- Samsung says it completed what it calls the first commercial call using its virtualized RAN software on a Tier 1 U.S. operator’s live network
- The setup ran on Intel Xeon 6700P-B processors, using a single off‑the‑shelf server plus a Wind River cloud platform
- The pitch: consolidate more network functions onto fewer servers to cut power, cost, and prep for AI-driven, “6G-ready” upgrades
Samsung says it has completed what it calls the industry’s first commercial call on a live network using its virtualized radio access network (vRAN) software running on Intel’s Xeon 6700P‑B processors. The company described it as a single-server setup on a Tier 1 U.S. operator’s network, building on a 2024 lab milestone and leaning on Intel Xeon 6 features like up to 72 cores, Intel AMX, and Intel vRAN Boost. (RCR Wireless News)
This matters right now because mobile operators are trying to turn parts of the cellular network into software — not fixed-function hardware — while also dealing with power bills, site constraints, and the slow grind of 5G expansion.
The subtext is competition and timing. If “AI-native” networks are the next marketing wave, carriers want the underlying infrastructure to look more like cloud computing: standardized servers, software upgrades, and hardware refresh cycles that don’t require ripping out everything at the tower.
vRAN is the jargon to watch here. In plain English, it’s the idea that the “brains” of a cell site (the baseband processing that makes radio access work) can run as software on general-purpose servers, instead of being locked into custom telecom boxes.
The single-server angle is what Samsung is really trying to prove. If more RAN functions can run reliably on one beefier machine, the promise is less gear on site, fewer moving parts to maintain, and more headroom to run optimization software alongside the radio workload.
It’s also an architectural bet on where networks are heading: more automation, more real-time tuning, and more compute at the edge. AI inside a network doesn’t have to mean a chatbot — it often means models that predict congestion, optimize radio parameters, or squeeze more efficiency out of a limited slice of spectrum.
Computer Weekly framed the milestone around consolidation: fewer, more powerful servers running both RAN and AI workloads, with potential savings in space, power, and total cost of ownership. Analyst Daryl Schoolar called the demo “an important milestone,” while Samsung networks R&D head June Moon described it as a “major leap forward,” and Intel’s Cristina Rodriguez pointed to a “compute foundation for AI-native, future-ready networks.” (Computer Weekly)
One detail worth noting: the operator wasn’t publicly named. That’s common in early-stage network validations, but it also means outsiders can’t yet judge how representative the deployment is across different spectrum bands, traffic loads, or existing vendor stacks.
There’s a broader platform shift buried in all of this. If RAN software keeps moving onto standardized compute, the battleground expands from traditional telecom hardware into cloud platforms, silicon acceleration, and the integration work needed to make “carrier-grade” reliability feel routine.
Samsung also pushed the milestone through its own press resources, publishing a media library entry dated January 14, 2026 that repeats the core claim about the commercial call using Samsung vRAN with Intel Xeon 6700P‑B on a Tier 1 U.S. live network. (Samsung Global Newsroom)
The risk is that “single-server vRAN” sounds cleaner on paper than it can be in the field. Real networks have messy RF environments, upgrades layered on upgrades, and operational teams that won’t accept even small stability regressions just to chase efficiency. And even if the tech works, adoption depends on whether carriers believe the economics beat sticking with more traditional, purpose-built RAN hardware.
Still, this is the direction the industry keeps nudging toward: software-defined networks, more open architectures, and more compute where the radio meets the cloud. If Samsung can keep showing live-network proof points — not just lab demos — single-server vRAN starts to look less like a slide deck and more like an upgrade path.
