When Waste Becomes Wealth: Using Data Centre Heat.
This edition covers the conversation with Mark Bjornsgaard from Deep Green at Data Centre World London. We explore a fascinating—and largely untapped—approach to data centre sustainability that could turn one of computing’s biggest “waste” product into valuable energy.
“97% of the electrons that go into a server come out as heat”.
The Staggering Reality of Computing Energy
Here’s a statistic that might surprise you: “97% of the electrons that go into a server (computer in a data centre) come out as heat”. This revelation from Mark at Deep Green fundamentally changes how we might think about data centres. Rather than seeing them as energy-hungry infrastructure, we could view them as what Mark calls “vehicles through which electrons first pass before you then get to use the energy twice”.
In our climate emergency, this dual-use approach isn’t just clever—it could be essential.
The AI Heat Wave Creates New Possibilities
With AI and GPU workloads driving exponential growth in data centre power densities, the amount of heat generated is becoming extraordinary. But rather than seeing this purely as a problem, Deep Green has flipped the perspective entirely. As Mark explains: “if you get to the point where you’re capturing 97% of the energy which we are then you start to actually see the problem through the other lens of a telescope which is actually you’re just kind of politicizing an electric heater”.
This reframing could be profound. If we conceived computers as electric heaters that happen to do useful computation, the sustainability equation might change dramatically.
The Economics Could Stack Up
What makes Deep Green’s approach particularly intriguing isn’t just the potential environmental benefit—it’s the economics. Mark is refreshingly direct about this: “we’re unfathomably cheap if you do this you just make yourself richer”. The company claims to achieve cooling costs that are 20% lower than traditional data centres whilst providing free heat to industrial users and district heating systems.
This aligns with a principle many sustainability advocates understand: “if you’re doing sustainability right it costs you less money”. But the question remains: why isn’t everyone doing this?
The Scale of Theoretical Opportunity
The potential impact could extend far beyond individual data centres—if the model proves scalable. Mark shared some striking projections:
- 30% of all industry could potentially use data centre heat
- The UK government targets 20% of home heating from district heating systems
- Deep Green believes 80% of the country could theoretically be heated through district heating
- Across Europe alone, this might represent a “3 and a half trillion retrofit issue challenge opportunity”
These are ambitious projections for what remains a largely experimental approach.
Market Reality Check
Despite common perceptions, Mark challenges the narrative around power scarcity: “there is plenty of power in the world there’s plenty of power in the UK”. In the US alone, there are “9 gigawatts of district heating systems alone which all need to be electrified”.
If accurate, this suggests the real challenge isn’t power availability—it could be efficiently capturing and redistributing the energy we’re already using.
Early Days, Big Dreams
Mark acknowledges the nascent nature of this shift: “we see a world where this becomes a very very important thing to do quite soon” as “governments get very energised about people wasting electrons”. But he’s realistic about the timeline and scale of change required.
The transition is only just beginning. Deep Green has expanded beyond AI workloads to serve “any racks that are above 20 kilowatts”, suggesting that whilst sustainability credentials help, superior economics might be the real driver of adoption—if the model proves itself.
Full Episode
Check out the short (under 10 minute) episode here (apologies for the audio and video gremlins on this one):
Embedded LinkedIn content: view it on the original article
Early Pioneers, Uncertain Future
Currently, very little waste heat reuse is actually happening in the data centre industry. Deep Green appears to be one of the early pioneers exploring this territory, alongside a handful of other innovators. The concept remains largely theoretical at scale, though the individual projects that do exist show promise.
Interestingly, we’ll be exploring another approach to this challenge in an upcoming conversation with David Kohnstamm from leafcloud, the Amsterdam-based sustainable cloud provider that’s working on using data centre heat to warm residential buildings and provide hot water. It will be fascinating to compare these different models and understand what barriers and opportunities exist across different markets. Keep an eye on the Architect Tomorrow YouTube channel for when that one comes out.
David K sharing a joke with Oliver on upcoming episode.
The Bottom Line
This conversation highlights a potentially shift in how we might think about digital infrastructure. Rather than accepting waste heat as an inevitable byproduct, we could design systems that turn computing’s energy consumption into a feature, not a bug. Of course, this doesn’t excuse wasteful practices.
As we navigate the intersection of AI’s growing power demands and climate commitments, solutions like Deep Green’s create interesting ecosystem regenerative architecture opportunities. Whether this vision becomes reality remains to be seen. Sadly note that this can’t be easily retrofitted to existing large data centres - unless they are in the right place - for example residential areas or next to industrial heat demand.
Want to learn more? Visit deepgreen.energy or catch the full conversation on the YouTube channel.
What are your thoughts on heat recovery from data centres? Do you see barriers to adoption that weren’t discussed? Feel free to share your thoughts in comments - thanks.
AI use disclosure: - Claude 4 Sonnet used to pull out and quote parts of the transcript to create a first draft - but it is based on a real life conversation that you can find on the YouTube Channel.
Originally published on LinkedIn. Comments and discussion live there.