By Kevork Kechichian, EVP and GM of Intel’s Data Center Group

With 6G on the horizon, operators are clear that success won’t come from architectural resets, but from evolving the strong compute foundations already built in 5G. Progress will come from deploying intelligence responsibly and at scale across existing infrastructure — not by adding complexity, but by strengthening what’s already working.

What we’re hearing from operators

• Inference should be inherent. There’s a need for AI built directly into the network — not added through new accelerators or disruptive architectural changes.
• Efficiency is imperative. To free resources for new revenue-generating services, decisions are driven by power savings, infrastructure consolidation, and lowering TCO — all while keeping pace with evolving end-user demands and fast-shifting usage patterns.
• Openness builds trust. Operators want a secure, stable, production-grade platform — one that’s open and proven in commercial networks, and offers a low-risk evolution path toward 6G.

Intel and its partners are bringing more AI compute to the RAN and Core right now with products like the Intel Xeon 6 with E-cores, Xeon 6 SoC, and Intel Ethernet 800 and 600 series. Intel’s approach is straightforward: deliver an open, secure compute foundation that can run critical workloads — network functions, security, enterprise services, and AI inference — on one platform. With a consistent foundation, operators can modernize each generation without rip-and-replace, turning infrastructure into a lever for faster services and better economics. These benefits extend to consumers as well, enabling more reliable connectivity, more personalized experiences, and greater cost-efficiency.

AI in networks: the right compute for the workload

It’s tempting to reduce this discussion to a binary debate: CPU versus GPU. But that’s not how infrastructure evolves — and it’s not how operators build networks.

Different AI workloads will use different forms of compute, and the most effective strategy is to match each workload with the architecture that delivers the best mix of performance, efficiency, cost, and ease of deployment. Intel Xeon 6 with E-cores and Intel Xeon 6 SoC can expand network capacity, enhance productivity, and AI capabilities in the Core and RAN respectively — all while maintaining openness and operator control.

What doesn’t scale is applying a GPU-first worldview indiscriminately to inference-heavy network workloads. That approach can increase cost and complexity, introduce new operational silos, and force architectural changes that aren’t justified by the workloads themselves.

In networks, the question isn’t “Can we run AI?” It’s “Can we run AI without re-architecting everything we already operate – and what impact will it have on our cost and power budgets now and for the foreseeable future?” AI will be everywhere in the network. But if we want it to scale, we have to anchor the conversation in reality.

Read more here.

And please visit the following resources to learn more about how Intel Xeon 6 with E-cores and Intel Xeon 6 SoC are driving momentum across 5G Core and the RAN respectively:

• Intel Xeon 6 SoC: Industry Partners Embrace Platform, Accelerate AI in the RAN
• Intel Xeon 6 with E-cores: Mobile Operators, Ecosystem Partners Show 5G Core Momentum