Nobody Has Left a Robot Working Overnight
Yulia Sandamirskaya · ZHAW Zurich University of Applied Sciences
EP. 12
AI Data Centres in Space: A Physicist’s Reality Check | Bilge Demirköz, METU
Physicist Bilge Demirköz examines AI data centres in space, cooling, orbital debris, chatbots and scientific discovery in this TechDrive Zürich interview.
Discuss this episode and meet other Swiss tech builders in the TechDrive WhatsApp group.
Key insights: executive summary
- More electricity does not solve the cooling problem. Bilge challenges the assumption that putting AI infrastructure in space makes waste heat easy to remove.
- Satellite fleets share an environment. She argues that their assessment must consider collision risks, debris and cumulative environmental effects alongside individual business plans.
- AI needs the care we associate with fire. Her analogy recognises its usefulness while questioning how quickly we put powerful tools into children’s hands.
- A local chatbot still needs checking. Her lab uses a local Qwen model adapted to its documents, but she acknowledges that it still produces errors.
- Scientific AI is a source of real enthusiasm. She describes analysing space-station data, improving simulations and an unexpected filter design that she says worked when built.
- Adding AI should solve a real problem. If an efficient existing method already meets the need, she asks what replacing it with AI actually improves.
Moving computation into space does not remove the need to manage heat.
The strongest case for AI starts with a useful result that someone can verify.
Why are AI data centres in space still a cooling challenge?
Space offers opportunities for solar power, but computing equipment still produces waste heat. Bilge argues that the discussion around AI infrastructure gives too much attention to electricity supply and too little to how that heat will be removed.
Her challenge is to examine the complete system: power generation, radiators, orientation and the surrounding environment. The conversation questions the idea of effortless cooling in space; it does not establish that orbital data centres are impossible.
What happens when many companies scale their satellite plans at once?
Bilge’s concern extends beyond whether one spacecraft works. Collisions can create debris that threatens other spacecraft, increasing the risk of further collisions, the process associated with Kessler syndrome.
She also raises questions about launches, re-entry and the cumulative effects of material returning through the atmosphere. Her broader point is that commercial plans need to be assessed together, because their consequences are shared. The discussion presents collision cascades as a risk, not an inevitable event with a fixed timetable.
Why does she compare AI to fire?
Fire can be extraordinarily useful while requiring judgement about how and where it is used. Bilge sees a similar tension in AI, particularly when children use conversational systems before developing the knowledge needed to question their answers.
As an educator, she worries about tools that replace the effort of learning or repeatedly reinforce what a user wants to hear. She calls for evidence, education and responsibility alongside technical progress. Her views in this interview are her own; membership of the UN scientific panel does not make them an official UN position.
Where does she see AI delivering real scientific value?
Bilge describes using AI to analyse International Space Station data in the search for dark matter and to improve scientific simulations. She also recounts an AI-assisted band-pass filter design that looked unfamiliar but, in her account, worked when built.
Her lab’s local Qwen model offers another example, with an important qualification: the team still catches hallucinations. Knowing the underlying documents helps researchers evaluate its answers.
The common thread is verification. An unexpected output becomes useful when researchers can test it against evidence or a physical result.
Could some space applications be explored closer to Earth?
Bilge discusses stratospheric balloons and concepts combining balloons with wings as research directions worth exploring. She describes student balloon projects and using AI to analyse their data.
These are possibilities to investigate for suitable applications, rather than a proven replacement for every satellite mission. Her wider question is whether teams have examined the available alternatives before committing to a particular technology.
When does a problem not need AI?
Bilge describes reviewing proposals that introduce AI to problems with working solutions already in place. Her challenge is straightforward: if an existing approach is efficient and meets the need, what does AI improve?
For builders, that means comparing performance, reliability and resource use with the existing method. Novelty alone is a weak reason to replace something that works.
Who is Bilge Demirköz?
Bilge Demirköz is a Professor of Physics at Middle East Technical University (METU) and a member of the UN Independent International Scientific Panel on Artificial Intelligence. In this conversation, she brings together space research, AI-assisted science, education and environmental concerns. She also shares a more personal perspective: physics and mathematics stretch the mind, music stretches the heart, and sport stretches the body.
Episode show notes
Bilge Demirköz joins Justinas Mišeikis for a drive through Zürich and a conversation about the physical and human limits behind ambitious AI plans.
Starting with her studies in physics, mathematics and music, the discussion moves into cooling, crowded orbits and environmental responsibility. It then turns to children’s use of chatbots, her lab’s carefully checked local model and the scientific applications that make her optimistic about AI.
The question running through the episode: what does this technology improve, and how do we know?
More conversations from the Swiss tech ecosystem
Physical AI Is Ready, Buildings Are Not
Kateryna Portmann · ANYbotics
Switzerland’s AI Movement Is About to Start
Katka Letzing and Anna Hakobyan · Swiss {ai} Weeks
The next drive, in your inbox.
A short note when a new episode drops: the guest, the hook, and the one question you won't want to miss. No spam.
Follow the Swiss tech ecosystem, one drive at a time
New in-car interviews with founders, investors and researchers building Zurich's deep tech, AI and robotics scene. Want to discuss this episode, ask follow-up questions and meet other builders?