As the 6G era draws ever closer, global telecom operators are facing a harsh reality: they need smarter networks, but lack the resources to rebuild from scratch. At this year's MWC 2026, Intel Executive Vice President and General Manager of the Data Center Group, Kevork Kechichian, pointed out that the next step in network infrastructure is not about blindly stacking expensive dedicated GPU accelerators, but rather about how to integrate AI inference capabilities naturally into the RAN (Radio Access Network) and core network based on existing 5G computing through "open platforms" and "optimized computing power allocation."
Debunking Myths: Internet AI is Not an Arms Race Between CPUs and GPUs
In today's era of generative AI, the industry often simplifies discussions about AI computing power into a "CPU vs. GPU binary opposition." However, this logic does not apply to the practical operation of telecommunications networks.
Kevork Kechichian stated bluntly that forcing the "GPU-first" concept into network workloads centered on "inference" would not only significantly increase hardware costs and power consumption, but also create an operational silo effect, forcing telecom operators to make unnecessary architectural changes.
The real key is to "match the most suitable computing resources to the workload".
In the RAN (Radio Access Module) field, Intel's answer with the Xeon 6 processor is to directly integrate AI acceleration capabilities into the server's underlying layer using built-in Intel Advanced Matrix Extensions (AMX) and Intel vRAN Boost technologies. This means that the server itself can handle the vast majority of AI inference tasks, and telecom operators do not need to allocate additional physical space or bear the power consumption and complexity of external accelerators, allowing them to easily expand AI capabilities across thousands of base stations.
Giant Companies Pay the Bill: Xeon 6's Real-World Performance in RAN and Core Networks
Intel's pragmatic AI networking logic has already received endorsement from several top global telecom operators at MWC 2026:
Breakthroughs in wireless access networks: AT&T, in collaboration with Intel and Ericsson, ran AI-native link adaptability technology on Cloud RAN powered by Xeon 6 processors, successfully boosting network throughput by 20%; Rakuten Mobile utilized its built-in AI acceleration capabilities to handle ultra-low latency RAN tasks; and European telecom giant Vodafone confirmed its large-scale adoption of Xeon 6 processors to deploy modern Open RAN and vRAN infrastructure.
A revolution in core network energy efficiency: Addressing the dual pressures of high data traffic and high electricity costs faced by 5G core networks, the Xeon 6 processor with E-core technology offers a solution. It not only provides extreme energy efficiency but also ensures zero-trust security for data transmission through TDX technology. Currently, companies including South Korea's SK Telecom and Japan's NTT DOCOMO have chosen to deploy it in their next-generation mobile core networks for commercial deployment.
A preview of the future: Xeon 6+ based on Intel's 18A process is coming soon.
In addition to showcasing its existing capabilities, Intel also previewed the next step in its roadmap at MWC 2026 – the Intel Xeon 6+.
This next-generation processor will be built using Intel's 18A process. Kevork Kechichian revealed that the Xeon 6+ will focus on achieving superior energy efficiency, further increasing core density while reducing power consumption, fundamentally improving the total cost of ownership (TCO) of data centers. It will shoulder the responsibility of significantly expanding AI workloads and supporting smarter network services, completely redefining the economics of the journey to 6G.
Analysis of viewpoints
Intel's message at MWC 2026 pinpointed a "collective pain point" for telecom operators worldwide: We need AI, but we don't have the money or the electricity to support a bunch of power-guzzling monsters.
In contrast to NVIDIA's absolute dominance in the cloud training market, Intel has chosen to maintain its massive market share advantage in "Network Infrastructure." By promoting the Xeon 6 series equipped with E-core processors, Intel is essentially telling telecom operators: "You don't need to replace your existing x86 servers; I'll directly integrate AI inference capabilities and hardware-level security (such as TDX) into the CPU for you to use."
This strategy of "prioritizing inference, being practice-oriented, and not disrupting existing architecture" not only effectively reduces telecom operators' hardware anxiety when upgrading to 6G, but also demonstrates Intel's strong ambition to defend against the GPU camp's invasion in the specific battlefields of "edge AI" and "telecom networks" with "high cost-effectiveness and high integration." The future details will depend on how far Intel's 18A process Xeon 6+ can push this energy efficiency advantage.
