Nvidia reportedly is preparing to end its troubled 16-month bid to buy chip designer Arm for $40 billion, a proposal that has drawn the ire of competitors and the scrutiny of skeptical regulators that have worried about the antitrust implications in the highly competitive and far-reaching semiconductor space.
According to a Bloomberg report that cited unnamed sources familiar with the matter, Nvidia officials are telling partners that they don’t believe the deal can be closed. At the same time, Softbank – the Japanese multinational that bought Arm for $31 billion in 2016 – is preparing to take the British chip designer public in the wake of the failing Nvidia deal.
Interestingly, analyst and InternetNews contributor Rob Enderle predicted this turn of events – and said he sees positives for Nvidia in this direction (see Why Not Buying ARM May Be Better for NVIDIA).
Bloomberg reported that no final decision has been made. However, in statements to the media, officials with Nvidia and Softbank said they are committed to the deal and Arm has referred inquiries to Nvidia’s statement, which reiterated the GPU maker’s belief that the deal would be good for both Arm and for competition in the industry.
The report comes after months of pushback by regulators in both the United States, the UK and European Union, who are closely investigating the deal over worries about anti-competitive and security issues. At the same time, the U.S. Federal Trade Commission (FTC) filed an administrative complaint aimed at blocking the acquisition, arguing that the combined company would inhibit innovation in data centers and in emerging markets like automotive technology.
It could take even more months to sort out of these regulatory investigations and lawsuits, which could put Nvidia plans on hold even longer. In addition, regulators in China were waiting in the wings and Nvidia officials said last year they expected tough scrutiny from them as well.
Some Nvidia Competitors Balk
Nvidia also caught flak from tech giants like Microsoft, Google and Qualcomm – a longtime Arm partner that has used the Arm architecture for its mobile chips – which argued that the deal would enable Nvidia to make changes to Arm’s design to favor its own processors over those of other manufacturers. However, other Arm partners, including Marvell, Broadcom and MediaTek, last year came out in favor of the deal, saying it would ultimately benefit their businesses.
Arm doesn’t make chips; rather it designs the architecture for chips and licenses those designs to a broad array of chip makers that make processors for everything from the smallest mobile devices to data center systems and supercomputers. The Fugaku, which sits at number one atop the Top500 list of the world’s fastest supercomputers – runs on the A64FX processor designed by Fujitsu and based on the Arm architecture.
Arm made its name designing low-power systems-on-a-chip (SoCs) for mobile devices like smartphones and tablets, but over the past decade has expanded its reach into the data center and the cloud, becoming a viable alternative to dominant chip maker Intel and its x86 brethren AMD. Nvidia officials have said the company’s success has been fueled by the broad ecosystem of partners Arm has built through its firm stance of neutrality within the industry.
When Nvidia in September 2020 announced its intent to buy Arm, concern quickly turned to the future of Arm’s neutrality and a possible hit to innovation in the semiconductor industry.
Competing with Intel and AMD
Enderle, principal analyst with The Enderle Group, didn’t give the Nvidia-Arm deal much chance of succeeding.
“I doubted Nvidia would get this done given the regulatory hurdles and felt that the restrictions put on the combined company would be excessively onerous,” Enderle told InternetNews. “This is a better path for both companies; Nvidia can still license and partner with Arm, getting the majority of the benefit while avoiding the crippling controls that might have been put in the combined entity.”
The outsized influence the tech world has over consumers’ lives also has hardened the regulatory environment in general for massive companies in the industry, which made the chances of Nvidia being able to buy Arm even more remote, he said.
“Facebook and Google have soured the well for significant acquisitions like this for now, and until something changes, these types of deals are likely not doable,” the analyst said. “They were the source of the concerns around big tech misbehaving and Amazon. All three – you could also include Apple – are the source of the huge antitrust cloud, making acquisitions like this virtually impossible. At the moment, there is a lot of regulatory pressure focused on not allowing big tech companies to get [bigger].”
In addition, much of that scrutiny is aimed specifically at U.S. tech companies.
An Evolving Chip Space
The semiconductor industry for several years has been a space in flux, with chip makers looking to grow their portfolios to address the rapid changes in IT brought on by such factors as the continued growth in mobility, the expansion of the cloud and rise of the edge, and such emerging workloads as AI and machine learning.
Intel not only builds CPUs but also field-programmable gate arrays (FPGAs) and is on the verge of rolling out its own GPUs, such as Ponte Vecchio and Arc. AMD continues to build out its CPU (Epyc for servers and Ryzen for clients) and GPU (Radeon) offerings, and there are a growing number of other established companies and startups looking to gain traction in the silicon space, many of them leveraging the Arm architecture.
There has been a wave of consolidation in the chip space over the past several years – Intel inherited its FPGA business when it bought Alterra for $16.7 billion in 2015. However, not every deal has made it through the regulatory process. In 2016, Qualcomm proposed buying fellow semiconductor manufacturer NXP for $44 billion, a deal Qualcomm ditched two years later in the face of strong resistance from Chinese regulators.
Collapse of Deal Could Help Nvidia
With Arm in the fold, Nvidia would have leveraged its new capabilities to put out its own combined CPU-and-GPU offering. According to Enderle, even that might be helped by the collapse of its effort to buy Arm.
“Overall, this might end better for Nvidia, which can move to build an alternative part immediately and are no longer held up by regulatory approvals,” he said. “This move might improve their time to market for an integrated CPU-GPU offering.”
The reports of the eventual demise of the deal come a day after Facebook parent company Meta announced it is building what promises to be the world’s largest and most power AI supercomputer – the AI Research SuperCluster – based on Nvidia’s Tensor Core A100 GPUs and other offerings, including its edge systems, InfiniBand fabric and software.
In addition, earlier this month, Nvidia bought Bright Computing, a company that makes high-performance computing (HPC) cluster management software for on-premises and cloud environments.
Read next: NVIDIA’s Huang Gets Overdue Recognition