Petya Petrova - Fotolia

Nvidia reveals its plans for Arm

Nvidia intends to acquire Arm for £31bn, as part of its plan to “create the computing company for the age of AI”

Arm used its Dev Summit virtual event to set out how its recent sale to Nvidia would provide a growth opportunity for the hardware and software developers who rely on its technology.

In his keynote speech opening the webinar, Arm CEO Simon Segars said: “We will be able to accelerate innovation – to unleash the computing technology potential, to help organisations build and release ideas.”

With the fifth wave of computing emerging, Segars believes that artificial intelligence (AI), the internet of things (IoT) and 5G are set to change the world, and Nvidia sees the acquisition of Arm as a vital step in putting the company that became a household name in PC gaming at the forefront of next-generation computing, beyond Moore’s Law.

Nvidia CEO Jensen Huang said: “We would like to create the computing company for the age of AI.” In a fireside chat with Segars, Huang described how AI is now able to write software that no human programmer would be able to create. “We want to unite the world leader in AI and the world’s most popular computing platform company, to create technology to help everyone grow,” he said.

During the discussion, Huang also spoke about “nurturing the Arm business model”, which is based on cultivating a vast ecosystem of hardware partners to build systems based on chip designs licensed from Arm. He described software development as the “engine of growth”.

With computers writing more and more software for other computers to run, Huang predicted that the expansion of computing over the next decade will be much faster than it was during the previous decade. “No computing platform has the reach of Arm,” he said.

It is this ecosystem that Huang hopes Nvidia can tap into. Last year, the company made its Cuda libraries for developing parallel computing software on Nvidia graphics processing unit (GPU) available on the Arm processor. Cuda provides domain-specific libraries to speed up the development of applications in areas such as quantum chemistry, fluid dynamics and scientific computing.

Read more about Nvidia/Arm

Huang described this as a big commitment. “Once you support software, you can’t stop,” he said. “This is why computing platform companies make a commitment to their ecosystem. You can’t hold back.”

During the talk, he announced that Nvidia would bring its GPU and data processing unit to Arm, which he claimed would complete the Arm hardware platform and make it a general computing platform.

Cuda is one of four parts that make up Nvidia’s overall strategy. The second is technology to support AI for training and inference; third is data acceleration across high performance, cloud and edge computing, as well as personal computing devices that require fast graphics and accelerated image processing; and fourth is the ecosystem.

In March 2018, Nvidia commissioned Forrester to analyse the cost of its first pre-built AI machine, the DGX-1. When the analyst firm assessed the software components, it noted: “The DGX-1 software has been built to run deep learning at scale. A key goal is to enable practitioners to deploy deep learning frameworks and applications on DGX-1 with minimal setup effort.”

The platform supports Cuda accelerator libraries as well as third-party one. Forrester’s Total Economic Impact website highlighted the savings a business would typically make by not requiring full-time data scientists to manage the hardware side of the system.

The DGX-1 platform is an example of where Huang is taking Nvidia. Arm is the next step on that journey. He believes that Moore’s Law has come to an end and the only way to gain performance improvements is by building better computing architectures.

Its intention to acquire Arm is focused on expanding the reach of its software platform across the Arm ecosystem, and to offer organisations a way to run high-performance workloads on pre-built hardware using Nvidia and third-party software libraries.

Next Steps

Apple takes its M1 chip to the iMac, iPad Pro

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management