Power measurement isn’t just a technical detail—it’s the foundation of every smart infrastructure decision we make.
After months of collaboration with Cisco engineering teams, I’m excited to share a three-part white paper series that tackles one of the most overlooked challenges in enterprise infrastructure: accurately measuring power consumption.
Why does this matter? Simple. You can’t optimize what you can’t measure accurately. And with AI data centers now costing $350 million annually in energy alone, precision isn’t optional—it’s survival.
Getting to the root of power measurement inaccuracy
Driven by the belief that the information and communication technology (ICT) industry deserves a common language for power assessment, we conducted thousands of hours of testing across diverse scenarios. Special recognition goes to the paper’s primary authors, who led the technical research and new methodology development: Beth Kochuparambil, Principal Engineer and Technical Lead; Joel Goergen, Cisco Fellow; and Anna Fessler-Hoffman, Sustainability Specialist.
Through this team’s research, we discovered that most ICT organizations measure power incorrectly. Traditional methods based on industry standards like the Alliance for Telecommunications Industry Solutions (ATIS)’ Telecommunications Energy Efficiency Ratio (TEER) miss the difference between “apparent” vs. “real” power. Without proper calibration, software readings can be wildly inaccurate, in some cases producing up to 50% error rates.
To solve this problem, we developed a standardized methodology that moves beyond TEER to capture traffic patterns, temperature fluctuations, load balancing dynamics, and other real-world variabilities. By understanding error patterns and implementing systematic corrections, we can now achieve +/- 2% accuracy in software readings. Compared to the typical +/- 30% accuracy produced by traditional methods, the results produced through the team’s new methodology represent a significant breakthrough.
AI creates a compelling case for better energy usage data
“Precision power measurement is fundamental to sound engineering decisions,” says Goergen. “When our teams can see real power consumption instead of guessing, they can optimize system design from the ground up. This level of measurement rigor needs to be embedded in every stage of our engineering process—from initial chip design through data center deployment. That’s how we build efficiency into the architecture itself, not retrofit it afterwards.”
This degree of improvement can deliver immediate impact across areas of business:
- Data center operators can improve capacity planning and infrastructure decisions.
- Operating costs can be reduced through power usage optimization.
- Environmental, social, and governance (ESG) and carbon footprint reporting can gain accuracy.
- AI infrastructure investments can be planned more strategically with reliable data.
“Accurate power measurement isn’t just about efficiency—it’s about enabling the next generation of AI infrastructure,” said Martin Lund, Executive Vice President of Cisco’s Common Hardware Group. “As we design silicon and systems that will power tomorrow’s data centers, having precise power telemetry at the hardware level is fundamental to delivering both performance and sustainability. This work provides the measurement foundation that allows our hardware innovations to operate at their full potential while meeting the stringent energy requirements that AI workloads demand.”
Preview the future of power telemetry
Ready to stop guessing about your infrastructure’s energy consumption? Dive deep into methodology, implementation, and the future of power telemetry through this groundbreaking white paper series:
What’s your biggest challenge with power measurement accuracy? Comment below and share your experience.
series to start optimizing with better power data.
