NVIDIA buys Arm - a good fit, but with points of friction and potential consumer ramifications

Profile picture for user kmarko By Kurt Marko September 15, 2020
Summary:
A poorly-kept secret confirmed - Arm falls to NVIDIA. What now?

Arm
(Arm)

NVIDIA confirmed persistent rumors by announcing late Sunday that it reached an agreement with Softbank to acquire Arm, creators of the processor architecture and reference designs behind the SoCs powering virtually all phones and tablets.

The deal is worth as much as $40 billion (depending on hitting financial performance targets): $21.5 billion, plus $1.5 billion worth of grants to Arm employees (66%) in NVIDIA stock and $12 billion in cash.

The deal's valuation is nominally about $2 billion higher than what Softbank paid four years ago, but the price is much easier for NVIDIA to swallow by primarily paying in equity and exploiting the massive 185% price appreciation of its stock over the past three years. 

In announcing the deal, NVIDIA founder and CEO, Jensen Huang, praised Arm's achievements and its "visionary team of computer scientists in Cambridge, UK", where he committed to building "a world-class AI research center" to serve as the company's European hub for research, business partners and customers. Echoing NVIDIA's shift from a consumer-centric provider of graphics accelerators to a powerhouse in AI technology, Huang positioned the acquisition as part of an overarching AI strategy, writing (emphasis added):

We are joining arms with Arm to create the leading computing company for the age of AI. AI is the most powerful technology force of our time. Learning from data, AI supercomputers can write software no human can. Amazingly, AI software can perceive its environment, infer the best plan, and act intelligently. This new form of software will expand computing to every corner of the globe. Someday, trillions of computers running AI will create a new internet — the Internet-of-Things — thousands of times bigger than today’s internet-of-people.

Huang highlighted the combination's synergies, namely  NVIDIA's AI expertise and R&D budget with Arm's developer ecosystem, massive customer base, "brilliant," open-licensing, customer-neutral business model and IP portfolio. "Our R&D scale will turbocharge Arm’s roadmap pace and accelerates data center, edge AI, and IoT opportunities," Huang wrote.

Good on paper, but numerous points of friction

The NVIDIA-Arm combination would cement NVIDIA as a viable Intel alternative in the data center, rounding out a portfolio of GPU-based AI accelerators, AI development tools and cloud-native software platforms, network components and equipment (from the Mellanox acquisition) and now general-purpose CPU cores and supporting logic. These could be combined and integrated in various ways to suit a range of performance, power, size and cost requirements. The combination is particularly timely now that the Arm architecture has finally emerged as a viable platform for cloud infrastructure, with Arm-based instances available from AWS via its custom Graviton processor, making their way to Azure and the foundation of enterprise products from traditional suppliers like HPE and Arm-centric startups like Bamboo Systems.

Huang expanded upon NVIDIA's data center strategy during a press conference where he emphasized " our seriousness of bringing Arm to data centers." He cited two business models for data center customers.

  1. IP licensing for customer-built, custom CPUs or SoCs
  2. Licensing or customizing NVIDIA-built CPUs. 

Huang noted that the flexibility of the Arm and NVIDIA IP portfolio and licensing model enables systems tailored to various needs that can't be satisfied by a static product list from one company (emphasis added):

So now with our backing and Arm’s serious backing the world can stand on that foundation and realize that they can build server CPUs. Now, some people would like to license the cores and build a CPUs themselves. Some people may decide to license the cores and ask us to build those CPUs or modify ours. So all of those options exist, the beautiful thing is this. There are going to be so many different types of data centers in the future, gigantic cloud ones, smaller ones for enterprise, high-powered density ones for supercomputing, ones they're going to fit in locations all over the world, distributed data centers that are secure and managed like fleets and they're going to be out there connected to IoT devices.

Indeed, even Intel has bent to the need of cloud builders to customize processors for different workloads by offering custom SKUs not on its general product list. 

The data center focus of the deal was reinforced by the fact that Huang completely ignored the phone market which made Arm rich and famous. However, detractors point to potential conflicts that an NVIDIA acquisition could create with Arm's consumer product licensees.

Arm co-founder appeals to UK PM Boris 

 Although Arm CEO and long-time employee Simon Segars publicly endorsed the acquisition and would stay on as CEO, one of Arm's founders, Herman Hauser, is leading the charge to quash, or at least significantly modify the union. In an open letter to the UK Prime Minister Boris Johnson, Hauser has three primary objections to NVIDIA's acquisition. 

  1. The potential for lost jobs at Arm facilities in Cambridge, Manchester, Belfast, Glasgow, Sheffield and Warwick "where thousands of Arm employees work."
  2. The risk to Arm's platform neutrality and business model selling to hundreds of licensees on equal terms as "the Switzerland of the semiconductor industry."
  3. Lost economic sovereignty by the UK, with its only remaining technology company coming under American control and subject to US. OFAC (Office of Foreign Assets Control) regulations. Notable is CFIUS, which has frequently been used to block sales of US-based technology products to countries deemed hostile by the US Treasury.

Hauser cited three mitigating and "legally binding" changes that the UK should demand before approving the deal, 

  1. "Legally binding job guarantees for Arm employees in the UK.
  2. "Legally binding agreement that Nvidia must not gain any preferential treatment over other Arm licensees.
  3. "Britain must get an exemption from the US OFAC regulation so that UK companies are guaranteed unfettered access to our own microprocessor technology."

Huang dismissed questions about regulatory challenges by citing NVIDIA's experience navigating the Mellanox deal saying (emphasis added):

You know that we have a fair amount of experience here recently with Mellanox and the thing that I can tell you about the regulatory discussion –it's logical and it's about protecting competition in the markets. If you look at look at our two companies, just like the case with Mellanox, NVIDIA and Arm are completely complimentary. NVIDIA doesn't design CPUs we have no CPU instruction set, NVIDIA doesn't license IP to semiconductor companies. So in that way we are not competitors.

Segars added that Arm's Chinese operations are already segregated as a joint venture with Arm as the minority owner. Both he and Huang expected approval after explaining to Chinese regulators and the JV owners the complementary synergies of the two companies and its pledge to "continue to support the Chinese ecosystem."

My take

Jensen Huang is a man on a mission to cement the era of GPU computing as the vanquisher of Moore's Law and rightful successor to traditional CPUs that can no longer meet the performance scaling timeline that had been inviolate for decades. Unfortunately, evangelists tend to have tunnel vision focused solely on their noble cause, a singularity of purpose that causes them to miss or minimize complexities, challenges and nuances that might slow their progress. 

The NVIDIA-Arm deal looks perfect on paper, with Arm supplying a critical piece missing from Huang's data center portfolio, however, it comes with a history, customer base and existing agreements that don't always align with NVIDIA's goals. The two key areas that Huang seems to be minimizing are:

  • NVIDIA's alignment with Arm's consumer business, which includes licensees like Apple, Samsung, Qualcomm that have built multi-billion dollar product lines around Arm technology.
  • Geopolitical tensions over the transfer of US-owned technology to foreign companies, particularly Chinese ones. Arm supplies (through Qualcomm and MediaTek) the brains behind phones from Huawei, Xiaomi, Oppo and others. If Arm becomes part of a US-based company, it could further crimp Chinese access to Arm IP, even for internally manufactured SoCs. It is also unclear how seriously the UK government takes the risks highlighted by Hauser of US interference in Arm operations.

Arm as an NVIDIA subsidiary could evolve from being a "Switzerland" into a "NATO," i.e. a less-independent actor that is more aligned with NVIDIA's interests and those of its home country. One possibility is a TikTok-like solution that splits Arm operations into a consumer-centric independent foundation and an NVIDIA-owned technology developer. We should get a better understanding of any objections from Arm licensees in the coming weeks, however, should NVIDIA not placate their concerns, I agree with my colleague Charlie Demerjian, that the deal could usher in a migration of mobile manufacturers to open alternatives like the RISC-V platform.