HEMS and virtual power plants at German utilities: build, partner or buy

Who owns the customer interface in the home?

Whoever runs the HEMS holds three things: the household's consumption and generation data, access to the controllable devices, and the revenue their flexibility will earn. The app on top is the smallest part of it.

The German utility trade paper ZfK put it bluntly in May 2026. For municipal utilities, the question is who owns the HEMS. Nicolai Kretz of evm, a regional utility around Koblenz, called it the heart of a new kind of systems sales. And the hurdle mentioned most often in the sector is not technical. It is getting retail, grid and IT to agree.

A HEMS on its own does not pay. That is in the same ZfK piece, and the vendors' own offers agree: HEMS, dynamic tariff, procurement optimisation and marketing as one bundle. So whoever decides on the HEMS decides on the virtual power plant as well.

What the market for white-label apps looks like, who integrates which devices and why Sections 41a and 14a of the German Energy Industry Act drive it, we covered in our piece on prosumer platforms and white-label apps . This one picks up the question after that. Which operating model fits a multi-utility, and how do you tell before you sign?

Why the decision is due now

Four developments meet in autumn 2026. None forces a utility into a HEMS. Together, though, they shift who will control the devices in its customers' basements two years from now.

The third one is underrated. If mandatory direct marketing for small PV arrives as drafted, every new rooftop system needs a direct marketer. Whoever runs the HEMS in that house is first in line with the customer. Everyone else sells the leftovers.

And module 3? It pushes from the other side. A time-variable grid fee is only more than a price sheet if something in the house reacts to it.

Three routes: build, partner, buy

In practice utilities choose between three operating models, and most end up with a mix. The examples below are publicly documented. They illustrate models, not a ranking.

Build

Your own platform means full data control and full responsibility. That covers integrating every inverter, every battery and every wallbox, the control logic, round-the-clock operation and market access. How big the device question alone is shows in a vendor statement from Heartbeat AI in April 2026: by its own account, its platform covers about 60 percent of PV manufacturers, half of the batteries, almost all heat pumps and 70 percent of EVs. That is where a company stands after years of building nothing else.

For a single utility, that is a software house of its own. For a group of utilities, maybe not.

Partner

Here the utility and its partner split the roles by contract. Two examples from this year. On 5 June 2026, enercity, the Hanover utility, and Kraken announced they would jointly develop a virtual power plant: enercity brings its power and heat generation portfolio, Kraken the technology, experience with aggregated assets and access to European markets. On 1 June 2026, Stadtwerke Tübingen started a pilot with Zählerfreunde and the Trianel Digital Lab that tests HEMS, PV, heat pump, e-mobility and dynamic tariff as one system and shares the results across the Trianel network of municipal utilities.

What separates this from buying is the contract, not the software.

Buy

Buying means the utility uses a finished platform, usually under its own brand. Heartbeat AI, part of 1Komma5°, has offered municipal utilities a platform with HEMS, Section 14a control, dynamic tariff and direct marketing since April 2026, at around 15 euros of software fee per household and month according to ZfK. gridX and greenblocks rolled out the HEMS liya with evm and say a utility can launch it in twelve weeks without building its own software.

The table will not tell you which route is right. Criteria will.

Five criteria for the decision matrix

Five questions separate the models cleanly. Price and feature list matter too, but they change with every offer. These five do not.

Data control

Where do the quarter-hour values, device states and forecasts live, and who may use them for what? If you build, the answer is simple. If you buy, it is in the contract, and only there. So ask for the right to export all data at any time, completely and in a documented format, and ask what the vendor itself may do with it.

Lock-in and exit

This is the one most often forgotten. What happens to 3,000 controlled households when the contract ends after five years? Do the devices keep working if the vendor's cloud stops answering? The installer PEAK.Energy asked this in public in May 2026: what still works sensibly if you no longer want the service or the tariff? It belongs in every tender.

Interoperability

A HEMS talks to devices through manufacturer interfaces, cloud APIs or open standards such as EEBUS under IEC 63380 . It talks to the grid through the smart meter gateway, the CLS channel and the control box. The more of that runs on open standards, the easier a later switch. The more runs through a vendor's cloud, the faster you get started today. Both are legitimate. Just know which price you pay for what.

Unbundling

In a multi-utility, three roles meet at the same meter cabinet. That deserves its own section, see below.

Rollout dependency

On 30 June 2026 the Bundesnetzagentur counted 3.17 million smart metering systems across 54.1 million metering points, which is 5.9 percent. 434 metering point operators are under supervisory proceedings. Many flexibility products need a smart metering system all the same; Tibber's Grid Rewards says so explicitly. A business model built on your own rollout grows exactly as fast as that rollout. No faster.

Unbundling: grid, metering and retail on one device

In a customer's basement, three roles meet that often sit under one roof in a German multi-utility. The grid operator sends the grid-serving control command under Section 14a. The default metering point operator runs the gateway the command travels through. Retail sells the dynamic tariff and the HEMS that distributes the command to the devices in the house.

Technically, that is one chain. Legally, they are separate roles. Section 6a of the Energy Industry Act (EnWG) requires commercially sensitive information from grid operation to stay confidential and to be disclosed only in a non-discriminatory way. Section 3(4) of the Metering Point Operation Act (MsbG) requires the metering point operator to act transparently and without discrimination and refers to accounting unbundling under the EnWG.

What does that mean for the HEMS decision? Something uncomfortable first. What your own grid knows about congestion in the low-voltage network, and what your own metering business sees at a metering point, is not a head start your own retail arm may use exclusively. A retail HEMS product has to work with the same signals a third-party provider would get.

The same separation applies once a partner or platform vendor comes in. The contract should spell out which data flows in which role. Whether and how far unbundling obligations apply to a particular company depends on its size and structure, and has to be settled in-house.

From leading a Section 14a programme at a multi-utility in northern Germany, we take one sequence away: first the control chain between grid and metering has to work, through the Section 14a control box and grid control and clean CLS management . Then the product. The other way round gets expensive.

The argument over the operating system in the basement

Two positions on the platform question met in public this year. Both are worth reading before a utility signs anything.

Heartbeat AI states its ambition openly. In the ZfK interview of 13 April 2026 the company said it did not want to keep the technology to itself but to make Heartbeat AI the standard operating system for every household. In May, 1Komma5° reported more than 100,000 controlled systems and named 20 gigawatts of controllable capacity by 2030 as its goal.

PEAK.Energy, an installation company, pushed back on 26 May 2026. Not against energy management, it says, but against unnecessary lock-in. Once control, visualisation and tariff are tightly coupled, switching gets hard. The actual intelligence runs as a service rather than locally, and owning the hardware does not mean controlling the system. PEAK.Energy therefore prefers manufacturer-independent systems with open interfaces.

Both have a point. A platform that controls a very large number of devices without the customer lifting a finger is a real lead that a single utility will struggle to close. And a platform whose exit nobody has settled is a risk that only shows up five years later. You do not resolve that tension with a verdict on a vendor. It goes into the contract.

Challenges and risks

The biggest uncertainty is regulation that has not been passed. The 2027 EEG amendment is a government draft. Whether mandatory direct marketing for small PV arrives as described is up to the Bundestag and the European Commission. A business model resting on that alone rests on a draft.

• The economics come late. In the ZfK round in May, utilities expected two to three years before the models pay off, depending on how fast the rollout goes.
• The rollout stands at 5.9 percent. Any product that needs a smart metering system reaches only a small share of customers today.
• Device variety never goes away. No vendor integrates everything, and every new inverter generation starts over.
• Inside the house. Retail, grid, metering and IT need to agree on a target picture before anyone picks a vendor.

That last risk is the most common one. And the only one no vendor can fix.

What utilities should decide now

Decide on roles and contracts, not on apps. A sensible order looks like this.

Five steps to the decision

1. Agree a target picture across all roles

   Settle which role the utility wants at the customer interface: platform operator, retail brand on someone else's platform, or partner in a group. Retail, grid, metering and IT sign off on it together, otherwise procurement ends up deciding.

2. Control chain first

   Without a working chain of gateway, CLS channel and control box, any HEMS stays a tariff calculator.

3. Settle the exit before the entry

   Write into the requirements what happens to data, device integrations and controlled households when the contract ends. Ask about local operation if the cloud fails. And about the export format.

4. Document data flows between roles

   Record which signals from grid and metering the HEMS product uses, and whether a third-party provider would get them the same way.

5. Measure with a pilot

   A pilot with a few hundred households, clear metrics and an end date tells you more than any presentation. Measure device coverage, controllability, customer retention and the effort in your own service team. The Tübingen approach of sharing results across a group lowers the cost of that insight for everyone involved.

Once these five are answered, the choice between build, partner and buy often turns out not to be that hard. The hard part was the question before it.

Further reading

Frequently asked questions