Germany sits last in Europe on smart meters as grid flexibility demands grow

The figure sits at the bottom of a 25-row table. Denmark and Sweden are at 100 percent, Estonia and Italy at 99.9 percent, and the EU average at 65.6 percent. The energy supplier 1Komma5° published the overview, and t3n reported it on 3 November.

Romania, Greece and Bulgaria are absent from the data. The gap is not for lack of a legal deadline. German households using more than 6,000 kilowatt hours a year, or running controllable loads, are supposed to get a smart meter. Metering operators were meant to reach 20 percent of that mandatory segment by the end of 2025. According to t3n, 434 of the roughly 800 operators recorded missed even that quota. In more than 40 grid areas not a single meter was installed. The Bundesnetzagentur now says it will pursue penalty payments against operators that missed their obligations.

The pattern inside the numbers is size.

The 19 largest operators each have more than 500,000 metering points. They are at just under 30 percent, past their target. The roughly 600 operators below 30,000 metering points average barely 19 percent. t3n attributes part of the national lag to earlier security rules. Meters had to be transported sealed, and German units must be able to actively control consumer loads. That adds engineering work.

Why the slow rollout matters beyond a compliance table is the thing smart meters are for. t3n notes they enable flexible tariffs and targeted switching of loads. That pays off where a household or business has large shiftable demand, such as a heat pump or an electric car. That is the same flexibility argument running through the rest of this year's grid reporting. The evidence for it is getting more specific.

Shifting compute is worth less than it sounds

An Ember analysis published in October 2026 and covered by pv magazine on 3 November puts numbers on how much relief data centres can offer grids. It looks at two mechanisms: moving computing workloads in time, and supplying power on site. The study covers nine markets in the US, Europe and Southeast Asia. Ireland, the UK and Germany are singled out as particularly exposed.

In Ireland, data centres already account for about 19 percent of peak demand. That is 1.1 GW of a 6 GW system peak in January 2025. Ember projects that share rising to 24 percent by 2030, with data centre demand at 1.6 GW against a 6.5 GW peak. The mismatch is timing. A large data centre needs 200 MW to 500 MW and can be built in 18 to 24 months. Conventional grid expansion takes four to 10 years, pv magazine reports.

Load shifting looks like the cheap answer until you look at the workload mix.

Cloud services, enterprise workloads and AI inference need low latency. They will be about 87 percent of global data centre capacity by 2030, per Ember. Ember assumes 35 percent flexibility for AI training, 5 percent for inference and none for traditional data centres. On those assumptions, roughly 6.5 percent of capacity could theoretically move. Average utilisation is near 48 percent, so the capacity actually available during grid stress falls to about 3 percent. In Ireland that means about 100 MW of peak reduction by 2030, or 1.5 percent. Other markets land between 0.1 and 0.6 percent.

Ember's alternative is bring-your-own-capacity. Keep the workload running and cover the electricity with on-site generation or storage. pv magazine reports this could deliver nearly five times more relief, with potential reductions of up to 2.7 GW in Texas. That is a grid-connection and interconnection story rather than a sensor story. But it sets the bar any flexibility product has to clear.

Location decides whether flexibility counts

Ireland's regulator has published its first national flexibility needs assessment. CleanTechnica's 29 October write-up teases out the detail. The Commission for Regulation of Utilities identifies about 234 MW and 58 GWh of potential upward flexibility requirements on distribution networks in 2028.

At the level of a single substation the numbers get concrete. The assessment models a requirement of 0.6 MW and 3 MWh at the Naas 38 kV/10 kV substation in 2028, concentrated in January. Reinforcement is planned by 2030, after which no corresponding requirement is identified. CleanTechnica is careful to call this a modelled network need, not a commitment to buy a service or install a battery.

A battery elsewhere in Ireland could provide useful balancing services, help integrate renewable generation and earn revenue in electricity markets while contributing nothing to the particular constraint at Naas.

That is the argument for metering and monitoring at the edge of the network. A national capacity estimate cannot tell you which transformer is short. CleanTechnica also cites UK Power Networks' March 2026 network assessment. It reports the utility stopped seeking flexibility tenders for certain single-transformer substations whose constraints typically lasted 24 hours. The market could not address them on technical and commercial grounds.

The other side of the ledger

Not every new load needs new wires. A study by clean energy think tank RMI was reported by Canary Media on 28 October. It found that in roughly half of US states there is enough spare winter grid capacity to add large numbers of heat pumps without triggering expensive buildouts. The grid peaks in summer, and heat pumps largely displace less efficient air conditioners. In the 24 states plus Washington, D.C. that RMI analysed, wider heat pump adoption could cut residential electricity rates by 13 percent on average from where they would otherwise be. That is about $200 a year per customer. Michigan could install 3.6 million home heat pumps and cut residential rates by as much as 27 percent.

Canary Media quotes Ryan Shea, a study co-author at RMI. He compares the grid to a toll bridge with too few cars: more users spread the fixed cost.

The caveat comes from Parth Vaishnav of the University of Michigan. He told Canary Media the work is sensible. But he noted that a lower price per kilowatt hour does not necessarily mean a lower total bill. The numbers are directional estimates, not forecasts.

On the supply side, the queue is the problem. Canary Media reported on 29 October that PJM Interconnection has agreed to start a stakeholder process on Surplus Interconnection Service. PJM is the largest US grid operator across 13 states. The service would let existing power plants add batteries behind existing connections. Jon Gordon of Advanced Energy United told Canary Media the Market Implementation Committee takes up next steps at its 4 November meeting. A filing to FERC is possible as early as April 2027. Gavin Ahern, co-founder of consultancy Surplus Interconnection, said his company has tracked 44 GW of historical SIS project requests as of mid-2026. About 80 percent of them are battery additions.

The stakes in PJM are visible in the prices.

Canary Media reported on 22 October that Google and Voltus are now targeting a gigawatt of virtual power plant capacity in the region. That is up tenfold from an initial 100 MW goal. Voltus CEO Dana Guernsey said the first 100 megawatts was always the first bite. PJM has seen the cost of securing peak capacity rise more than eightfold in a few years. It forecasts large loads adding 30 to 34 GW of demand by the early 2030s on a system that recently peaked at 162.6 GW.

What the sensor layer has to prove

None of this runs without measurement at the distribution level. Germany's 5.9 percent is the clearest illustration of the gap between a flexibility target and the hardware that would make it real. The Bundesnetzagentur's penalty payments will test whether enforcement moves the small operators that make up most of the shortfall.

The Ember and CRU numbers suggest the harder test is granularity. A national megawatt figure can be met by a battery in the wrong place. The Naas example asks for a requirement in one substation in one month. That is a data problem before it is a hardware problem. On current evidence, the metering layer in Europe's largest electricity market has a long way to go before it can answer that question.