At the 2026 Mobile World Congress (MWC) in Barcelona, technology providers made clear their commitment to supporting the shift toward AI-enabled telecom infrastructure. Some infrastructure companies are opting for distributed GPUs, while Intel is focusing on CPUs with integrated AI acceleration.

Intel’s main announcement was the Xeon 6 platform, a processor architecture designed to handle radio access network (RAN) workloads, core network functions, and AI inference on a single chip.

Intel executives say that building AI directly into the CPU, instead of using external accelerators, is key to their strategy. They argue this is a more efficient and cost-effective approach for telecom operators who must manage costs and energy use, directly addressing the industry’s main pain points.

Intel’s telecom hardware strategy improves processor performance within existing power limits. The company offers several Xeon 6 processor versions, including an SoC for RAN and efficient core options for the 5G core network.

On Intel’s A18 platform, Xeon 6+ scales up to 288 cores and improves performance-per-watt by 60% over previous models.

In an interview with EE Times at MWC, Cristina Rodríguez, VP and general manager in Intel’s Network and Edge Group, stated that the company has focused strictly on network requirements. “We have been very intentional year over year on what we are doing with our road map and how we approach the requirements of the network,” Rodríguez said.

The newest Xeon 6 SoC, called XCC, can scale to 72 cores, reducing the amount of hardware needed at cell sites. Rodríguez noted that in the past, some operators had to use two servers.

“With 72 cores, we eliminate one server,” she said. “Now we have 50% of everything right where cost, power, complexity, and supply chain matter. We moved from two servers to one.”

This consolidation means the whole wireless stack, from the first layer to the core and the edge, can run on a single software-programmable platform.

Integrated CPUs versus distributed GPUs

Adding AI to networks has led to different hardware strategies among major telecom suppliers. At MWC, Nokia announced a major move toward artificial intelligence-radio access networks (AI-RAN) with Nvidia.

Nokia’s approach is to use a distributed Nvidia GPU system for fast machine-to-machine communication and AI workloads. They say this lets them separate hardware from software, making operators more like flexible cloud providers.

Intel presents its Xeon 6 platform as a direct alternative to GPU-focused strategies, forming the core of its argument. The company asserts that for telecom workloads, especially for inference, using GPUs adds unnecessary complexity and cost, whereas integrated CPUs with AI acceleration can meet these demands efficiently within telecom constraints.

Also, Intel uses advanced matrix extensions in Xeon 6 for AI inference. Rodríguez explained that cell towers don’t require the training capabilities of GPUs.

By running AI inference for radio algorithms and infrastructure directly on the CPU, operators don’t need to buy extra hardware. “All that inference can be done in our Xeon SoC, and what is good about that is that then you don’t need to have an external component,” Rodríguez stated. “You don’t need a GPU basically, which will consume more power, which will add cost to the solution complexity.”

Stakeholders and transition to 6G

The telecommunications industry sees today’s investments in software-defined 5G as the base for future 6G networks. To help with this transition, Intel announced it is extending its long-term partnership with Ericsson.

The two companies are working together to create high-performance, energy-efficient computing systems that combine programmable networks with real-time sensing and AI. At MWC, Ericsson showed its Cloud RAN solution running on Xeon 6 servers, processing real-time 8K immersive media and core network functions without needing extra hardware.

Additionally, Intel is working with major global cellular service providers (CSPs) to roll out the Xeon 6 architecture. AT&T, Intel, and Ericsson are showing an AI-based link adaptation, which has improved throughput by 20% compared to older rule-based systems.

“Together with AT&T and Intel, Ericsson is demonstrating how our domain expertise, combined with AI-native RAN software, can drive transformative advancements in both Cloud RAN and purpose-built deployments,” said Mårten Lerner, head of networks strategy at Ericsson. “This milestone underscores Ericsson’s commitment to helping operators advance their networks by deploying AI functionality across the RAN stack.”

Vodafone is using the Xeon 6 SoC for its virtualized RAN upgrades in Europe. Rakuten Mobile relies on the built-in acceleration for demanding, low-latency tasks. In the core network, SK Telecom and NTT DOCOMO use Xeon 6 processors with E-cores and Intel Ethernet 800 Series products.

CSPs power consumption dilemma

A key point in Intel’s message to operators is the need to manage energy use. As AI applications create more global data traffic, telecom providers are seeing higher operating costs, especially for electricity. Intel’s Xeon 6 processors with E-cores use Intel Infrastructure Power Management software to cut power consumption by 60%.

Furthermore, even as the need for computing power grows, operators are still limited by utility costs. Rodríguez said that network hardware efficiency must show up directly in financial results. “The intention of being conscious about the energy consumption still needs to be there, and the operators at the end of the day, the operators get an electric bill, and they need to make sure that the electric bill goes down,” she said.

Because of these economic factors, Intel questions whether it makes sense to use GPUs at edge locations like cell towers. “When I hear the thought of putting a GPU in a cell tower, that makes no sense,” Rodríguez asserted. “You can’t justify the increase in cost, power, and complexity without real use cases.”

By offering a common hardware foundation from the core to the edge, Intel aims to give telecom operators a reliable, cost-effective path for transitioning from today’s 5G networks to future 6G infrastructure.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved