Arm’s Growing Cloud Server Momentum

Forbes Innovation Cloud Arm’s Growing Cloud Server Momentum Steve McDowell Contributor Opinions expressed by Forbes Contributors are their own. Industry Analyst covering all aspects of Enterprise IT Infrastructure. Following Feb 26, 2023, 04:23pm EST | Press play to listen to this article! Got it! Share to Facebook Share to Twitter Share to Linkedin Arm-based processors grow in the cloud.

NAND Research Arm is the most prevalent processor architecture in the world. Odds are good that the car you drive contains multiple Arm-based processors hidden within, while the phone you carry does. According to Mercury Research, Arm-based processors accounted for 13% of all PCs sold during the fourth quarter of 2022.

Where Arm is making its most significant gains, however, is in the server market, with the public cloud providers leading the way. I recently spoke with Tab Schadt, CEO of Liftr Insights , about Arm’s penetration in the public cloud market. Liftr Insights offers some of the most comprehensive tracking of instance types offered by the top public cloud providers available in the market.

Mr. Schadt told me that Arm-based offerings in the public cloud increased by an incredible 290% in 2022. In addition, arm-based instance types grew to account for over 6% of public cloud offerings at the end of 2022, up from 4.

6% at the start of the year. That’s significant momentum. Arm, the Business While Arm Holdings, Ltd controls the Arm architecture, the company doesn’t actually sell any processors.

Instead, Arm, the company, manages the evolution of the architecture and licenses its technology to processor vendors. There are various licensing models offered by Arm. Some vendors incorporate IP blocks from Arm into their silicon, while others implement Arm’s architecture from scratch, optimizing to solve the problem.

This is a lucrative business. During its most recent fiscal quarter, Arm Holdings reported total revenue of $746M, up 28% year-on-year. Most of that revenue, $446M, was from royalties from products shipping in the market, while $300M was from licensing activity.

Arm’s licensing business was up 65% year-on-year. That’s impressive momentum. Arm Holdings, owned by the private equity firm SoftBank Group, is planning an IPO later this year.

The IPO plans follow a failed acquisition by Nvidia, which was scuttled following intense regulatory pushback in several countries. MORE FOR YOU The Inside Story Of Papa John’s Toxic Culture This Tradition On Georgia’s Jekyll Island Has Locals Looking For Hidden Globes Extensively Drug-Resistant Shigellosis Has Risen This Much Since 2015, CDC Warns Arm’s Journey towards Server Arm has traveled a long road to find a place in the server market, one littered with failed attempts. Two of the earliest Arm-based server processor vendors, Cavium and Calxeda, emerged with grand visions of replacing x86 in general-purpose servers.

Neither company exists today, though Cavium’s intellectual property has entered Marvell’s offerings. After a decade as a stand-alone company, Marvell Technology Group acquired Cavium in 2018. SeaMicro also emerged as an early potential Arm-based server vendor, only to be acquired and rapidly discarded by AMD.

There are several reasons for Arm’s slow ascent into a server. There were, for example, valid questions in the early days about software compatibility and ecosystem enablement. Server software, including applications and development tools, simply didn’t exist for Arm until recently.

Enterprise IT teams have been deploying x86-based servers for over two decades, where Arm never played. The most substantial obstacle that Arm faced in the server space wasn’t technical at all. I recently spoke to a vice president of one of the tier-one server manufacturers.

This executive explained, quite candidly, that the problem with building Arm-based servers is that it simply doesn’t expand the market. Arm-based products, instead, replace existing offerings for anyone building servers. So why spend the money to do that? It’s a valid question.

Arm in the Cloud Public cloud service providers (CSPs) have different concerns than traditional server OEMs. The CSPs, such as Amazon Web Services (AWS), sell IT organizations a business and operating model that increases efficiency. While important, the underlying server technology is secondary to the convenience and flexibility these companies provide.

What CSPs care most about is operational efficiency. It’s basic accounting to say that profits rise and fall based on operating expenses, but operational costs can rapidly increase within a hyperscale data center. The most significant operational expenses for a hyperscaler are those associated with power and cooling.

Microsoft Corporation, for example, reported that energy costs directly impacted income from its Azure cloud business. The company’s 10-Q filing for its quarter ending December 2022 shows that its Azure gross margin declined by approximately 3 points primarily due to higher energy costs. The Arm processor architecture is differentiated by its ability to deliver high performance with incredibly high levels of energy efficiency.

Arm’s energy efficiency story has led to the architecture’s dominance in both the embedded and smartphone markets. Energy efficiency also led Apple to replace all its Intel-based offerings with its own in-house designed Arm-based processors. So it’s no surprise that energy efficiency is also a key driver for Arm in the cloud.

AWS Graviton Amazon was the first company to bring an Arm-based processor to market, even though it only sold the processors to itself. AWS introduced its in-house designed Arm-based processor, Graviton, in late 2018, and the company has released a new generation of Graviton nearly every year since. AWS Graviton delivers value along three dimensions: cost, energy efficiency, and an architecture ideal for cloud-native applications.

The energy efficiency claims are real. For example, NEC Corporation published a study in mid-2022 that demonstrated that building a 5G core infrastructure atop instances powered by AWS Graviton processors resulted in a 72% reduction in power consumption as compared to similar x86-based processors. That efficiency likely has a tremendous impact on AWS’s electrical and cooling bills; Amazon, after all, hasn’t cited energy costs as impacting its AWS margins.

Acquisition cost for the CSPs is another critical driver. Arm-based processors are cheaper than comparable x86-based offerings from Intel and Advanced Micro Devices (AMD). Some of the cost reduction comes from Arm-processor vendors not being burdened with the bloated cost structures of the larger competitors.

Instead, the reductions are passed on to the CSPs, who then pass the savings on to their customers. AWS, by designing and overseeing the manufacturing of its own Graviton processors, has a nearly ideal cost structure. Vertically integrating processor development removes all the margin it would otherwise sacrifice to a processor vendor.

The challenge for AWS shifts from managing acquisition costs for a processor to managing the efficiency of the engineering teams designing the processor. Amazon, without question, is a master of efficiency in a complex environment. Graviton is paying off for Amazon.

According to the latest data from Liftr Insights, Graviton-based instances today account for nearly 19% of the instances offered by AWS, a 6% increase from the year prior. The Rise of Ampere Computing Credible rumors abounded after Amazon demonstrated success with its Graviton offerings that Microsoft Azure and Alphabet’s Google Cloud teams were working on their own competitive Arm-based offerings. However, some of those rumors began to quiet down when a little silicon company called Ampere Computing began shipping its Arm-based server processor.

Founded in 2018, Ampere Computing quickly shipped its first Arm-based Ampere Altra and Altra Max processors. Its primary customers are the public cloud hyperscalers. Still referring to data from Liftr Insights, Ampere grew from having no presence in the public cloud in 2020 to nearly 2% of public cloud instance types today.

Microsoft Azure offers the most Ampere-based instance types, accounting for more than 69% of Ampere-based instances. Alibaba Cloud, Google Cloud, Oracle Cloud Services, and Tencent Cloud also offer ampere-based instance types. Ampere Computing is, so far, the only vendor that has found success delivering Arm-based processors into the server market.

Competitive Environment The success of AWS Graviton and Ampere Computing’s Alta processors paves the way for additional competition. In addition, Nvidia is expected to ship its first Arm-based Grace server-class processor later this year. While I expect Nvidia’s Grace will be initially used in Nvidia’s own DPUs and its DGX deep learning offerings, there’s too much opportunity in the broader market for Nvidia to ignore.

There will also be increased competition from home-grown offerings. Alibaba, for example, has already announced that it is deploying its in-house designed Yitian 710 processor, which offers 189 Arm-compatible CPU cores. Speculation that some of Amazon’s closest competitors are pursuing internal development efforts continues, despite the embrace of Ampere.

Arm’s road into the server space has ironically paved the way for even more alternative architectures. The adoption of Arm within public cloud ecosystem has shown that Arm-based servers are viable. IT organizations have adapted software and operational toolchains to be processor agnostic.

As a result, the barriers to entry are lower for a new processor architecture, such as RISC-V. RISC-V, which doesn’t force royalty payments and license fees, could remove even more margin from the value chain. On the flip side, RISC-V isn’t controlled by a central body like Arm, so there is little guarantee of interoperability across different RISC-V parts.

That won’t stop the market from trying. Ventana Microsystems has emerged with a plan to bring RISC-V to the server and embedded worlds with its Veyron processors. It’s still very early for both RISC-V and Ventana, but keep watching both.

The value proposition is strong, and Ventana has the right management team to drive it to the goal. The Analyst’s Take Arm-based processors deliver real value. While this article focused on what makes Arm attractive to hyperscalers, it’s mostly ignored the compelling value delivered to IT end-customers.

AWS Graviton, Ampere’s Altra, and even Nvidia’s Grace, deliver a nearly ideal environment for cloud-native workloads. The x86 processor is in no danger of going away. Intel and AMD jointly control nearly every on-prem server offering in the industry and over 90% of available public cloud instance types.

The value of alternative architectures such as Arm and RISC-V is compelling, which will only lead to greater adoption in the public cloud and hyperscale space. The real question is whether the server OEMs will deliver Arm into their on-prem and as-a-service offerings. If, as I was told, OEMs truely don’t see market expansion opportunities, then perhaps the OEMs will hear customers asking for Arm-based servers.

Nearly every enterprise today has ESG concerns that Arm-based servers would help mitigate, all while quietly delivering the performance needed. Its a compelling story. Disclosure: Steve McDowell is an industry analyst, and NAND Research an industry analyst firm, who engages in, or has engaged in, research, analysis, and advisory services with many technology companies, which may include those mentioned in this article.

Mr. McDowell does not hold any equity positions with any company mentioned in this article. Follow me on Twitter or LinkedIn .

Steve McDowell Editorial Standards Print Reprints & Permissions.