Nvidia-Arm deal ‘will be good for customers,’ while Arm GPU and NPU development will continue…
Nvidia CEO Jensen Huang is confident that the Nvidia-Arm deal, the biggest in semiconductor history at $40 billion, will get past global regulators. Speaking during a “fireside chat” as part of Arm Dev Summit, the company’s developer conference, Huang said that the two companies are complementary and as such, the acquisition would drive innovation forward and be good for customers.
“I am confident. We are confident that [the Nvidia-Arm deal] is going to go through,” Huang said. “As soon as we explain the rationale of the transaction and our plans to regulators around the world, they will realize that these are two complementary companies.”
Arm CEO Simon Segars backed Huang up, while noting that the regulatory process will take some time.
“This is a deal that’s about expanding,” Segars said. “It’s about creating more and new technology. It’s about putting that technology in the hands of people who are going to build really cool stuff with it. So from that point of view, this is about enabling more people to do more things. That’s the positive thing, that as regulators do scrutinize this, they’re going to be able to see.”
Huang’s reasoning behind the Nvidia-Arm deal is that the combination of the world leader in AI compute with the world’s most popular CPU architecture will bring AI capabilities to the Arm ecosystem as well as give Nvidia’s accelerated computing extended reach.
“I’ve always wished that we could take the Nvidia architecture and put it through Arm’s network, so that companies of all different sizes and shapes, and different industries could take advantage of our IP and allow us to extend the accelerated computing vision we have to all industries,” he said.
Huang is sure that customers will get behind the deal once they realize “that we love [Arm’s] business model, we’re going to protect the business model. We’re going to continue to nurture this neutral, committed, consistent trusted platform, so that our ecosystem can continue to flourish and even grow. I think that when people realize that, they’re going to be really quite excited.”
The Nvidia CEO also confirmed that the Nvidia-Arm deal would not affect Arm’s plans for its popular GPU line (Mali), nor its neural processing unit (NPU) line (Ethos), both of which ostensibly clash with Nvidia’s offering.
“We’re going to keep doing it,” Huang said, without hesitation. “We want to give customers as many choices as possible. Every customer is not trying to build the same type of system and some of the customers probably want to design their own. And so we want to give the customers in the market as many opportunities to innovate and differentiate as possible. So we’re going to continue to advance [Arm’s GPU and NPU product lines].”
Nvidia announced last year that it would begin to support Arm CPUs for exascale applications, bringing its full stack of AI and HPC software to the Arm ecosystem. Huang made it clear that this decision was not taken lightly.
“Once you decide to make a commitment to an ecosystem, you can’t take it back,” he said. “It took us several years to come to the conclusion that Arm is finally at a point where it’s going to evolve way past mobile devices. It’s going to go into high performance computing, into the cloud, into the edge, into autonomous machines, and that we ought to make the first commitment to bring our architecture to the platform. The response has been fantastic.”
Since this was announced last summer, Huang said Nvidia had been working closely with Arm in four key areas. The first is bringing all Nvidia’s GPUs and DPUs, plus all the software and acceleration stacks to Arm, to make a combined (CPU, GPU, DPU) general computing architecture for data center/HPC. Secondly, Nvidia will bring all of its domain specific acceleration libraries to Arm’s architecture – this includes HPC, AI and Rapids (Nvidia’s data science platform). Third, doing so for all Nvidia’s computing platforms — this means HPC, cloud computing, edge computing and personal computing. Finally, Nvidia has started to build an ecosystem, selecting three Arm-based HPC processor partners as a starting point (Fujitsu, Ampere and Marvell).
Huang made it clear that this marriage is a long-term commitment.
“Take the permutation of all of that: you have processors, you have systems, you have libraries and you have an ecosystem, and we’ll support it for… as long as we all shall live,” he said.