Intel Empowers Network AI with Precise Computing Power, Confidently Moving Toward 6G

Author: Kevork Kechichian, Executive Vice President and General Manager of Data Center Group (DCG) at Intel

The 6G era is just around the corner. More and more operators understand that success doesn’t come from complete architectural overhauls, but from continuously strengthening the solid computing foundation established during the 5G era. True industry breakthroughs come from deploying intelligent capabilities on existing infrastructure in a cautious, scalable manner—avoiding blindly increasing system complexity, and instead continuously reinforcing proven, mature systems.

Based on this, we offer a new perspective for the 2026 Mobile World Congress. The next chapter of network evolution will be written by leaders who possess both deep industry experience and ecosystem collaboration capabilities. They are committed to making networks simpler, more reliable, and more efficient—ensuring AI inference can perform well in real-world scenarios, balancing performance, power consumption, and cost, while enabling smooth deployment and scalable applications.

Insights from Operators

Through conversations with operators, we have identified three core industry consensus points:

• AI inference should be embedded within the network: AI must be built directly into the network system, rather than achieved through adding accelerators or disruptive architecture modifications.

• Efficiency is an eternal theme: To free up resources and expand innovative revenue-generating services, decision-making should always focus on energy saving, infrastructure integration, and reducing total cost of ownership (TCO), while closely aligning with end-user needs and rapidly evolving usage scenarios.

• Openness builds trust: Operators urgently need a reliable, stable platform that meets commercial deployment requirements. This platform should have an open architecture, be thoroughly validated in live network environments, and provide a low-risk, smooth evolution path toward 6G.

Intel is working with industry partners to inject greater AI computing power into wireless access networks (RAN) and core networks through Intel Xeon Scalable processors, Xeon 6 system-on-chip (SoC), and Intel Ethernet 800 and 600 series products. Intel’s approach is straightforward: create an open, reliable computing foundation capable of supporting various critical workloads—including network functions, reliable computing, enterprise services, and AI inference. With this continuously evolving computing base, operators can upgrade each generation without complete rewiring, turning infrastructure into a strategic asset that accelerates business innovation and optimizes business models. Ultimately, these advantages will benefit end users by providing more reliable connections, personalized services, and higher cost efficiency.

Moving Beyond CPU and GPU Battles, Empowering Network AI with Precise Computing Power

The industry often simplifies this as a binary opposition between CPUs and GPUs. However, this does not align with the evolution of infrastructure nor reflect how operators build networks.

Different AI workloads require different types of computing. Therefore, the most effective strategy is to match each workload with an appropriate architecture that balances performance, efficiency, cost, and deployment convenience. Intel Xeon Scalable processors and Xeon 6 SoC can enhance capacity, operational efficiency, and AI capabilities for core networks and RAN, while maintaining open architectures and operator autonomy.

Applying GPU-first thinking blindly to inference-intensive network workloads cannot achieve scalable expansion. This approach not only increases costs and complexity but also creates operational silos and leads to architecture changes driven by tools rather than actual workload needs.

In networking, the question isn’t “Can we run AI?” but “Can we run AI without reconstructing all existing operations? And what impact will this have on current and future cost and power budgets?”

In RAN, AI’s key is matching the right compute power to different workloads, rather than using standalone accelerators universally. Intel Xeon 6 SoC integrates AI acceleration capabilities directly into the virtual RAN (vRAN) software stack via Intel Advanced Matrix Extensions (AMX) and Intel vRAN Boost, enabling most inference tasks to be completed locally on servers without additional dedicated AI hardware—reducing costs, power consumption, complexity, and space. This delivers tangible value: lower TCO, better utilization of existing infrastructure, and AI deployment in current real-time networks without large-scale architecture overhauls. For operators prioritizing AI deployment with efficiency and cost-effectiveness, Xeon 6 SoC offers more predictable AI performance, simplified operations, and easy scalability across thousands of base stations.

Below are some real-world examples of operators applying Intel Xeon 6 SoC. These cases demonstrate that, with Xeon 6 SoC, there’s no need to choose between CPU and GPU—just the right compute power for AI workloads.

• Lotte Mobile is collaborating with Intel to utilize the built-in AI acceleration of Intel Xeon 6 SoC for training, optimizing, and deploying advanced AI models designed for demanding RAN workloads, meeting ultra-low latency requirements.

• Vodafone has decided to adopt Intel Xeon 6 SoC to support its large-scale Open RAN and vRAN modernization efforts across Europe. This builds on Vodafone’s earlier deployment in the UK, where Intel Xeon processors supported their first commercial O-RAN rollout.

Partnering with Customers and Ecosystem to Build More Reliable, Energy-Efficient Networks

With deep experience in virtualized core networks and RAN, Intel now supports the vast majority of commercial 5G networks worldwide. Leveraging Xeon 6 processors, including products tailored for core, RAN, and edge computing scenarios, Intel continues to expand its leadership—delivering higher performance per watt, more comprehensive integration, and unified software platforms to help operators smoothly modernize their networks.

In the 5G core network space, Intel is increasing investment. Operators urgently need mature, reliable platforms to handle growing traffic, stricter energy controls, and rising operational costs. Against this backdrop, the Intel Xeon 6 energy-efficient processors are gaining attention and accelerating toward larger-scale deployment. This is backed by over a decade of collaboration with ecosystem partners: building 5G core and telecom cloud solutions based on Intel hardware and software, achieving significant energy savings while balancing performance, user experience, and deployment efficiency.

As 5G core demands grow, simply improving energy efficiency is no longer enough to meet increasing operational challenges. Intel Xeon 6 energy-efficient processors address key operator needs—delivering reliable, long-term flexible platforms that combine energy efficiency with built-in reliability. Their architecture, built with TDX for security, protects sensitive data; QAT accelerates data processing during transmission; and support for 5G core AI inference on existing infrastructure enables new capabilities without disruptive network reconstructions, allowing scalable expansion.

Over the past year, Intel Xeon 6 energy-efficient processors have been deployed in various real-world scenarios, such as:

• SK Telecom deploying Intel Xeon 6 energy-efficient processors and Intel Ethernet 800 series controllers in its live mobile core network.

• NTT DOCOMO planning to use Intel Xeon 6 energy-efficient processors and Intel Ethernet E830 adapters for next-generation mobile core deployments.

Looking ahead, network equipment vendors (NEPs) and service providers have validated in live networks that Intel’s energy-efficient architecture can meet current core network infrastructure needs—improving energy efficiency and deployment density while maintaining performance and reliability. As customer demands shift toward platforms with predictable performance, high reliability, and high scalability to further reduce TCO, Intel is advancing the Xeon 6 roadmap toward the next stage—Intel Xeon 6+.

Built on Intel’s 18A process technology, Xeon 6+ is designed for exceptional energy efficiency, helping operators expand workloads more efficiently, reduce power consumption, and support smarter network services. This platform also increases core density while further lowering power consumption, directly optimizing TCO. From 5G infrastructure to cloud-native applications, Intel Xeon processors are engineered to optimize performance, efficiency, and cost—further refining data center economics and ROI as we move toward 6G. At MWC, Intel previewed its next-generation products, with more details to be announced in the coming months.

Highlights

In networks, AI will be ubiquitous. To achieve scalable commercial deployment, we must rely on practical, deployable, and operable foundations:

• Prioritize inference: Deploy real-time decision-making capabilities at RAN, core, and edge.

• Be efficient and easy to operate: Design for telecom-grade demanding scenarios, not just laboratory performance.

• Be open and trustworthy: Provide interoperable open platforms that strengthen operators’ autonomy and industry robustness.

• Focus on results: Use measurable, tangible outcomes as benchmarks.

MWC is a key milestone for industry leadership and consensus-building. Our clear and pragmatic goal is to leverage open, reliable, and efficient computing foundations to help operators transform infrastructure into growth drivers, enabling continuous technological evolution. At the same time, we aim to enable AI deployment without high costs or extensive architecture overhauls—making AI truly practical and scalable.

This is a reliable evolution path from realizing 5G value to unlocking 6G potential.