Assessing NVIDIA’s Opportunities, Challenges, and Options for Acquiring Arm

Earlier this week, SoftBank confirmed it had reached an agreement with NVIDIA on Arm, under which NVIDIA will acquire the British semiconductor and processor design company in a deal valued at US$40 billion. SoftBank acquired Arm in 2016 for an outstanding value of US$32 billion. In the meantime, Arm has made several acquisitions in excess of US$1 billion, including Apical, an imaging company specializing in computer vision, Allinea, a software tool provider targeting High-Performance Computing (HPC) devices, and Treasure Data, a device management software company targeting the enterprise sector.

However, to complete this acquisition, NVIDIA still needs approval from several regulators around the world, as well as from Arm’s Intellectual Property (IP) licensees, which include major chipset makers competing with NVIDIA across multiple markets. So, the deal should not be considered done until all parties concerned give their blessing.

When SoftBank acquired Arm back in 2016, ABI Research was not overwhelmingly enthusiastic about the outlook of this acquisition for numerous reasons, chief among them the lack of synergy needed by Arm to transform itself and grow beyond just licensing its IP. For more details on this, refer to this article in The Wall Street Journal.

Under SoftBank’s ownership, Arm managed to grow its revenue only slightly from US$1.5 billion in 2016 to US$1.8 billion in 2019. This growth was essentially streamed from licensing its IP to new sectors, including HPC and the Internet of things (IoT). While these two markets have witnessed double-digit year-on-year growth within the last 4 years, the revenue share of Arm from the total processor market remained very slim and far from enough to justify a rapid Return on Investment (ROI) for SoftBank. Based on Arm’s current performance, ABI Research estimates that it could take up to 20 years for SoftBank to see any ROI from Arm.

In fact, Arm is in a crucial need of revamping its business model and technology offering beyond just licensing its IP if it wants to become a key solution provider for the markets it services. SoftBank did not possess the technology and innovation muscles to lead Arm in this direction, but NVIDIA may have what it takes to transform Arm’s business.

NVIDIA is one of the most aggressive semiconductor companies out there that managed to transform itself from selling graphic accelerators, essentially targeting the gaming market, into a clear leader in the Artificial Intelligence (AI) industry and an authoritative leader in many new and lucrative areas, including autonomous driving, robotics, Augmented Reality (AR)/Virtual Reality (VR), HPC for data centers, and open mobile networks.

It is this very visionary mindset that Arm needs to its value proposition from a traditional IP supplier to a key market leader capable of translating technology innovation into true business opportunities.

Arm’s business can be categorized in two major areas, namely Arm’s architectures business and Arm’s cores business. Arm’s architectures are basically Reduced Instruction Set Computer (RISC) processor designs configured to address various use cases ranging from mobile phones and IoT devices to HPC applications. Arm’s cores business focuses on creating end-to-end processor blueprints, including instruction sets, interfaces to memory, peripherals, radios, and other accelerators. Examples of Arm’s core designs include Cortex-A series, Cortex-M series, Cortex-R series, CoreLink, Artisan, and Mali. While virtually all mobile and IoT chipset suppliers use Arm architectures as a foundation for their products, not all of these companies license Arm cores. Some companies prefer to use their own core designs to differentiate, including Apple, Qualcomm, and Huawei, mainly for their high-end product portfolios. Others like MediaTek and UNISOC rely 100% on Arm’s cores for building their chipsets.

While Arm’s architectures have become the foundation for the entire RISC ecosystem, including mobile and IoT devices, Arm’s cores business is the asset on which NVIDIA really needs to capitalize. This business is very complementary to NVIDIA’s product portfolio and could enable the company to make a huge difference in markets in which it has yet to establish itself, including the mobile device and IoT markets.

Since its inception, NVIDIA has lacked a solid Central Processing Unit (CPU) business and this is becoming a serious gap in its technology portfolio. Arm’s Cortex-A series could be a good starting point for NVIDIA to develop its CPU technology. A solid CPU technology is becoming essential for NVIDIA if it wants to maintain its leadership in markets it is established in today. For example, the AI processor market, where Nvidia is currently the number one chipset supplier, is moving very fast toward heterogeneous architectures where both training and inference workloads are distributed across multiple execution environments, including Graphics Processing Units (GPUs), Field-Programmable Gate Arrays (FPGAs), CPUs and Application-Specific Integrated Circuits (ASICs). Each of these architectures is used to its strength on different project types and targeted use cases. The CPU technology is also becoming essential for NVIDIA to be able to grasp new opportunities in markets where Arm is increasingly challenging Intel’s x86 architecture, notably the Personal Computer (PC) and the server markets.

This potential acquisition is not a done deal yet, but is still pending approval by Arm licensees, as well as by key regulatory authorities across the world, notably in the United States, the United Kingdom, Europe, China, and Japan.

Several Arm licensees have already expressed concerns regarding this potential acquisition, but NVIDIA promises that Arm licensing will remain business as usual. However, NVIDIA still needs to make actionable guarantees on this if it wants to win approval from most of Arm licensees. NVIDIA has two potential options to address this issue. The first option is for NVIDIA to keep the Arm architectures division independent from the rest of Arm’s business. While this option could enable NVIDIA to meet the requirements of Arm licensees and the regulatory issues, this is a highly unlikely option because it could break the synergy Arm has created between Arm architectures and Arm cores. The second option, which ABI Research qualifies as most likely, is for NVIDIA to open-source Arm architectures or even harmonize this part of Arm’s business with RISC-V. This option could be an interesting proposition for Arm licensees. At the same time, it would enable NVIDIA to have clear visibility and control of Arm’s roadmap. If this option materializes, it could be a major development within the semiconductor industry, enabling the RISC ecosystem to expand beyond the mobile device and IoT markets. It will allow the RISC ecosystem to challenge Intel’s x86 in many areas, including the HPC and PC markets.

In terms of approval from key regulators around the world, ABI Research does not expect any major objections from the United States, Europe, Japan, and the United Kingdom. However, there could be some major issues with Chinese regulatory authorities. In fact, China fears this acquisition could escalate its current trade war with the United States, which could potentially harm its interest. Indeed, if the NVIDIA acquisition of Arm is approved, Arm architectures will become U.S. technologies and this would mean that several dozen Chinese companies blacklisted and on the U.S. Department of Commerce’s entity list would not be able to license Arm architectures or use any product using Arm’s IP. This would make it hard for a great majority of Chinese technology companies to compete in the mobile and IoT markets.

China could, therefore, use its control over Arm China to challenge NVIDIA’s acquisition of Arm. The Chinese branch of Arm is considered a Chinese entity that abides by Chinese laws. In fact, SoftBank sold a 51% majority of the Chinese branch of Arm to a group of Beijing investors in 2018, which allows Arm China to control all of Arm’s interest in China, including licensing Arm’s IP, its marketing activities, and customer care. Arm China claims that 27% of Arm’s global licensing revenue comes from China. This provides enough incentive for Chinese regulators to block the NVIDIA-Arm deal if no compromises are made between Arm China and the Cambridge-based headquarters. Again, migrating Arm’s architectures to open source may be a good fix to address the Chinese conflict, but this strategy may take months to implement properly.

In conclusion, the road to closing the Arm deal could be quite long and bumpy for NVIDIA, but, if successful, it could prove quite rewarding. If successful, the Arm acquisition could enable NVIDIA to become the most innovative semiconductor company across many industries and markets. However, this process may be full of legal hurdles that could distract management from focusing on what it is best at: technology innovation and market transformation.