Ericsson Conducts First Live 6G Trial in Texas

The trial's use of centimeter-wave (cmWave) spectrum in the 7-15 GHz range is significant; this band is considered a "sweet spot" for 6G, offering a crucial balance between the wide-area coverage of traditional mid-band and the high-capacity of millimeter-wave. This approach aims to reuse much of the existing 5G cell site grid, controlling deployment costs while delivering multi-gigabit speeds. A core component of the trial is its cloud-native infrastructure, reflecting a fundamental shift in network design for 6G. Unlike the monolithic systems of the past, 6G will extend the 5G core's service-based architecture (SBA) to the Radio Access Network (RAN), disaggregating hardware and software to enable more flexible, scalable, and automated network operations. This demonstration of AI and robotics workloads points to a key business model evolution for telecom: offering "AI as a Service" (AIaaS). The 6G network is being designed as an "AI-native" platform, capable of not just transporting data but also offering distributed computation, model training, and inference services at the edge, creating new revenue streams beyond simple connectivity. The trial leverages Ericsson's own custom silicon, a strategic choice in a competitive landscape. While rival Nokia is partnering with NVIDIA to utilize general-purpose GPUs for a flexible AI-RAN compute fabric, Ericsson is integrating its own energy-efficient ASICs with built-in neural network accelerators directly into its radios to optimize for Layer 1 processing and AI-driven beamforming. This represents a key architectural bet on whether specialized or general-purpose hardware will dominate the 6G era. The demonstration's focus on uplink-heavy applications like robotics and video streaming is intentional. Future use cases, from industrial automation to immersive extended reality, are expected to generate massive amounts of uplink data, shifting network design priorities. Ericsson and its partner KT have specifically trialed mid-band FDD Massive MIMO to boost uplink capacity and indoor penetration where AI workloads are expected to intensify. This move towards AI-native networks is forcing a re-evaluation of data center and edge infrastructure. The disaggregation of the RAN and the need for low-latency AI inference will push more compute resources toward the network edge. This creates a new dynamic between telecom operators and hyperscalers, who are both positioning to provide the distributed cloud infrastructure necessary to run these next-generation applications. Ericsson's collaboration with large language model specialists like Mistral AI is crucial for developing the AI agents needed for network automation and operations. These partnerships aim to co-develop AI models tailored for telecom, addressing complex workflows from legacy code translation to assisting in 6G research and development.