06
Apr

How compelling is Nvidia’s rationale for buying Arm?

RCR Wireless News, April 6, 2021
Right after the last Nvidia quarterly earnings release, Jim Cramer, host of CNBC Mad Money spoke to Jensen Huang, CEO of Nvidia regarding the deal with Arm. Most of his questions were softballs, but what caught my attention was Jensen’s comment that Arm was not a must for Nvidia’s success, but a nice to have. That got me thinking and made me take a deeper dive into the rationale for the merger. Here are some of my thoughts on why Nvidia needs Arm more than vice versa.
Nvidia’s announcement of its intent to acquire Arm from Softbank has brought Arm out of the shadows and into the limelight. Arm has always been a silent performer, quietly powering the modern smartphone revolution. Its inner workings have been an enigma for many industry observers. And now, many are scrambling to understand what Arm does, and how Nvidia’s buyout will affect the semiconductor industry, the competitive landscape, and the future of tech at large. If you do not yet know the importance of Arm, consider this: almost any tech gadget you can think of, be it a simple IoT device, a game console, a smartphone, or even a modern car, has been touched by Arm technology in some shape or form. Its importance and reach are only going to expand, as the whole world moves toward untethered and low-power computing, as I explained in my earlier article here. Hence, the impact on the industry of its buyout by Nvidia is going to be oversized and impossible to overstate. 
Side note: You can read the full article series here.
Arm’s licensing model
To scrutinize Nvidia’s rationale effectively, one has to really understand Arm’s business model, especially its licensing model. In simple terms, Arm is the design house of power-efficient processors (aka cores) for the entire tech industry. It makes money by licensing those technologies in different forms. It offers three types of licenses—Processor, Optimized Processor, and Architecture. Let us look at each of these more closely.
The first, Processor License, is simply the permission to use processor cores designed by Arm. Licensees cannot change Arm’s designs but are free to implement them however they like in their own solutions. For example, Qualcomm, Samsung, and Huawei have this type of license. They combine multiple types of Arm cores (e.g., CPU, GPU, or other types, and in some cases, different sizes of cores) alongside other proprietary cores to make their semiconductor Systems on a Chip (SoC’s).  They also optimize the cores to achieve greater performance and to provide differentiation from other SoC’s. You might have heard about how Qualcomm SnapdragonSamsung Exynos, and (Huawei) HiSilicon Kirin platforms perform differently. That difference is because each company uses and optimizes Arm cores differently. So, such a license is for players that have the technical and financial wherewithal to do such optimizations.
The second, Optimized Processor License, is a bit more involved and detailed, where Arm not only provides the basic processor design but also, optimizations to achieve a certain level of guaranteed performance. This license is well-suited to companies that do not have the capabilities to implement and optimize a design, for example, smaller IoT chipset providers. This is probably Arm’s most popular option, with thousands of licensees.
The third, Architecture License, is also sometimes referred to as an Instruction Set Architecture (ISA) License or simply, Instruction Set License, and is the most minimalistic option. ISA licensees only get access to Arm’s instruction set and can design their own cores that run those instructions. Apple is such a licensee. Its A-series processors used in iPhones, iPads, and the new M1 processor used in Macs are designed by Apple but, run Arm’s instruction set.  Nvidia, Google, Microsoft, Qualcomm, and Tesla also possess architecture licenses.
Why is Nvidia buying Arm?
The reasons Nvidia has given for buying Arm can be grouped into three categories of benefits: 1) Using Arm’s vast ecosystem to distribute Nvidia’s Intellectual Property (IP); 2) Invest in Arm architecture to consolidate and expand its reach in the data center market; 3) Co-invent the Edge-cloud with Arm’s and Nvidia’s technologies.
In general terms, these reasons seem very attractive and complementary, benefiting both companies and their shareholders. They seem to benefit the industry at large as well, by giving others access to Nvidia’s market-leading graphics IP and accelerating the growth of data center and Edge-cloud markets. However, when you remove the covers and dig a bit deeper, there are quite a few peculiarities to consider.
First—Nvidia distributing its IP to the Arm ecosystem: from a business model perspective, Nvidia and Arm could not be more dissimilar. Arm is a pure-play licensing company that derives most, if not all, of its revenues from licensing. That means it is a neutral player across the whole ecosystem because it licenses its technology to all, and does not compete with any of its customers. On the other hand, to my knowledge the only thing Nvidia licenses is its CUDA software, and at no charge. One reason CUDA is free is because it only runs on Nvidia GPUs. Nvidia makes most of its money from its highly differentiated, high-margin GPU hardware and integrated software. Given this lucrative revenue stream, it is hard to fathom Nvidia’s willingness to license its GPU IP to Arm’s ecosystem, which would diminish its differentiation and destroy those sky-high margins. This could be particularly problematic, as some Arm licensees are in the process of developing products for the data center, where Nvidia makes most of its money. Nvidia’s licensing revenues and margins, like Arm’s, would be a pittance compared to Nvidia’s existing product revenues and margins. Unless there is another more plausible explanation where margins and revenue stream are not sacrificed, it is hard for anybody to buy this argument.
Second—Helping Arm expand into the data center market: this seems like a novel idea… the significant financial and other resource infusions Nvidia can make into the program could certainly accelerate Arm’s current trajectory. However, the “Arm for data center market” effort is already well underway, mainly because the data center service providers themselves have realized the importance of power-efficient processing, for financial as well as environmental (carbon footprint) reasons. Cloud giants such as AmazonGoogle, and Facebook have reportedly been working on their own in-house, Arm-based platforms. Arm already seems to have the financial and market support it needs. On the contrary, Nvidia with its high-performance, but energy-guzzling GPUs will need low-power CPUs to complement (and improve) its portfolio, especially since the data center market is becoming extremely energy conscious. Additionally, it is likely Arm, with its decades-long experience in low-power design, that can teach a trick or two to Nvidia to help reduce the power consumption of its GPU designs. So, although Nvidia’s resources might help Arm, it seems Nvidia needs Arm equally, if not more. 
Third—Co-inventing the Edge-cloud: unlike Arm in the data center market, this ship has sailed for some time. Power-efficient design is a basic necessity for edge compute, and one of the reasons that Arm is at the center of this universe. Thousands of small and large companies, including the cloud titans, are investing in, and developing technologies for the Edge-cloud. Nvidia will be a noteworthy addition to that ecosystem, but only one of many such players. Also, with power at a usability premium for Edge-cloud use cases and workloads, Nvidia has to pivot from its performance-only focused design philosophy to more power-efficient architectures. In this market, Arm will be of greater value to Nvidia than the other way around.
Upon closer examination of the three main reasons cited by Nvidia for the acquisition, one seems unconvincing, and the other two seem counter to Nvidia’s logic because it appears Nvidia would benefit more from Arm than vice versa. Moreover, Arm and its customers are already on the path with which Nvidia is proposing to help Arm. But, if the merger goes through, Arm, instead of being a neutral supplier with no conflicts of interest with its customers, would instead become a technology supplier as well as a competitor for its customers in the Cloud, the Edge-cloud, PC’s, the automotive industry, and AI. This dichotomy might affect Arm’s vast ecosystem and its unwavering support for the architecture. Also, Arm has developed its architecture and its business with significant inputs from its ecosystems. Ecosystem players would likely be disincentivized to share their inputs with a competitor, Nvidia-Arm. Nvidia’s resources, it seems, would not come without opportunity cost to Arm. 
I am sure you are aware of news reports citing many concerned ecosystem players reaching out to the FTC and other antitrust agencies about the acquisition. You might even be wondering what these players, including behemoths like Google, Samsung, Qualcomm, and even Apple, are worried about? Well, that is the topic of my next article… so be on the lookout! 
Meanwhile, for more articles like this, and for an up-to-date analysis of the latest mobile and tech industry news, sign-up for our monthly newsletter at TantraAnalyst.com/Newsletter, and listen to our Tantra’s Mantra podcast.