When Nvidia announced that it was in the process of buying Arm from Softbank, many analysts and industry observers were exuberant about how it would transform the semiconductor industry by combining the leading data center artificial intelligence (AI) CPU company with the leading device AI processor architecture company. While some see the potential advantages that Nvidia would gain by owning ARM, it is also important to look at the risks that the merger poses for the ecosphere at large and the course of innovation.
An understanding of the particular business model and its interplay highlights the importance of the proposed merger.
Nvidia became the industry leader in data center AI almost by accident. Nvidia became the largest graphics provider by combining strong hardware with frequently updated software drivers. Unlike its competitors, Nvidia’s drivers constantly improved not only the newest graphics cards but also past generation graphics cards with new drivers that made the graphics cards faster. This extended the useful life of graphics cards but, more importantly, it also created a superior value proposition and, therefore, customer loyalty.
The software also added flexibility as Nvidia realized that the same application that makes graphics processing on PCs efficient and powerful – parallel processing – is also suitable for other heavy computing workloads like bitcoin mining and AI tasks. This opened up a large new market as its competitors could not follow due to the lack of suitable software capabilities. This made Nvidia the market leader in both PC graphics cards and data center AI computation with the same underlying hardware and software. Nvidia further expanded its lead by adding a parallel computing platform and application programming interface (API) to its graphics cards that has laid the foundation for Nvidia’s strong performance and leading market share in AI.
ARM, on the other hand, does not sell hardware or software. Rather, it licenses its ARM intellectual property to chip manufacturers, who then build processors based on the designs. ARM is so successful that virtually all mobile devices use ARM-based CPUs. Apple, which has used ARM-based processors in the iPhone since inception is now also switching its computer processors from Intel to ARM-based internally built CPUs.
The ARM processor designs are now so capable and focused on low power usage that they have become a credible threat to Intel, AMD, and Via Technology’s x86-based CPUs. Apple’s move to eliminate x86 architecture from its SKUs is a watershed moment, in that solves a platform development issue by allowing developers to natively design data center apps on their Macs. Consequently, it is only a matter of time before ARM processor designs show up in data centers.
ARM vs. Nvidia
This inevitability highlights one of the major differences between ARM and Nvidia’s business model. ARM makes money by creating processor designs and selling them to as many companies that want to build processors as possible. Nvidia’s business model, on the other hand, is to create its own processor designs, turn them into hardware, and then sell an integrated solution to its customers. It is hard to overstate how diametrically different the business models are and hard to imagine how one could reconcile these two business models in the same company.
Currently, device AI and data center AI are innovating and competing around what kinds of tasks are computed and whether the work is done on the device or at the data center or both. This type of innovative competition is the prerequisite for positive long-term outcomes as the marketplace decides what is the best distribution of effort and which technology should win. With this competition in full swing, it is hard to see how a company CEO can reconcile this battle of the business models within a company. Even more so, the idea that one division of the New Nvidia, ARM, could sell to Nvidia’s competitors, for example, in the datacenter or automotive industry and make them more competitive, is just not credible, especially for such a vigorous competitor as Nvidia.
It would also not be palatable to shareholders for long. The concept of neutrality that is core to ARM’s business would go straight out of the window. Nvidia wouldn’t even have to be overt about it. The company could tip the scales of innovation toward the core data center AI business by simply underinvesting in the ARM business, or in industries it chooses to deprioritize in favor of the datacenter.
It would also be extremely difficult to prove what would be underinvesting when Nvidia simply maintained current R&D spend rather than increasing it, as another owner might do as they see the AI business as a significant growth opportunity rather than a threat as Nvidia might see it.
It is hard to overestimate the importance of ARM to mobile devices and increasingly to general purpose computing – with more than 130 billion processors made as of the end of 2019. If ARM is somehow impeded from freely innovating as it has, the pace of global innovation could very well slow down. The insidious thing about such an innovative slow down would be that it would be hard to quantify and impossible to rectify.
The proposed acquisition of ARM by Nvidia also comes at a time of heightened anti-trust activity. Attorney Generals of several states have accused Facebook of predatory conduct. New York Attorney General Letitia James said that Facebook used its market position “to crush smaller rivals and snuff out competition, all at the expense of everyday users.” The type of anti-competitive conduct that was cited as basis for the anti-trust lawsuit against Facebook was also that of predatory acquisitions to lessen the threat of competitive pressure by innovative companies that might become a threat to the core business of Facebook.
The parallels are eerie and plain to see. The acquisition of ARM by Nvidia is all too similar to Facebook’s acquisitions of Instagram and WhatsApp in that both allow the purchasing entity to hedge their growth strategy regardless of customer preferences while potentially stifling innovation. And while Facebook was in the driver’s seat, it could take advantage of customer preferences. Whereas in some countries and customer segments, the core Facebook brand is seen as uncool and old, Instagram is seen as novel and different than Facebook. From Facebook’s perspective, the strategy keeps the customer in-house.
The new focus by both states and the federal government, Republicans and Democrats alike, on potentially innovation-inhibiting acquisitions, highlighted by their lawsuits looking at past acquisitions as in Facebook’s and Google’s case, make it inevitable that new mergers will receive the same scrutiny. It is likely that regulators will come to the conclusion that the proposed acquisition of ARM by Nvidia looks and feels like an act that is meant to take control of the engine that fuels the most credible competitors to Nvidia’s core business just as it and its customers expand into the AI segment and are becoming likely threats to Nvidia.
In a different time, regardless of administration, this merger would have been waved through, but it would be surprising if that would be the case in 2021 or 2022.
Roger Entner is the founder and analyst at Recon Analytics. He received an honorary doctor of science degree from Heriot-Watt University. Recon Analytics specializes in fact-based research and the analysis of disparate data sources to provide unprecedented insights into the world of telecommunications. Follow Roger on Twitter @rogerentner.
“Industry Voices” are opinion columns written by outside contributors—often industry experts or analysts—who are invited to the conversation by FierceWireless staff. They do not represent the opinions of FierceWireless.