I didn’t set out to build an energy company. I was looking at something else entirely — agent infrastructure, decentralised identity, the memory market problem — and kept running into the same question from a different angle: what happens when you treat information as infrastructure?
Not information as content. Not information as data. Information as the structural layer that makes a market legible.
The thing that nagged me was waste heat. Specifically, data centre waste heat in Europe. I’d been working on CarbonBench — tracking carbon intensity across AI inference providers — and the numbers kept pointing to a related problem. Data centres consume enormous amounts of electricity. Roughly 40% of that becomes waste heat. In a continent that imports natural gas to heat buildings, that seemed like a problem worth understanding.
So I started looking. And I found something surprising: nobody had mapped it.
There are directories of data centres. Cloudscene tracks thousands of facilities. Data Center Map plots them on a map. EUDCA publishes industry reports. But none of these include thermal profiles. None tell you the temperature of the waste heat, the volume, whether it’s continuous or intermittent, or whether the operator would connect to a district heating network.
You can find out where data centres are. You cannot find out what heat they produce. The first question has been answered many times. The second question has barely been asked.
Why the gap exists
Thermal data is operationally sensitive. Waste heat output correlates with IT load, which correlates with revenue. No standard format exists across borders — one operator measures in MW thermal, another in GWh/year, another in BTU. And until recently, operators had no incentive to disclose and no marketplace to sell into. Classic chicken-and-egg.
The result: a continent with thousands of data centres producing terawatt-hours of waste heat annually, and a separate continent with millions of buildings that need heat. The supply exists. The demand exists. The market doesn’t, because the information layer between them doesn’t.
What regulation changed
Germany’s Energieeffizienzgesetz — EnEfG — is a strange piece of legislation. On the surface, it’s compliance overhead: data centre operators above a certain threshold must file annual waste heat reports with BAFA, the federal energy efficiency authority. The reports include thermal output, temperature grade, availability, PUE, energy reuse factor.
What struck me is that this compliance document contains exactly the data a heat buyer needs to evaluate a connection. A district heating operator planning to route pipes needs to know: how much heat, at what temperature, with what reliability. The EnEfG filing answers all three.
The compliance cost is also a product listing. The operator doesn’t know this yet, because there’s no marketplace to list in.
France has a similar mechanism emerging — Décret 2025-1382. The Netherlands has RVO/EED implementation. Ireland has CRU rules. Each country is independently building the supply side of a European waste heat market. Different formats, different archives, different languages. None of them coordinating.
The EU is accidentally building a waste heat market, one national regulation at a time. The data is being created. It’s just not being aggregated.
The information infrastructure problem
This is where the agent infrastructure thinking actually helped. I’d spent months working on problems where the core challenge was the same: you have distributed data producers, each creating valuable information in isolation, and no aggregation layer making it searchable or tradeable. Memory markets for AI agents. Attestation registries. Reputation systems. Different domains, same structural problem.
Waste heat is the same pattern at physical scale. Hundreds of data centres across Europe, each producing thermal data through compliance filings, each filing going into a national archive that nobody searches. The data exists. The infrastructure to make it useful doesn’t.
So the question became: what if you built the aggregation layer? Not a technology company — the technology works. Not a consulting firm — the analysis is straightforward. An information infrastructure company. One searchable map of European waste heat supply, verified through compliance filings, that makes the market legible to buyers for the first time.
What I got wrong at the start
I assumed technology was the bottleneck. It’s not. Heat pumps work. Pipes work. The bottleneck is information.
I assumed operators wouldn’t care about waste heat. They do. They just had no marketplace to signal availability.
I assumed the Nordics had solved this. They’re ahead — Stockholm and Helsinki have high district heating connection rates — but even there, the matching between specific data centres and specific networks is manual and slow.
I assumed regulation was purely a burden. It’s actually the most efficient dataset-building mechanism available. Each filing creates a verified, searchable data point.
The thing I keep coming back to: this isn’t a software problem. It’s an information infrastructure problem. The value isn’t in the platform. It’s in the verified, searchable dataset of who produces what heat, at what temperature, and who needs it.
Where this sits now
Ardor maps 111 facilities across 10 European countries. Most profiles are estimated from public data today. Verified profiles grow with every compliance filing and every operator who claims their facility. The gap between estimated and verified is the gap between a directory and a marketplace.
The compliance business — generating EnEfG reports for German operators — is the customer acquisition funnel. Every filing adds a verified profile. The marketplace is the product.
Still figuring out a lot of this. Whether the municipal heat planning angle opens a separate market. Whether waste heat certificates become a tradeable instrument. Whether the EU Energy Efficiency Package in 2026 creates a cross-border standard that makes aggregation easier.
But the core observation hasn’t changed: the information infrastructure for this market doesn’t exist yet, regulation is creating the data to build it, and nobody is aggregating it into something searchable. That felt like a problem worth working on.