could be positioned for a new type of evolution. 

The Wall Street Journal dubs it “Huang’s Law,” after Nvidia co-founder and chief executive Jensen Huang. Nvidia chief scientist and senior vice president Bill Dally reported that, “between November 2012 and this May, performance of Nvidia’s chips increased 317 times for an important class of AI calculations,” which equates to a doubling of performance every year — twice the speed of Moore’s Law.

Nvidia’s graphics processing units (GPUs) specialize in accomplishing “many independent tasks to be done simultaneously,” a more efficient operation than the CPUs that Intel specializes in, which excel at “executing a single, serial task very quickly.”

The AI these GPUs enable, among many products and tasks, include that used in cashierless checkout. Startup Standard Cognition inked a deal with Circle K to transform some of its stores into cashierless versions, similar to Amazon Go stores. Standard founder and chief executive Jordan Fisher noted that the company, “could do nothing and just wait and Nvidia will drop our prices every year.”

Also impacted are autonomous vehicles. Startup TuSimple, is “making a self-driving system that can fit the power and space limitations of a diesel-powered semi-trailer truck” and co-founder and CTO Xiaodi Hou said the company “is seeing performance double every year on its Nvidia-powered systems.”

Mobile phones are also seeing the impact, with Apple iPhone 8’s neural engine. At Nexar, which makes AI-powered dashboard cameras for cars, co-founder and CTO Bruno Fernandez-Ruiz noted that Apple’s decision to make that chip “accessible to any app on the phone — as well as the introduction of comparable chips and software on Android phones” has enabled his company’s cameras to “alert drivers to imminent hazards.”

Arm Holdings, whose technology is licensed by Apple among many other tech companies, “is at the center of this revolution.” That company’s vice president of marketing in the machine learning group Dennis Laudick said that, “over the last three to five years, machine-learning networks have been increasing by orders of magnitude in efficiency … [and that] now it’s more about making things work in a smaller and smaller environment.”

He noted that Arm’s smallest chips are “tiny enough to be powered by a watch battery … [and] can now enable cameras to recognize objects in real time.”

At Nexar, co-founder and chief executive Eran Shir pointed out that Nvidia’s interest in Arm is due to the “movement of AI processing from the cloud to the ‘edge’ — that is, on the devices themselves.” Nvidia, which has a “near monopoly on AI processing in the cloud,” uses Arm-based chips to “do much more of that processing in mobile devices, and faster, since it doesn’t have to be transmitted over the Internet first.”