As featured in Fierce Electronics.
On December 2nd, 2021, the FTC announced that it would move to block Nvidia’s acquisition of Arm Holding. This most recent complaint has become another link in a global chain of regulatory legal challenges and reviews that could sink Nvidia’s $40 billion bid first announced on Sept. 13, 2020.
The deal was bound to be a tough one to close from the outset as marquee Arm licensees such as Qualcomm, Alphabet, and Microsoft protested the union. However, Nvidia did manage to muster public support from Mediatek, Marvell, and Broadcom support for the transaction shortly after.
A Union in Doubt
What has prompted regulatory bodies in the UK, EU, and now the United States, to apply added scrutiny to Nvidia’s acquisition of Arm? The FTC complaint can be considered a culmination of months of arguments against the transaction that have matured since the UK’s Competition and Markets Authority launched its initial investigation into the deal almost a year ago.
The crux of the problem is the treatment of Arm’s IP core technology which it licenses to largely fabless chip designers such as Nvidia, Apple, Mediatek, Qualcomm, and hundreds of other firms. Is core IP a key input into the value chain of the logic sector or not? Should IP core and system-level IP such as Arm’s Neoverse be treated as a tool or as an essential input into finished logic devices such as the smartphone SoC, a market that Arm’s IP core dominates?
To properly answer these questions in the context of antitrust, we need to be very clear that Arm is not a chipmaker. For this reason, Nvidia would not acquire end market share through their acquisition of Arm. They would gain influence across the fast-growing expanse of end markets for electronics that use Arm-based chips. More concerning per the FTC complaint, Nvidia would acquire effective control over Arm’s ecosystem and a vital chevron in the logic industry value chain – the IP core.
Arm’s licensee roster is comprised of many of Nvidia’s key competitors in markets such automotive, data center, personal computing, gaming, mobile edge computing, and the vast IoT. Herein lies the conflict of interest at the core of the FTC complaint. As the controlling entity, Nvidia could intentionally or unintentionally influence the direction of Arm’s technology in a way that gives them an unfair advantage over Arm/Nvidia licensees who compete directly with Nvidia in strategic markets.
High Hurdles and Risks to Competition
Nvidia faces some significant hurdles and growing concern among regulators regarding the risk to competition in the market. For one, the company’s assertion that its leadership in AI would enable the combined company to become the “premier computing company for the age of AI” suggests incorporation of its GPU technology and proprietary CUDA (Compute Unified Device Architecture) platform into the Arm technical roadmap in some substantial form. Why wouldn’t they? After all, Nvidia is a very profitable GPU company.
For Arm’s licensees who compete with Nvidia in the coveted “AI market,” Nvidia’s ambitions come off less as an acceleration of Arm ecosystem innovation and more as a threat to the future technological competitiveness of their products. The fear is Arm under Nvidia’s influence could potentially promote and prioritize architectural optimizations and roadmaps in favor of Nvidia GPU devices over competing neural processing logic and system designs by AWS (Inferentia) for data center and Qualcomm (Adreno) for automotive and edge computing.
In its current position, Arm is not obligated to play favorites to any licensee, including Nvidia. While it is a stretch to characterize Arm as the “Switzerland” of the semiconductor industry, its perceived neutrality is an outcome of its role in the industry value chain. Arm and its competitors, the emerging yet nascent RISC-V and established licensed IP core ecosystems, have all played an important role in “democratizing” chip design. This has fostered competitive diversity across the semiconductor industry along with the popularization of the fabless business model.
Nvidia’s ownership of Arm’s IP core assets and technical roadmap would undoubtedly put the current competitive dynamics at risk with potentially deleterious implications for Nvidia’s competitors who happen to also be Arm licensees. These risks will be tough to mitigate. Regulators will need to thoughtfully contemplate them as they deal with Nvidia in negotiations and legal challenges.
Remedies in Waiting?
Early on, Nvidia CEO Jensen Huang provided assurances that if the deal went through appropriate “firewalls” would be put in place to mitigate apparent conflicts of interest that might foster anticompetitive behavior in the future. Surprisingly, Nvida has yet to publicly disclose details on remedies that it is working on with regulators and unsettled Arm customers to temper antitrust concerns that have only worsened over time.
While Nvidia asserts that it intends to maintain Arm’s “neutrality” after deal close, the question boils down to how? Remedies that have been suggested with no substantial detail include treating Arm as an independent entity and brand under Nvidia. It is difficult to imagine how such a mechanism would work without a creative “neutral” governance structure. Would Nvidia be open to having Arm steered by a standalone corporate board comprised of representatives from Nvidia’s primary competitors among Arm’s licensee community?
There lies the rub. Will the remedies that would be required to make the acquisition of Arm palatable to regulators and the company of disgruntled Arm licensees dilute or even eliminate the synergies of the transaction? It is difficult to imagine a construct that would allow Nvidia to maintain the perceived value of the acquisition of Arm for itself. It’s also likely that a “neutral” arrangement would limit the combined company’s ability to deliver suggested technological innovations which would water down the value of the merger to the Arm ecosystem.
As Jensen Huang has said on several occasions, the success of Nvidia is not dependent on the Arm deal going through. It’s difficult to disagree with this statement. Nvidia is already engineering highly competitive and potentially disruptive systems for strategic growth markets including data center, automotive, 5G networks, and edge AI. Nvidia’s Project Grace is a great example of how it is applying Arm IP into a system design with the potential to be as game-changing for data center compute as Apple’s MI has been for personal computing.
Over the years, Nvidia has proven to be a formidable partner for Arm and Arm licensees without owning the IP core technology. With or without Arm, Nvidia has game, no pun intended.