It’s a testament to the collective professionalism of Arm executives and employees that the company hasn’t let the seemingly endless review processes for NVIDIA’s proposed acquisition distract it from product development across its expansive portfolio.
Amidst the lingering uncertainty of potential absorption into a larger company, many acquisition targets lose focus on projects as key employees with highly sought skills tire of the organizational ambiguity and seek greener pastures elsewhere. So far, Arm has resisted this hazard, with 2021 being one of its most fruitful years for new architectures and products. Although the company did freeze hiring and effectively cut employee pay, so far, it hasn’t derailed progress on many fronts.
As I detailed this Spring, Arm introduced the next v9 generation of its core architecture and new reference designs suitable for cloud, HPC and high-end mobile applications. Last week, it unveiled several enhancements to improve IoT product and software development at the Arm DevSummit that could shorten the IoT design cycle by 40 percent, to “as little as three years.” Meanwhile, the company waits for regulatory reviews that could stretch well beyond NVIDIA’s March 2022 completion target into next summer.
Arm initiatives stretch from cloud data centers to smart devices
Arm’s server-class Neoverse family gets most of the attention lately since it’s the foundation for a new class of cloud-optimized products such as AWS Graviton and Ampere Altra, however, the bulk of Arm’s business is in mobile, embedded (IoT) and automotive processors. Arm announced three significant IoT products and services at its recent DevSummit.
- Arm Total Solutions for IoT are a set of reference designs that combine embedded Corstone processors with software, including an OS, middleware, application software and development tools, optimized for one of four scenarios: general purpose CPU, general-purpose machine learning (ML), keyword detection CPU and keyword detection ML. These are available as locally installable SDKs or as an AWS system image using Arm’s next big -announcement, Virtual Hardware.
- Arm Virtual Hardware is a cloud-based product available on the AWS Marketplace that provides a virtual model of the Corstone SoC that can be used for software development in place of actual silicon. Using a virtual ‘digital twin’ of a device speeds the development process by eliminating the need to acquire and test hardware farms (often with pre-release silicon) and allowing the IoT development process to exploit agile methodologies like continuous integration and development (CI/CD) and DevOps automation tools. The device models allow simulating memory, I/O and peripheral interfaces and embedded ML hardware. Initially, Arm provides Virtual Hardware for the Corstone-300 subsystem using the Cortex-M55 processor and Ethos U55 NPU.
- Project Centauri is an analog for the Cortex-M (embedded microcontrollers) product line to Project Cassini, launched last year, for the Cortex-A (cores for high-performance mobile and data center SoCs) with a set of device and platform standards and reference implementations for device boot, security and cloud integration. Corstone is Arm’s IoT reference platform. Corstone is Arm’s IoT reference platform that includes Cortex-M cores and various System IPs depending on the configuration.
NVIDIA is still a willing (and waiting) suitor
NVIDIA knew the approval process for the Arm deal would be lengthy, complicated and fraught with political peril, however, its quest will likely stretch well into next year. The company recently made unspecified concessions to the European Commission reviewing the deal that will extend the process for several months. The Commission will now seek feedback from rivals and customers before ruling on the concessions which could lead it to extend to the investigation. Such an extension could cause Arm and SoftBank to reconsider Plan B, an IPO aided by outside investment from Qualcomm, which has previously indicated its willingness to capitalize Arm, other Arm partners or an investment group.
NVIDIA and Arm aren’t ready to pull the ripcord since they remain convinced that the deal benefits both parties in several ways.
- Benefits for Arm stem primarily from NVIDIA’s deep pockets, where profits from its lucrative gaming, deep learning, HPC and visualization hardware would allow Arm to better exploit the tremendous opportunities in AI, mobile devices, edge computing, robotics and autonomous transportation. Arm’s financial situation makes it difficult to compete with tech industry giants attacking these markets, accounting for the capital it would raise via an IPO. Indeed, as Arm CEO Simon Segars said when the deal was first announced:
The combination of Arm and NVIDIA is a better outcome than an IPO. The level of investment that will be needed to lead in AI will be unprecedented. We’ve been down this road before when we predicted a major market shift in 2016. We knew that we needed to invest heavily in our products, talent, and technology to take advantage of the opportunities ahead.
It remains unclear whether Arm’s current financial position would support a strong IPO, however, the combination of its robust development pipeline and perhaps some outside VC or sovereign wealth fund (SWF) investment would strengthen its market appeal.
- Synergies for NVIDIA come not just from its ability to more closely collaborate with Arm, which provides many technologies critical to its products, but via better incorporating the development model embodied (and arguably pioneered) by Arm. Indeed, Arm facilitated a significant shift in semiconductor design over the past decade as process miniaturization, merchant foundries, component libraries and design automation software enabled an exploding number of product developers to combine previously discrete functions on composite systems-on-a-chip (SoCs). Arm’s CPU IP is crucial to many NVIDIA products and its computing ecosystem complements NVIDIA’s leading artificial intelligence, GPU and networking technologies. Combining companies with complementary strengths creates many synergies that lead to greater innovation, R&D efficiency and product development. Not only will Arm have the necessary resources to accelerate and expand its processor roadmap, but the two will be able to combine elements of each company’s IP portfolio in ways not feasible as separate entities.. The combination will also increase NVIDIA’s developer ecosystem and reach more than seven-fold by adding millions of Arm programmers and designers.
- Benefits to the Arm ecosystem will emerge from NVIDIA’s pledge to “expand Arm’s IP licensing portfolio with NVIDIA technology.” Counter to the FUD surrounding NVIDIA’s announcement that predicted dire implications for the Arm ecosystem of handing its Crown Jewels over to a potential competitor, NVIDIA isn’t about to kill the Golden Goose since it needs Arm licensing revenue to make the $40 billion deal financially viable. Skeptics of the deal reasoned that NVIDIA would turn Arm into a subsidiary more focused on technology beneficial to its AI, rendering and robotics products than general-purpose architectures applicable to the broad swath of existing Arm licensees. These doubters were not appeased by NVIDIA’s commitment to maintaining Arm’s open, customer-agnostic licensing model nor its pledge to “expand Arm’s IP licensing portfolio with NVIDIA technology.” To them, these were perfunctory statements made to assuage regulators, not principled pledges. In responding to EU and UK regulators, NVIDIA has undoubtedly reiterated, if not strengthened, its commitment to an independent Arm, however, it might not be enough to assuage abstract concerns over “national security.”
There’s a strong case that Arm and NVIDIA are better together, but Arm wisely isn’t letting its development efforts wane while waiting for a deep-pocketed saviour. Oddly, Arm’s success bringing new technologies to market, like the updated Neoverse platforms with v9 architecture, Corstone and the other IoT enhancements announced last week strengthen the case for its independence via an IPO.
Given the enormous success of Apple’s Arm-architecture M1 Silicon, which has delivered performance, power efficiency and feature integration to its MacBook and iPad Pro products that Intel or AMD can’t match and the AWS Graviton instances which power an increasing percentage of its internal systems, it’s conceivable that key Arm licensees like Apple, AWS, Google, Qualcomm and Samsung might chip into a RISC-V type structure to guarantee an independent Arm.
Regardless of the financial structure, an independent Arm remains Plan B, and in the near-term expect NVIDIA to do whatever it takes to assuage EU, UK and Chinese regulators via deal concessions, stronger regulatory oversight or additional investments (beyond those already announced and here) in the UK.
Postscript [added 09:45 PT]: Reuters reports that “Nvidia suffered a setback on Wednesday as EU antitrust regulators opened a full-scale investigation into its $54 billion bid for British chip designer ARM on concerns the deal could lead to higher prices, less choice and reduced innovation.” The Commission stated that it would make a final decision on whether clear or block the deal by March 15, 2022.