Nvidia shrugged off its failed bid to acquire Arm Holdings to end the fourth quarter of fiscal year 2021 on a high note. The chipmaker’s net income doubled to $3 billion from the year-ago period on $7.6 billion in revenues.

“When we entered into the transaction in September 2020, 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,” CEO Jensen Huang said. “We gave it our best shot, but the headwinds were too strong, and we could not give regulators the comfort they needed to approve our deal.”

While much of the conversation during this week's earnings call centered on the chipmaker's failed Arm bid, Nvidia’s Arm ambitions are anything but dead. In fact, Nvidia plans to launch a series of new Arm processors, Huang told investors, according to a transcript.

Nvidia is no stranger to the Arm architecture or even Arm CPUs, however, these products have historically targeted consumer and developer applications, like entertainment consoles and single-board computers.

Nvidia early last year broke this mold, announcing its first data-center focused Arm CPU dubbed Grace.

Named for Grace Hopper, the U.S. computer programming pioneer, the CPU was based on Arm's Neoverse N2 core architecture and is specifically designed to accelerate high-performance compute (HPC) environments, like artificial intelligence (AI) training.

Huang, at the time, vehemently denied any interest in competing with CPUs from Intel, AMD, Ampere, or Amazon. Instead, Grace was designed to address a problem unique to HPC environments: memory bottlenecks throttling the performance of larger AI workloads. Grace is slated to launch in 2023 alongside a yet-to-be-announced GPU.

Meanwhile, Huang’s stance on this appears to have softened over the past year.

“We will deliver on our three-chip strategy across CPUs, GPUs, and DPUs. Whether x86 or Arm, we will use the best CPU for the job. And, together with partners in the computer industry, offer the world’s best computing platform to tackle the impactful challenges of our time,” he told investors this week. “You’re going to see a lot of exciting CPUs coming from us, and Grace is just the first example.”

“Our strategy is accelerated computing. That’s ultimately what we do for a living,” he added.

Whether we’ll see a data-center class Nvidia CPU that can go toe-to-toe with AMD and Intel’s best remains to be seen. Nonetheless, Nvidia’s Q4 revenues offered a compelling reason why the chipmaker might want to bring such a chip to market.

Data center revenues accounted for nearly half of the company’s revenues during the quarter, up 11% sequentially and 71% from the prior year.

For the 2021 fiscal year, data center-related products accounted for $10.6 billion of the vendor’s $26.9 billion in annual revenues.

“Data center growth was once again led by our compute products on strong demand for Nvidia AI. Hyperscale and cloud demand was outstanding, with revenue more than doubling year on year,” CFO Colette Kress said.

The past year has seen strong adoption of Nvidia’s A100 GPUs in major HPC projects and by large enterprises to advance AI applications. Kress highlighted deals with Square’s parent company Block and Snapchat, which are using Nvidia GPUs for appointment scheduling and customer recommenders respectively.

“AI is unquestioningly growing, and it’s growing incredibly fast,” Huang said.

While Nvidia has long held the crown when it comes to data center and AI acceleration, its reign may be short lived as the chipmaker faces its stiffest competition in years.

Long-time rival AMD’s Instinct MI200 GPUs, announced late last year, claim a 490% lead over Nvidia’s top-of-the-line A100 in “highly-precise, floating-point 64 calculations,” according to Forrest Norrod, SVP and GM of data center and embedded solutions at AMD.

The GPUs will make their debut early this year in the U.S. Department of Energy’s Oak Ridge National Laboratory’s Frontier supercomputer. Built in collaboration with Hewlett Packard Enterprise, Frontier is expected to deliver 1.5 exaFLOPs of computing power.

And AMD has something Nvidia, at least for now, does not: CPUs of its own to pair with its GPUs.