MWC 2026 delivered yet another compelling showcase of innovation across the telco industry, with AI taking center stage across the show floor. One of the questions I was asked most often was how does Samsung expect to use GPUs for AI, and is AI just hype? The answer was simple: AI does not equal GPU. It’s much more than that. It was clear at this year’s event that AI is not just hype — it’s a lasting shift, and it requires deeper discussions to realize its full potential over the network platform. 

One of the key points is to think about how to bring AI into existing telco networks. It’s not possible to replace the underlying infrastructure overnight, so it’s crucial to get prepared for the changes that will unfold over the coming decades. That’s why Samsung has already been moving in this direction, building on our leadership in network virtualization and end-to-end software-driven networks to help operators future-proof their infrastructure.

One of the consequential debates in network infrastructure today revolves around whether GPUs are a prerequisite for AI in the RAN. While GPUs serve as high-performance platforms for networks, AI is not synonymous with GPUs. Samsung’s approach embraces both CPUs and GPUs to run AI workloads — the optimal choice depends on an operator’s deployment economics, service requirements and use cases. Samsung’s commercial vRAN today runs on CPU infrastructure, enabling AI-driven enhancements without the need for a separate GPU. While GPU integration is also possible, it remains an option where the economics support it.

I believe that operators’ AI strategy shouldn’t be locked in by their hardware, but defined by the vision and business models that drive revenue. Samsung provides an adaptive, flexible, AI and software-driven architecture supporting diverse compute engines. This approach was highlighted at MWC through successful multi-cell tests across Intel, AMD and NVIDIA/Arm. The point is that operators aren’t locked into a single hardware trajectory. 

Earlier this year, Samsung completed the industry’s first commercial vRAN call using Intel Xeon 6 SoC on a Tier 1 U.S. operator’s live network, presenting a pathway for single-server vRAN deployments that can optimize TCO. Other collaborations include:

• Validated AI-RAN features to improve network performance using Intel Xeon 6 SoC (with up to 72 cores) 

• Demonstrated 5G Core with AI-driven autonomous network capabilities using live network data for time sensitive operations

Additionally, Samsung achieved a number of milestones with AMD recently, demonstrating progress from lab tests to field deployment:

• Collaborating on commercial vCore deployment in Canada, enabling more flexible network deployments

• Validated Samsung’s next-generation enterprise solution, Network in a Server (NIS) in Japan, simplifying enterprise deployment and operations

• Completed multi-cell tests using 100% software-based solution powered by an AMD CPU (Sorano), without any hardware accelerators, achieving commercial-grade vRAN performance 

Another standout theme at MWC 2026 was AI and automation. Automating each network function is not a significant challenge anymore. The more critical aspect is managing the full network lifecycle intelligently — planning, deployment, operation, troubleshooting, optimization and customer experience — with minimal human intervention. The ultimate goal is to move toward a truly autopilot network, where efficiency and intelligence are seamlessly integrated.

That is the exact scope of Samsung’s end-to-end network automation solution, CognitiV NOS (Network Operations Suite). At its center is the AI Agent Fabric, a decision intelligence layer that lets multiple AI agents plan, evaluate, and act autonomously using their own reasoning and coordinating with each other beyond operating as isolated point solutions. 

At MWC 2026, Samsung demonstrated multi-agent automation scenarios that drastically reduce manual intervention. By cooperating to detect, analyze and solve network issues autonomously, these AI agents address network anomalies more quickly, making real-world tasks easier. Already deployed and proven in Tier 1 operators’ networks globally, CognitiV NOS paves the way toward our goal of actualizing fully autonomous networks by 2027.

Building on our industry-leading virtualization experience for macro networks, Samsung is bringing this expertise and innovative approach to the enterprise. At the show, we introduced the Network in a Server (NIS) solution, which integrates all private network functions — mobile core, RAN, transport and even AI agents, applications — into a single commercial off-the-shelf (COTS) server from Supermicro with an AMD CPU and Wind River cloud platform. This architecture also allows for simple GPU integration upon customer’s choice. 

As AI processing runs locally within the same server, NIS reduces latency for use cases in manufacturing, logistics and public safety that need instantaneous service and can’t tolerate a round-trip to a remote data center. The entire deployment fits in a single rack unit and can be placed directly at the point of use.

At the show, Samsung also introduced a key NIS use case, validated with a Japanese operator at a convenience store: AI video analytics and intelligent robot services, all running over a private 5G network. This use case isn’t just a demo scenario — it’s a real-world deployment model showing how a single, unified infrastructure can power distributed environments where each location needs edge AI and local resilience. Samsung is the only major vendor offering both traditional hardware-based and software/AI-based private network solutions, giving enterprises more flexibility.

MWC 2026 served as another great opportunity to see where the telco industry is heading. Samsung’s software-driven strategy, based on RAN disaggregation, AI-native/resilience by design, is delivering commercial benefits and tangible value to operators in real-world deployments. 

Moving forward, we are committed to two critical missions: empowering operators to activate new technologies without the need for installing new equipment, and more importantly, not just keeping pace with the future, but shaping it.