After months of speculation, GPU king NVIDIA announced Sept. 13 that it is acquiring chip developer/designer Arm from SoftBank for $42 billion, comprising $12 billion in cash, $21.5 billion in stock, a $2 billion payment at signing, $1.5 billion in NVIDIA stock for Arm employees and an addition $5 billion payment based in Arm’s performance.
Arm will continue to operate in its Cambridge, UK headquarters and will function as a division of the broader NVIDIA corporation.
The enormity of this deal highlights just how massive Santa Clara, Calif.-based NVIDIA has become in a relatively short period of time. When SoftBank purchased Arm in 2016, it paid about $32 billion, and NVIDIA’s market cap was only about $30 billion. That was a mere four years ago, and NVIDIA is now worth $300 billion, or 10X its valuation back then. The company’s growth has been fueled by the demand for its graphics processing units, the main computing unit used to power accelerated computing systems, such as artificial intelligence, ray tracing, self-driving cars, super computers and a bunch of other leading-edge use cases.
NVIDIA-Arm will be industry changing
This deal will have a profound impact on the broader computing industry because it helps pave the way to expose Arm’s massive customer base to the power of GPU computing and can help NVIDIA build better integrated “end-to-end” systems. This becomes increasingly important as the world relies on accelerated computing to solve some of the planet’s biggest problems, such as finding a cure for COVID-19, building autonomous vehicles and doing seismic exploration.
Arm builds CPUs, similar to Intel and AMD, with one major difference. Intel and AMD design the chips, manufacturer them and ship them to systems companies that install them onto a motherboard; Arm designs silicon and then turns the architecture over to other companies to build the chips. This enables the manufacturer to optimize the system for that processor, making it more power- and space-efficient, compared to the pluggable model of an Intel or AMD processor.
An easy way to think of the difference is that Intel and AMD based systems are optimized in software, while Arm systems can be optimized in hardware and software, creating more efficient systems.
Arm has been used in mobile but is rapidly expanding
This is why Arm has been the preferred CPU for mobile devices for years and can be found in iPhones, Samsung and Qualcomm devices. But the efficiency and improved performance of Arm is just catching on. Microsoft now makes an Arm-based Surface laptop and has released Windows for Arm. Also, Apple recently announced it was planning to switch its future Mac computer systems to Arm processors.
It’s important to understand that, as powerful as GPUs are, they aren’t great at everything. While they handle high-performance computing tasks, like video analytics and AI well, CPUs are still used to boot systems and run a variety of other processes so even the most advanced systems use a combination of CPUs and GPUs and now NVIDIA has both. This lets the company create better end-to-end designs when Arm processors are used. This is similar to the approach it is taking with data center networks with the acquisition of Mellanox.
NVIDIA commits to leaving Arm open
One of the important aspects of this announcement is that on the acquisition call, NVIDIA CEO Jensen Huang made it crystal clear that Arm will continue to operate with the current open licensing model while maintaining its customer neutral go to market approach.
The open approach has made Arm the company it is today, and NVIDIA won’t disrupt that. However, that doesn’t mean it’s the only approach the company must take. This acquisition opens the door for NVIDIA to take Arm’s design and build the CPUs, similar to the way it does GPUs. Or it could license its GPU design using Arm’s model. Is either better? Not really; it depends on what the customer wants, and I believe there is a market for both and that opens more doors for both companies, particularly NVIDIA, when we are just at the start of the GPU-in-everything cycle.
On the analyst and press call Sept. 13, Huang did a good job of outlining the size and scope of Arm compared to NVIDIA. He stated that last year alone, Arm shipped 22 billion chips. In the same time frame, NVIDIA shipped about 100 million. The latter is a nice number, but it’s orders of magnitude smaller than Arm. The reason is that NVIDIA has historically served very select markets, such as supercomputers, self-driving cars and gaming systems. Now it has exposure to the entire Arm pie.
Similarly, NVIDIA can actively look to use the performance and power efficiency in many of its GPU systems. One low-hanging fruit use case is edge AI systems, where space is limited as is power, because many systems operate on batteries, but these systems need to perform complex tasks.
Another blow to Intel
This acquisition also will be another nail in the coffin of rival Intel. For years, NVIDIA was considered a niche gaming company (which it was), while Intel was the king of silicon (which was also true), but over time and through a number of good decisions by NVIDIA and Intel’s inability to build a GPU, NVIDIA continued to grow while Intel flat-lined. In July 2020, NVIDIA caught Intel with respect to market cap, and both were worth about $250 billion. Today, Intel has slipped to $209 billion, and NVIDIA is at about $300 billion. The acquisition of Arm will help NVIDIA accelerate the replacement cycle of Intel to Arm by building better-optimized systems in which both CPUs and GPUs are needed.
I believe AI is the single biggest disruptive technology to happen since the birth of computing. Speech analytics, video recognition, translation, automation and more will become standard across almost all devices--both consumer and business--and increase the scope of where GPUs are needed.
With the acquisition of Arm, NVIDIA is now able to offer its customers greater flexibility in how things are designed with improved performance. This is a well-timed acquisition by the company, because we are just hitting that inflection point.
Zeus Kerravala is an eWEEK regular contributor and the founder and principal analyst with ZK Research. He spent 10 years at Yankee Group and prior to that held a number of corporate IT positions.