Rumors of a union between the world’s leading GPU provider and CPU IP vendor have been swirling for a couple of months now, but on Sunday night it became official: NVIDIA is purchasing Arm from Softbank for a reported $40 billion.
So now there are more questions than answers. Yes, the move does fill a big hole in NVIDIA’s computing portfolio by adding CPU technology, which will help them compete with Intel (who acquired FPGA company Altera in 2015 and is said to be working on a GPU architecture), AMD and others for heterogeneous computing sockets from the data center to the edge.
However, what does the acquisition mean for Arm licensees? NVIDIA doesn’t currently – or didn’t, anyway – license IP, or any hardware for that matter. While the company did begin licensing virtual GPUs earlier this year, they have used traditional channel partners to get their physical chips to market.
Mixing the two models could get messy. And while it’s not impossible to have a traditional semiconductor business alongside a licensing business – we’ve seen Qualcomm navigate this dance floor for decades – it’s certainly not ideal.
Nevertheless, NVIDIA CEO Jensen Huang says that they “love Arm’s business model.”
“We will maintain Arm’s open licensing model and customer neutrality, serving customers in any industry across the world,” Huang said. “In fact, we intend to expand Arm’s licensing portfolio with access to NVIDIA’s technology. Both our ecosystems will be enriched by this combination.”
Wow. So Now What?
Well, on the surface this appears to mean that developers of the 180 billion Arm units that have shipped will now have access to a larger technology ecosystem based on NVIDIA GPU technology, CUDA, and so on. The converse is also obviously true. According to NVIDIA, the acquisition will result in a combined software development universe of approximately 15 million engineers, which sounds like a big win for open-source software innovation
And while the headliner application of Sunday’s press conference was AI, the combined technology stack is a serious competitor for workloads in the IoT, cloud/data center, 5G networks, autonomous vehicles, robotics.
Huang also reinforced that Arm headquarters will remain in Cambridge, and that the company name and branding will remain intact. Arm IP will continue to be registered in the United Kingdom, and there are also plans to build an AI center of excellence there.
Around the World
Despite NVIDIA’s stated plans to continue supporting the U.K. semiconductor market, the deal is still subject to approval by authorities in the U.S., Britain, China, and the European Union. This is when the rubber will meet the road, and more definitive requirements around jobs, trade, and a range of other economic details will be ironed out.
Back to the tech. We already knew that NVIDIA was making use of Arm CPUs in their Tegra and Xavier offerings (which feature Cortex-A9s, A-57s, etc.), and we can presumably expect a tighter integration of the two compute technologies in future SoCs. But just how tight? And when can we expect it?
Only time will tell.
About the AuthorFollow on Twitter Follow on Linkedin Visit Website More Content by Brandon Lewis