For years, the wireless industry has pushed toward faster speeds, denser deployments, and unprecedented connectivity. But as our appetite for data continues to surge, the pressure on global energy systems is becoming impossible to ignore. Today, sustainability isn’t a “nice to have,” it’s a defining challenge for the Information and Communications Technology (ICT) sector and a core design principle shaping 6G.

By 2040, industries worldwide aim to dramatically cut (or eliminate) their carbon emissions. Yet ICT’s footprint is on a trajectory to surpass aviation. The biggest driver? Radio access networks (RAN). Between power-hungry amplifiers, always-on base stations, and soaring data demands, RAN operations make up the majority of wireless emissions. In the United States alone, cellular base stations consume an estimated 21 million megawatt-hours annually, equivalent to powering nearly two million households. The message is clear: If we want a greener future, we need greener networks.

To its credit, 5G is the most energy-efficient wireless generation when measured per bit transmitted. But efficiency gains have been overwhelmed by exponential data growth. Video streaming, IoT expansion, cloud gaming, and AI-driven applications put constant strain on networks, keeping base stations running at high power even when traffic levels drop.

Operators have acted where they can by tuning sleep modes, improving cooling strategies, deploying smarter resource management, and introducing more efficient amplifiers. These steps help, and the OpEx savings are real. But because most of these improvements came after 5G was already deployed, as retrofits and incremental upgrades, the industry quickly hit the limits of “low-hanging fruit.”

With 6G, the approach must be different. Rather than designing the network first and adding sustainability later, energy efficiency has been a core architectural requirement from day one.

Unlike previous generations, 6G will emerge in the middle of a historic AI surge. Machine learning (ML) is proving uniquely suited to optimizing complex systems, and few systems are as complex as modern wireless networks.

AI can:

• Predict traffic patterns and dynamically scale network resources
• Manage sleep/wake cycles far more intelligently than static rules
• Adjust beamforming and scheduling based on real-time load
• Optimize massive MIMO configurations
• Identify inefficiencies invisible to traditional algorithms

In short, AI enables a shift from reactive energy management to predictive, contextual, system-level optimization.

This isn’t theoretical. Early research shows that ML-driven control loops can dramatically reduce power consumption without compromising performance—and in some cases, improve both simultaneously.

One of the institutions on the front lines of this shift is Yonsei University, where researchers are exploring how ML models can improve both downlink and uplink energy efficiency in realistic cellular network scenarios. Their work includes developing learning algorithms that adjust radio parameters based on real-time load and channel conditions, identifying when components can be safely powered down and modeling how network performance responds to dynamic ML-based control strategies.

These insights are shaping the early vision for 6G sustainability. But as promising as the models are, a fundamental challenge remains:

How do we accurately measure the energy impact of an ML model at the system level?

Traditional test methodologies weren’t built for ML-driven architectures. Training data changes outcomes. Model behavior changes with context. And the relationship between performance KPIs and actual power consumption is subtle, even nonlinear. To unlock the full potential of AI, researchers need ways to tightly correlate network behavior with real-time, high-fidelity power measurements.

To address this complexity, we have partnered with Yonsei University to build a testbed designed specifically for evaluating ML-driven sustainability gains in wireless networks.

This system brings together:

• Our synchronization expertise for precise timing across devices
• NI CompactRIO (cRIO) power monitoring hardware to capture real-time consumption at sub-millisecond resolutions
• USRP software-defined radios delivering flexible 5G/6G-ready PHY and MAC experimentation
• System-level KPI observation to measure throughput, latency, reliability, and spectral efficiency
• Test orchestration that correlates performance with power draw to quantify ML model effectiveness

Together, these capabilities allow researchers to understand what the real impact at the base-station level is when an ML model claims an energy efficiency gain. For 6G research, this kind of transparency is essential. Without it, operators may overestimate ML’s potential or implement models that save power in one area only to increase it elsewhere. With it, engineers can finally close the loop between design, test, verification, and optimization to push toward aggressive net zero goals with confidence.

While research is ongoing, early findings indicate that ML-driven optimization could reduce total base station power consumption by up to 33 percent. If scaled across the US, this reduction translates to the electricity usage of approximately one million households. Globally, the impact would be immense. A 33 percent reduction isn’t just a sustainability milestone, it’s a fundamentally different cost structure for operators. Lower OpEx means:

• Greater profitability
• More accessible rural coverage
• Smoother transitions between generations
• Flexibility to support new applications without new carbon burdens

For operators like AT&T who have publicly stated that 6G must deliver efficiency gains for the technology to be economically viable this shift is not optional. Sustainability and affordability are interconnected.

The industry can’t wait until 6G deployments are underway to tackle power consumption. Unlike 5G, the economics won’t work if operators retrofit sustainability later. This is why standards bodies, research institutions, and industry partners are aligning around a shared priority: build a sustainable foundation into the architecture itself.

Three principles are emerging as core to that vision:

1. Efficiency as a first-class KPI—Throughput and latency have traditionally dominated wireless design. 6G is adding energy per bit, and system-level energy efficiency, as equally important performance metrics.
2. Intelligence everywhere—From the RAN to the core to the edge, ML will be deeply embedded. And because AI itself also consumes power, efficient ML architectures must be part of the equation.
3. Holistic, system-level measurement—The industry must evaluate energy usage not in isolated components but across the entire network stack. This is where testbeds like ours and Yonsei’s become foundational.

The path to a greener wireless future won’t come from hardware alone. It will come from intelligent systems that learn, adapt, and optimize themselves. 6G is the first wireless generation where AI isn’t an enhancement, but a requirement. Sustainability isn’t the obstacle, it’s the opportunity.

By pairing advanced measurement systems with ML-driven algorithms, researchers are beginning to unlock breakthroughs that were impossible just a few years ago. Our collaboration with Yonsei University is one example of how academia and industry can accelerate progress together.

If the early results hold, the next generation of wireless networks won’t just be faster, more reliable, or more connected. They’ll be cleaner, smarter, and more aligned with our global climate commitments than any generation before. The work is far from finished. But for the first time, sustainability isn’t an afterthought—it’s the blueprint.

Read the full application note on energy-efficient 6G network design.