Samsung Networks posts worst quarter in nine years as 5G sales slump

Samsung Networks generated only about 400 billion South Korean won ($300 million) in third-quarter sales, its worst quarter for at least nine years and down from roughly KRW600 billion ($450 million) a year earlier. The collapse lands seven years after Vodafone's then network strategy chief Yago Tenorio rated Samsung the strongest challenger to Ericsson and Nokia, a judgment that raised hopes of a genuine third force in the radio access network (RAN) market.

Those hopes have faded. In the UK, Vodafone had placed Samsung at the center of an open RAN plan meant to replace Huawei across 2,500 sites. After Vodafone's merger with Three last year, the project was scrapped in favor of a quick turn to Ericsson and Nokia.

Why did VodafoneThree abandon open RAN?

Regulatory deadlines forced the operator's hand. "The CMA [Competition and Markets Authority] have imposed very strict deadlines in terms of number of sites and in terms of spectrum solutions on those sites, by certain deadlines," explained Andrea Dona, VodafoneThree's chief technology officer, at a meeting with equity analysts last month.

"Unfortunately, the current-technology open RAN roadmap did not enable us to meet those deadlines. That's why we took the difficult decision to abandon open RAN for now in the UK network."

The implication that Samsung could not keep pace with a government-mandated, double-time rollout in a sizeable country clearly stung. An absence of other visible work compounded the damage.

How bad were the numbers?

The year-over-year decline was not as bad as a third, Light Reading understands. Because parent Samsung Electronics rounds figures to the nearest hundred billion won, the most favorable scenario puts the revenue drop at less than a fifth. That still looks steep against the competition:

• Ericsson's comparable business group grew third-quarter sales 9%.
• Nokia's radio networks unit increased revenues 10%.
• Samsung's MX/NW unit, spanning consumer gadgets and network products, stayed unprofitable, with an operating loss of KRW600 billion, narrowed slightly from KRW700 billion ($520 million).

Samsung blamed lower sales in North America and of network software. Its stated aim this quarter is to "secure profitability" through growth in North America, Japan and South Korea.

According to Omdia, a Light Reading sister company, Samsung held just a 5% share of the RAN market last year. Results for the first nine months of this year show a 10% year-over-year sales decline, indicating no material improvement in 2026.

Is this a Samsung problem or a market problem?

Largely the latter. After an early 5G splurge, operators have resisted further investment, and every vendor has felt it. Omdia put global RAN revenues across all vendors at about $45 billion in 2022. By 2025, they had fallen to $35 billion.

Samsung's reliance on a smaller number of deals in critical markets leaves it more exposed to the quarter-by-quarter budget decisions of a few big operators. "Telecom investment cycles naturally create fluctuations in quarterly performance," a Samsung Networks spokesperson said by email.

The parent company cushions the blow. At group level, Samsung reported sales of KRW195.3 trillion ($145.3 billion), a 127% year-over-year increase, and an operating profit of KRW107.4 trillion ($80 billion), nearly eight times the level a year earlier. The AI-driven memory chip boom filled the company's pockets. Samsung could massively outspend Ericsson on RAN R&D, run Networks at an enormous loss, and still report record profits.

"Samsung Networks is backed by the scale, diversified business portfolio, and financial strength of Samsung Electronics – which reported a record quarterly operating profit of approximately $80 billion in Q3 2026 from our semiconductor business," the spokesperson noted.

Can Samsung's Intel-based virtual RAN compete?

Samsung champions a virtual RAN built on Intel's general-purpose processors, while Ericsson sticks with its own custom silicon and Nokia leans on its Nvidia partnership. That positioning could appeal to telco executives worried about custom silicon longevity or the power appetite of Nvidia's GPUs.

Doubts once surrounded Intel CPUs matching purpose-built RAN silicon. The latest Xeon 6 generation, codenamed Granite Rapids, has significantly narrowed the gap, according to numerous telco executives. A recent white paper coauthored by Intel and Samsung says Xeon 6 delivers a "70% performance-per-watt gain" over its predecessor.

Intel's own troubles complicate the story. The company tried to sell its network and edge group (NEX), the unit behind the virtual RAN flavor of Xeon 6, until plans were scrapped last December; those parts were absorbed into larger Intel units. Intel posted a net loss of $11.4 billion for the first nine months, versus a small profit of $324 million a year before, despite sales growing 18% to $46.2 billion. Concern about an AI bubble sent its shares down 29% on October 30, the worst day in Intel's history, after disappointing guidance.

The supposed virtue of virtual RAN is decoupling hardware and software. Samsung could lift code written for Intel chips and move it to another CPU platform with minimal tweaks. It has recently been flaunting a partnership with AMD, which shares Intel's x86 architecture, and a shift to Arm-based CPUs also looks possible.

A startup called Kenyi Technologies, founded by former Qualcomm developer Vincent Loncke, is working on a hardware accelerator to handle tasks too power-hungry for CPUs, including forward error correction and fast Fourier transform. Unlike Intel's vRAN Boost, a vendor-agnostic product usable with Intel, AMD or Arm CPUs would address telcos' portability dilemma.

Samsung also argues CPUs can carry AI-RAN, the Nvidia-popularized concept of replacing conventional RAN algorithms with AI. Unless processing happens within a transmission time interval of about 500 microseconds, performance suffers, so AI models must stay lightweight and may not need a GPU. Xeon 6 includes Advanced Matrix Extensions (AMX) for AI workloads; the white paper says AMX lets "operators can begin deploying AI in the RAN today, without the costs, complexity, power consumption, or space requirements of discrete AI compute." Samsung has not ruled out GPUs in the RAN but has so far only experimented with them.

What happens next?

Samsung's outlook commentary made no mention of Europe or European telcos. Beyond losing VodafoneThree, it won work at other Vodafone businesses and was due to roll out infrastructure in Germany this year, though financials suggest it remains stuck at a relatively small number of sites. Vodafone, heavily reliant on Huawei and Ericsson across Europe, has shown little appetite for removing either.

The best outcome for Samsung would be sterner government measures forcing Huawei's eviction. Ericsson's former CEO Börje Ekholm estimated Huawei's European share at between a third and 40% as recently as June. Fear of an Ericsson-Nokia duopoly could then drive telcos toward Samsung.

Nearer term, Samsung hopes its KDDI partnership in Japan begins translating into revenue growth this quarter. In the US, Verizon remains heavily invested in Samsung's virtual RAN and is arguably its most important global customer. Tenorio, who once championed Samsung at Vodafone, now does the same as Verizon's chief technology officer, and his advocacy may prove decisive in convincing other telcos.