Prying Open the NVIDIA-Arm Acquisition: The Controversies, Actions, and Results
Last week the NVIDIA-Arm deal fell through, but it left us some questions to answer. What were the controversies and issues standing in the way? What did NVIDIA hope to gain? Where does that leave Arm?
The semiconductor industry can finally rest easy. With Arm firmly remaining in SoftBank's hands—and instead slated to hit the New York Stock Exchange (NYSE)—NVIDIA's landmark IP play has stalled indefinitely.
Arm and NVIDIA deal called off. Image used courtesy of NVIDIA
This acquisition has been in the works for quite a while, with controversies surrounding the talks from the beginning.
In this article, let's look at this deal and attempt to see why the negotiation might have been doomed and what NVIDIA was trying to gain from the acquisition.
Nvidia Targets Emerging Technologies
In late 2020, All About Circuits first highlighted NVIDIA's conflicts of interest in purchasing Arm. The skyrocketing popularity of AI and machine learning (ML) drove chipset design philosophies—headlined by rapid computation, raw power, and efficiency.
While graphics processing units (GPUs) excel at ML tasks involving deep learning (DL) and parallel processing, they're not suitable for all ML applications.
These components often require greater specificity in algorithm design, and they're generally more power-hungry while performing tasks. This power requirement is a major problem for portable and Internet of Things (IoT) devices.
Conversely, Apple's A15, Qualcomm's Snapdragon 888, and Huawei's Kirin 9000 CPUs are lauded for shouldering this computational burden without siphoning excess battery power.
Additionally, one should also consider—if COVID-related shortages and crypto-mining mania have taught us anything—the overall volatility of the GPU market.
Meanwhile, purchasing Arm would give NVIDIA an excellent opportunity to expand into the central processing unit (CPU) space. It'd be an ideal pathway for NVIDIA to internally own the entire processing hardware stack, encompassing AI and ML.
Arm itself owns an overwhelming majority of the mobile processing IP market. It would indeed be advantageous to gain ownership over the firm's Cortex CPU architectures—plus its Ethos family of neural processing units (NPUs).
The Underlying Problems With the NVIDIA-Arm Deal
The problem arises when assessing Arm's current place in the semiconductor market.
The Cambridge-based giant is a crucial supplier for companies like Samsung and Qualcomm. Currently and historically, Qualcomm's Snapdragon processor designs have also famously featured numerous Arm Cortex-A and Cortex-X series cores.
Example of Arm Cortex being used in a Snapdragon CPU. Image used courtesy of Qualcomm
Arm has long licensed its technology out to major players in the mobile space. The company has aided in designing many leading CPUs, and its instruction sets power tens of millions of devices worldwide.
Despite NVIDIA's insistence otherwise, these Arm-reliant companies were afraid of having the rug pulled out from underneath them.
What would happen if those technologies become unavailable, more costly, or locked down?
Though NVIDIA's ambitions have stalled, engineers and tech companies fear the consequences of consolidation in the semiconductor space. They remain wary of chipmaking superpowers who favor themselves over longstanding customers.
Developing Alternative CPU Architectures
Since 1978, the x86 architecture has dominated the CPU market. All Intel and AMD CPUs currently use x86 to power server-centric data centers, supercomputers, or personal devices.
However, NVIDIA's view is that the x86 platform isn't optimally efficient for modern AI and ML workloads, especially in a server context.
NVIDIA wants to capture a greater piece of that market with its newest Grace CPU—which it claims is ten times faster (paired with NVIDIA GPUs) than its existing DGX-based systems.
NVIDIA's Grace CPU. Image used courtesy of NVIDIA
Such a boost could have massive implications for massive-scale AI and high-performance computing (HPC). NVIDIA's Grace CPU development was made possible by licensing Arm's designs.
Instead of merely building the best Arm-based server platform—and selling it for a handsome profit—NVIDIA could effectively lockdown that implementation entirely. Longstanding x86 leaders like Intel and AMD could lose ground if Grace had widespread viability.
They also might not be able to design their own Arm-based counterparts. Both Intel and AMD have experimented (albeit unsuccessfully) with their own Arm architectures in the past; would they have those same opportunities given an NVIDIA takeover?
All in all, competition could suffer, and the industry might struggle to move forward despite the takeoff of next-generation technologies.
Regulators Take Exception
The U.S. and E.U.'s governing bodies met the pending Arm deal with copious skepticism.
America's Federal Trade Commission (FTC) moved to block the acquisition, citing a need to foster competition in the semiconductor space.
The case summary on behalf of the FTC explicitly mentioned data centers and automotive as two focus areas in their grievances. That trial was set for August 9th of this year, though that won't come to fruition.
In its rebuttal, NVIDIA highlighted how its technology could complement Arm's—and how its robust R&D pipelines could help crack x86's firm grip on the data center.
Accordingly, the E.U.'s antitrust officials recently resumed their probe into the NVIDIA-Arm deal back on January 11th.
Overall, regulators worried that a potential buyout could cause price hikes, stifle competitive innovation, and reduce choice. The European Commission anticipated a May 25th decision on the matter. However, with last week's news, that motion is no longer necessary.
Future Prospects
While NVIDIA and Arm have tabled talks, there's no guarantee that the duo won't reconsider at a later date.
Similarly, other players in the semiconductor realm can make a similar play. The proposed NVIDIA-Arm deal has encouraged solidarity across the industry—which would rather have healthy competition than monopolistic growth. While those appeals might be motivated by self-interest, they've been quite strong.
After everything is said and done, two things are clear: Arm is running a lucrative business and is poised to keep turning large profits in the years to come. NVIDIA also remains a top player in its niche. How each continues to expand remains to be seen.
