This Network Energy Data Has Never Existed Before

Energy efficiency · Network infrastructure · Data strategy

Jon Baars  ·  April 2026

There is something at the center of what we do at ClaritechIO that is frequently misunderstood. The energy telemetry data we produce does not already exist somewhere.  It isn't waiting to be moved or cleaned. It isn't billing data from an energy company, and it is not legacy energy management output from a proprietary hardware or software system. We are creating standardized energy telemetry that has never existed before.

The Equipment Has Been There. The Data Hasn't.

In any data center or telecom site built in the last twenty years, equipment has been designed to report granular operational data: SNMP-capable generator or rectifier controllers, power distribution units with management interfaces, AC meters, environmental sensors, battery management systems, cooling systems with digital controls, and various proprietary and partial solutions like RTU and IoT units.

Most of this equipment has been capable of reporting valuable data all along — rack-level power draw, load percentages per circuit, supply air temperatures, rectifier loads or generator fuel levels. Hundreds of variables, minute by minute, if you use the right tools.

This Is Not a Data Pipeline Problem

Companies like Confluent, Snowflake, Cloudera and the broader modern data stack are excellent at what they do. They move data faster, store it more securely, unsilo it, and make it more accessible.  Every company is faced with the problems related to the existing data in their organizations.

ClaritechIO is solving a data creation problem. The granular, standardized, equipment-level, real-time energy telemetry we produce has no predecessor in operator environments. There is no historical version of it sitting in an aging SQL database. There is no slower batch process we are speeding up.  None of the existing proprietary hardware and software tools you have are creating it.  This data simply did not exist before modern data-streaming methodologies recently enabled it.

The tools that make our work possible are the same tools the modern data stack is built on, but the work is fundamentally different. These are complementary systems, not competing ones.   The data ClaritechIO produces can and should flow directly into the platforms the companies mentioned above provide, giving them a valuable new dataset while also giving users built-in infrastructure and access to AI tooling.   

What Gets Created, Specifically

For example, in a typical data center deployment, FOCUS ingests data from power distribution units to standardize rack-level power draw and load percentage across every circuit and every PDU vendor in the facility. It collects supply and return air temperatures, fan speeds, compressor utilization, and humidity from every CRAC/CRAH in the building. It captures UPS output load across phases, generator status and fuel levels, and facility-wide environmental sensing, standardized across multiple vendors to a common schema by equipment class, streaming continuously.

Data Value Compounds Over Time

A single cell site measured with FOCUS produces roughly 100 normalized energy metrics every minute. At 10,000 cell sites, a carrier would produce 1 million data points per minute — roughly 500 billion per year. Twenty larger telecom data center facilities producing 30 thousand data points per minute could contribute another 300 billion annually.

But the more important fact is not the volume.  The methodology described would work equally as well in a 100 site telecom network, or a single data center.   What is important is that these data points did not exist before. FOCUS doesn't look at existing data in a new way. It causes this data to exist.

A week of normalized telemetry lets you find outliers. A month lets you establish baselines. A year reveals seasonal patterns and validates whether efficiency interventions — like RRU power-saving features — are actually working. Multiple years enable predictive work: forecasting equipment behavior, anticipating failures, building the training data that AI optimization models will need to become genuinely useful.   You gain value very rapidly and the value grows over time.

An operator who begins building this dataset today will have years of normalized, structured, queryable history by the time AI-driven infrastructure optimization matures into standard practice. An operator who waits will be starting from scratch at exactly the moment when that history will be most valuable.

The AI Connection Is Direct, Not Aspirational

There is a lot of discussion in infrastructure circles about AI-driven operations — autonomous optimization, predictive maintenance, intelligent energy management. Most of it is framed as a future capability.

The limiting factor, in almost every case, is not the AI. It is the underlying data. Specifically, the absence of normalized, granular, time-series infrastructure data that AI systems need to do anything useful. Standardized telemetry is the prerequisite without which AI-driven energy optimization remains theoretical.

Where We Are

This is not theoretical.   The datasets we have built for customers span years in some cases and already contain billions of standardized energy records. The normalization approach works across heterogeneous equipment from different vendors, at different ages, using different protocols, across every site in your network.  

Energy costs are rising. Sustainability mandates are tightening. AI infrastructure is driving power demand to levels that make efficiency a strategic rather than operational concern. The competitive window in which not having a viable network energy data strategy is closing.