NVIDIA has abandoned its plans to buy Arm in a $40 billion deal, which fell apart amid stern pushback from regulators about anticompetitive concerns. But the deal's collapse, executives said, would not change NVIDIA's strategy or technology roadmap.
CEO Jensen Huang said the company will forge ahead with plans to launch a series of new Arm-based processors for markets such as autonomous cars, data centers, supercomputers, and industrial robots.
"You should expect us to do a lot of CPU development around the Arm architecture," Huang said during a conference call with analysts last week.
NVIDIA already has a close relationship with Arm and it takes advantage of Arm CPU cores in a wide range of products, including its family of data processing units (DPUs), which can offload networking, storage, and other behind-the-scenes workloads in the data center from the CPU. NVIDIA also taps Arm CPU cores in its Orin SoC targeting self-driving systems in cars and single-board computers (SBCs) for autonomous robots.
In addition, the company is leveraging Arm to develop a data-center CPU called Grace to take on Intel's and AMD’s chips in the high-performance computing (HPC) market. Huang said NVIDIA is slated to launch the Grace CPU in the first half of 2023.
A Deal Disarmed
In September 2020, NVIDIA agreed to buy Arm for about $40 billion from Japan's SoftBank Group, in what would have been the largest semiconductor deal ever.
The deal would have given NVIDIA control over the instruction set architecture (ISA) at the heart of most of the world’s smartphones. Arm is also playing an increasingly prominent role in a range of other markets, including in cars and on factory floors. On top of that, the deal could have opened the door for NVIDIA to collaborate more closely with Arm, giving it another way to challenge AMD and Intel in the lucrative data center market.
"We believed that would accelerate Arm’s focus on high-performance CPUs and help expand into new markets, benefiting all our customers in the entire ecosystem,” said Huang.
For around a year and a half, NVIDIA worked to reassure regulators that Arm was in safe hands and that the company, dubbed the "Switzerland" of the semiconductor industry, would remain a neutral and independent partner to any company that wanted to use its IP. But NVIDIA struggled to win over officials that worried the deal would tilt the balance of power in the industry too far in its favor, harming competition in the process.
In December, the Federal Trade Commission sued to block the deal, and other regulators around the world opened probes into whether the deal would hurt competition. "We gave it our best shot. But the headwinds were too strong, and we couldn't give regulators the comfort they needed to approve our deal," said Huang.
Architecture Lover
Even though the deal fell apart, NVIDIA said it would keep its 20-year architecture license with Arm. That will allow the company to keep building chips based on Arm's technology, with "the full breadth and flexibility of options across technologies and markets." The license gives the Santa Clara, California-based company the freedom to create its own custom Arm-compatible CPU cores from scratch for whatever purpose it wants.
This is the same license that Apple uses to develop its A-series SoCs for smartphones and tablets as well as the M1 chip in its Mac laptops. Qualcomm is also leveraging its architecture license to build out chips for PCs.
Companies without an architectural license have to choose CPU cores from Arm's catalog of chip designs.
Huang explained that the Grace CPU is the first in a whole range of Arm-based processors NVIDIA plans to build. He remains convinced that Arm—increasingly an alternative to Intel’s x86 architecture in data centers—will play an even bigger role in the technology industry going forward.
"We intend to bring the full spectrum of NVIDIA's accelerated computing platform to Arm CPUs," said Huang.
At the same time, Huang said NVIDIA is not trying to play favorites and will continue to support industry-standard x86 platforms. For instance, the company already taps AMD's EPYC processors in its DGX A100 systems.
"Our strategy is accelerated computing. That's ultimately what we do for a living," he said.
"Whether x86 or Arm, we will use the best CPU for the job," added Huang. Together with partners in the computer industry, we will offer the world's best computing platform to tackle the impactful challenges of our time. You’re going to see a lot of exciting CPUs coming from us, and Grace is just the first example."
Future Fortunes
Even though it abandoned its deal for Arm, the collapse of the merger is not going to derail the company's fortunes. NVIDIA is already the most valuable U.S. semiconductor maker, having passed Intel back in 2020.
NVIDIA reported sales of $7.64 billion in the fourth quarter, up 53% from a year ago, driven by booming demand for its graphics processing units (GPUs) used in data centers, game consoles, and personal computers, said NVIDIA finance chief Colette Kress. Net income more than doubled to $3 billion, or $1.18 per share, she added.
The company is also getting a boost from cloud-computing giants and other corporations as they add its graphics processors in their sprawling data centers to carry out machine-learning workloads, such as recommending content or classifying photos.
Sales in NVIDIA's data-center business soared more than 70% in the fourth quarter to $3.26 billion, a record high, bringing the business to more than $10 billion in annual sales for the first time. The company said revenue in its core gaming GPU business jumped 37% year-over-year to $3.42 billion.
It is also muscling through the supply-chain issues the electronics industry has grappled with for months. Rising demand for chips has caused a global chip shortage that some executives estimate could last into 2023.
NVIDIA has $9 billion in long-term supply obligations from foundries and other suppliers, up from $2.5 billion in the same quarter a year ago. “We expect supply to improve each and every quarter going forward,” said Huang.