The grid is becoming telecom’s problem, and its opportunity

This Industry Viewpoint was authored by Stephen Douglas, Head of Market Strategy, Spirent, a Keysight company

Most of the energy conversation in our industry right now is about consumption. How much power AI data centers will pull, where the chips go, whether the models can run cheaper. Fair questions. But there’s a quieter shift underneath them that I think matters more for telecom operators, and it has nothing to do with how much energy they use. It’s about energy they could supply.

Start with the demand picture, because it sets up everything else. By 2030, data centers alone are projected to need an additional 1,000 TWh a year, somewhere around 5 to 9% of global electricity generation. Europe may have to find more than 80 TWh annually, roughly Belgium’s entire consumption added to the grid. Transport electrification piles on another 700-plus TWh. The grids carrying this load were not built for it, and in much of the world they’re aging. The question stopped being whether demand grows. It’s whether the grid can take it without breaking.

That pressure is pushing utilities toward generating, storing, and managing power closer to where it’s used instead of hauling it long distances from central plants. And here’s the part operators should pay attention to: building a decentralized energy system needs exactly the assets telecom already owns.

What operators are sitting on

Think about what a network actually is, physically. Tens of thousands of distributed sites. Backup batteries at most of them. Highly reliable communications between all of them. A growing layer of edge compute. That’s most of the ingredients for distributed energy resources, already deployed, already maintained, already connected.

The batteries are the obvious example, and the most striking. Much of that backup capacity sits idle up to 99% of the time, waiting for an outage that mostly doesn’t come. Some operators already hold gigawatt-hours of distributed storage across their footprint. One European operator spanning several countries has over 1 GWh, enough to power tens of thousands of homes for a day or keep a mid-sized colocation facility running. Right now, that capacity does nothing but wait.

Orchestrate it intelligently and the network starts to look like a virtual power plant. It can help with frequency regulation and grid balancing, shield the operator from price volatility, and open revenue from energy markets and arbitrage. This isn’t a thought experiment. In Finland, an operator has regulatory approval to use base station backup batteries in the national balancing market. The model exists. Others can follow it.

The hard parts are real

I am hesitant to say it is going to be easy because it won’t be. The transition from an existing backup battery (with no ability to sell excess) to being a participating player in the energy marketplace involves changing regulatory models that are state-specific. This process typically requires replacing old batteries with new lithium-ion systems that are designed to cycle frequently and have a full depth of discharge. Proving both the battery and inverter are capable of meeting all applicable grid safety standards such as anti-islanding, which requires that when the grid fails the system must disconnect within milliseconds; having millisecond response time for fast frequency service; having secure, low latency communication for real-time market participation; and ensuring there is sufficient reserve capacity at the site to provide power during the period the site is supplying electricity back onto the grid.

That last point is the one operators can’t compromise on, and it’s why the discipline around this matters as much as the ambition.

Where AI fits, and where it has to be watched

The challenge of coordinating millions of small power generation units (like residential solar) on the grid with renewable sources (such as wind and solar), cannot be managed manually. That’s why an Artificial Intelligence system will have to manage this process. The issue here is that the grid is critical infrastructure where failure is not acceptable. Therefore, if the confidence level of the artificial intelligence makes a decision based on incorrect information or “hallucinations” it may result in actual failures. Also, by incorporating an AI component to the grid, we are expanding our vulnerability to cyber-attacks since AI adds new vulnerabilities (for example, Data Poisoning), which were previously nonexistent.

The guardrails will have to be mandatory. Mandatory guardrails will require both operational constraints based on physical limits including voltage and protocol restrictions that restrict the AI from suggesting any actions which may compromise the integrity of the grid. In addition, human involvement for consequential events (i.e. Load Shedding), explanatory AI so humans can validate the rationale behind the suggestions made by the AI; and digital twins constantly tested to validate that current protective measures can disable the AI if needed.

That is the glue binding these three components together. Running an AI driven grid orchestration process and implementing a large-scale grid resiliency program using battery-based energy storage products are complex processes and the costs associated with errors made while executing them show up as power outages, not bug reports. They require a high degree of verification from initial design through real-time operation. This is why simulation, testing and ongoing assurance become required rather than optional.

We invest most of our resources in closing this gap between how something appears to operate during design versus how it actually operates once it is deployed. For operators who are considering entering the energy business, that gap between planned operations and the physical behavior of the actual equipment is where the risk resides and that is where it must be addressed.

The convergence of AI, telecom, and energy is already underway. The operators who treat their networks as energy assets, not just connectivity, have a real opening. The ones who get there will be the ones who proved it works before they switched it on.

If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!