00:00Dr. Connor McCone is a lecturer in digital media at Stirling University in Scotland.
00:05Joining us now, Connor, it's fascinating hearing about underwater data centres.
00:09We also know there are plans for satellite space centres as well.
00:15How much innovation are we seeing in the sphere?
00:19Oh, tons. There's so much innovation happening right now in data centre technology
00:25and there's so much more left to go.
00:27One of the biggest enhancements that's happened recently,
00:30one of the biggest areas of research is actually in hot liquid cooling.
00:35We're not talking about very hot liquid cooling,
00:37but basically requiring liquid cooling that doesn't require the constant flowing of cool water
00:43is actually going to make things a lot more efficient.
00:44As you said, space centres, nuclear-powered centres,
00:50although those are a lot trickier and contingent upon heating,
00:54there's also a huge drive for solar-powered centres, just like in the Shanghai model.
01:01So there is an absolutely huge amount of innovation left to go right now.
01:06Something that's even more interesting than that, though,
01:09is that the data centre technology itself might not be the final frontier for this.
01:14If you look at things like what Michael Berry has been betting on recently,
01:17you're seeing a lot of shifts away from traditional data centres
01:21and perhaps towards more small-scale local data centres
01:25that could make things a lot more interesting very soon.
01:27So does that suggest that this massive power demand that we're currently seeing
01:33might go away or be managed somehow?
01:37It's actually more of a question of how long it will take rather than if.
01:43It's definitely going to become far more efficient.
01:47It's worth bearing in mind that the chips that are currently used to power AI
01:51are actually graphics processors most of the time.
01:55They're computer chips made for running computer games.
01:58They're not generally...
02:00There are some, but they're not generally chips
02:02specifically trained for training AI.
02:05NVIDIA are coming up with these.
02:07The Chinese government is investing heavily in research
02:10for specifically driven AI chips.
02:13But right now, we're kind of just...
02:15We've stumbled upon a happy accident with this technology.
02:18So we are going to become a lot more efficient very soon.
02:22The problem is, however long it takes to get to that more efficient paradigm,
02:27that's going to have real climate impacts.
02:30And it's really going to affect how much carbon we're using
02:33and the footprint of the planet within that time period.
02:36Yeah, we're supposed to be looking after our environment at the same time,
02:38and we're seeing these crushing heat waves.
02:40So there's a lot of problems coming at us.
02:43The US and China are leading the way on data centers.
02:46Europe lagging behind somewhat.
02:48What do you think we're going to see in the next five or so years on this front?
02:53Yeah, I think it's difficult to try and say that Europe and the UK aren't lagging behind.
02:58Maybe it's more easy to say that they've missed that first wave
03:02of what AI investment has really been about,
03:05providing these big transnational LLM APIs that we've seen
03:10in terms of like DeepSeek, OpenAI, Anthropic, those kinds of companies.
03:16But I think they're actually looking at the market in a very different way.
03:20If you look at things like Oriel photonic connections that are networking
03:25that's happening out of London,
03:26and some of the interesting things that are happening in France and Germany right now,
03:30it looks more like what Europe and the UK have got their target shifted on,
03:35is how can we facilitate AI in the coming future?
03:40And that, I think, is really interesting and points to just a kind of diversification
03:46within the AI market for those countries.
03:48Thank you so much.
03:49Dr. Connor McCohan is a lecturer in digital media at Stirling University in Scotland.
Comments