Artificial intelligence was a dominant theme at this year’s Mobile World Congress, the show that foretells the industry's future direction. The most exciting use of AI in mobile networks is for vRAN. Since AI requires high-caliber computing and radio networks require near-real-time, low-latency processing, in the future, there will be an intense focus on the processors used in vRAN infrastructure products.
Among the top global infra vendors, when it comes to vRAN acceleration, Ericsson has historically been firmly in the Intel/x86-based CPU camp, and Nokia in the Arm ASICs camp. However, unlike its peers, Samsung Networks has been offered both solutions, giving operators a greater choice.
The introduction of AI and the increasing role of GPUs make these options even broader and more diverse. Ultimately, the major vendors have to provide solutions to most of these options, if not all, to cater to operators' needs. In such a scenario, Samsung has already gotten a head start and is leading the race now, and is also actively collaborating with ecosystem partners on upcoming solutions.
Diverse vRAN configurations and processor architectures
Intel has pioneered and owns a majority share of the vRAN processor market. Its Xeon scalable processors power most global vRAN/Open RAN deployments today. With the evolution of its CPU cores, Intel has been offering increasingly higher vRAN performance and features, the latest being the sixth-generation Xeon (Granite Rapids-D). Intel has chosen the “lookaside” configuration with a built-in vBoost accelerator.
However, as 5G networks and technology evolved, especially with higher-order MIMO such as 64T64R, many vendors proposed specialized ASIC-based accelerators working in an “in-line” configuration, instead of Intel’s general-purpose processors. Qualcomm and Marvell are the primary vendors supplying these ASICs, viz. Qualcomm Dragonwing X100 Accelerator and Marvell Octeon series. Both use the Arm microarchitecture. The commercial deployment of these Arm-based solutions is still in its infancy, but is starting to ramp up.
Cloud giant Amazon Web Services (AWS) and AI giant Nvidia have recently entered the vRAN market. AWS supports vRAN accelerators on its Arm-based Graviton processors (ASICs) instances, and Nvidia has developed its GPU-based AX800 accelerator. Both these accelerators are yet to be commercially deployed.
Nvidia's AI-RAN effort is the latest sensation in the mobile industry. It has been gaining considerable industry attention, with many operators and vendors collaborating. Most of these are still in the proof-of-concept and early trial stages.
With AI poised to play a significant role in different aspects of mobile networks, including RAN, deciding what kind of processors and architectures will be used for what applications and use cases has become a major industry discussion topic: Is it general-purpose CPUs, ASICs, or GPUs, whether x86, Arm, or something else?
vRAN/Open RAN infrastructure vendor landscape
As vendors started developing vRAN /Open RAN systems, they adopted one or the other processors and architecture to optimize their R&D investments. For example, Ericsson opted for Intel’s Xeon CPUs with “lookaside” configuration. They even forcefully defended this approach through detailed analysis and whitepapers. Since Intel was the pioneer and offered a software reference design called FlexRAN based on its processors, it was easy for many early vRAN players, including Samsung, Ericsson, Mavenir, Rakuten and others, to adopt it.
On the other side, another major infra provider, Nokia, espoused Arm ASICs-based “inline” configuration, working closely with Marvell. Other vRAN players like Fujitsu and NEC have adopted Qualcomm’s in-line Arm solution. Recently, Mavenir, Rakuten and Viettel have announced support for Qualcomm’s solution as well. Nokia, Marvell and Qualcomm have also presented their own analysis and reports claiming the benefits of the dedicated inline accelerator.
Because of Nvidia's outsized influence in the AI domain, almost all vendors are either already collaborating with it or planning to collaborate on GPU-based vRAN accelerators and AI-RAN efforts.
In my opinion, because of the diversity of global operators, no matter where they have started, major vendors have to support most of these processors and configuration options.
Samsung embraced processor diversity early, setting the trend for the industry
Realizing the importance of choice, vRAN leader Samsung has embraced all the options: Intel, AMD, Arm and Nvidia, early on, offering choice to operators.
As mentioned, adopting Intel was a natural decision because of the early availability. Samsung has supported successive generations of Intel Xeon processors, including the latest sixth-generation, on which it claimed the industry’s first vRAN call in 2024. It even showed its multi-cell vRAN testing with sixth-generation Xeon at MWC 2025. Because of Intel's dominant vRAN market share, most of Samsung’s vRAN deployments are also Intel-based. Samsung has also offered ARM-based lookaside accelerators. Although not publicly known, it seems to use Marvell Technology’s ASICs.
Keeping the spirit of diversity, Samsung started collaborating with AMD very early. It successfully completed the industry's first end-to-end call with AMD EPYC 8004 (aka Siena) processor in 2024 and multi-cell testing using both the Siena as well as the latest generation (Turin) processor in early 2025. Further, Samsung was the first to make a data call on AWS Graviton instances in 2024. More importantly, it demonstrated AI-RAN proof-of-concept with NVIDIA in its research lab later in the year.
To support this multitude of processor platforms, Samsung has developed a highly versatile and flexible vRAN software platform that can be easily ported to all processors and architectures. As industry moves toward this diverse processor landscape, Samsung has gotten an early start in the processor race and a formidable edge against competitors. It will be interesting to see how the company can leverage this edge to win the marketplace.
Prakash Sangam is the founder and principal at Tantra Analyst, a leading boutique research and advisory firm. His monthly newsletter is at TantraAnalyst.com/Newsletter, or listen to Tantra's Mantra podcast.
Op-eds from industry experts, analysts or our editorial staff are opinion pieces that do not represent the opinions of Fierce Network.