The Digital Climate Footprint: What AI Means for Bees (Ep. 70)
As AI transforms the digital economy, its hidden climate costs are becoming harder to overlook. A new report from Australia’s Centre for AI, Trust, and Governance offers the first big-picture look at how the rapid expansion of AI‑driven data centers connects to ecological impacts on native pollinator communities.
Between 2024 and 2035, data center growth could add 4.8 to 15.4 gigatonnes of CO₂ emissions, nudging global temperatures up by 0.0026°C to 0.0084°C. Those numbers might sound small, but the report applies a “bee barometer” framework that translates global climate shifts into local consequences for pollinators. Even slight temperature increases can spark extreme weather events that put ecosystems under stress. By zeroing in on key bee species, the report shows how our tech choices ripple into biodiversity risks and makes the case for stronger conservation, smarter innovation, and coordinated policy action.
Sharing his research on this episode is Dr. Rob Nicholls, a policy and regulatory specialist at the University of Sydney. This is the paper, The Digital Climate Footprint, that we discussed.
Good to know
Rob’s report shows that two Australian native bees are in real trouble: the Carbonaria stingless bee and the green carpenter bee. The stingless bee can’t survive when temperatures climb above 42°C which could lead to colony collapse. The carpenter bee, on the other hand, is losing its nesting spots as climate change drives more intense fire patterns. Between wildfires, direct heat stress, and a shrinking range, this species could be pushed to extinction.
Transcript
Jacy: [00:00:00] Welcome to the Bees Knees a podcast wild about native bees. Wild and native bees are under threat worldwide. In each episode, we look at actionable things we can do to support these adorable little guys whose pollination work is crucial for maintaining biodiversity. I’m Jacy Meyer and I thank you for being here.
Did you use chat GPT Gemini, copilot or another AI tool today? Did you say please, when you asked your question and thank you when you received it, and did you ever imagine those quick actions could be affecting bees? This episode considers the bee barometer a way of showing how the hidden energy demands of AI connect directly to the health of our ecosystems.
The digital economy is racing ahead, but the climate impacts aren’t just data on a page. They show up as real risks for pollinators, for [00:01:00] agriculture, and for the systems we all depend on. By 2035, data center growth alone could pump billions of tons of CO2 into the atmosphere, nudging global temperatures, just enough to push vulnerable species past their survival limits.
That’s the story we’re unpacking today with Dr. Rob Nicholls from the University of Sydney in Australia, and the author of the Digital Climate Footprint from Global Data Center Growth to Local Ecological Consequences. So your report introduces the bee barometer. Could you walk us through how this framework works and why pollinators are such sensitive indicators of climate disruption?
Rob: The idea behind the bee barometer was to come up with a way of actually expressing growth in global climate change in a way that lots of people could understand. So rather than saying a [00:02:00] world, there’s a temperature rise of 1.5 degrees by 2050, well. In my part of the world, if you walk outside and it’s 22 or 24, most people can’t tell the difference.
So coming to something which is quite specific, what effect will temperature rises have on pollinators? Actually just brings it a lot closer to home for people. The other thing that I wanted to do with the bee barometer was to be very specific. What’s the climate effect of just one component of global temperature rises?
And specifically it was the impact of data centers and even within data centers, it’s the growth of the use of AI and that. Increase in energy consumption by data centers just to answer [00:03:00] AI questions and the impact that that would have on pollinators. And as an example, we, we’ve already heard from Google that when you do a Google search, it takes a certain amount of energy.
If you do a Google search, which has a Gemini answer, it uses 10 times as much energy. So. When you start to think about that in the context of, uh, lots of people asking questions of chat GPT, it’s significant amounts of of energy. If you go a little bit further, Sam Altman, who’s the CEO of Open AI, was asked the question well.
for people who use the words please and thank you, in their interactions with chat GPT, how much does that cost? And his answer was, oh, about $10 million. But I don’t want to stop them [00:04:00] doing it. That $10 million is not in GPU chips. It’s in energy. It’s a power amount paid for, for power, and that has an impact on the climate.
Jacy: Those numbers are shocking.
Rob: It’s very scary. And I suppose the other thing that’s scary is looking at that narrow field. So what does the effect of data center growth to map how AI is growing? What does that effect, does that have on pollinators? I must admit, when I first looked at doing this, I was thinking what happens if I find that I can’t actually measure the effects?
And that’s, you know, that’s a normal question when you’ve gotta start doing some research and you might then go back and say, oh, well perhaps I won’t, won’t bother with that and look for something else. It really is significant. It’s significant on [00:05:00] farmed bees, on honeybees. It’s significant on different types of Native bee in Australia and the bee Barometer did honeybees because.
People recognize honeybees, but also looked at, uh, stingless bee and the Green Carpenter Bee, both of which are native pollinators that are actually incredibly important in agriculture in Australia.
Jacy: So we often think of these data centers as kind of abstract or invisible infrastructure. Um, talking about the, how their expansion is impacting these pollinator habitats you looked at, was there something that was very surprising to you or something that’s maybe been overlooked?
Rob: I think the, perhaps a surprising thing was to think about the growth of use of data centers, even just looking at Australia. So Australia’s going through the same type of [00:06:00] transition that. All countries are going through move to renewables, and part of that is increasing the amount of transmission in the grid.
Why? Because actually the transition is electrification moving away from other sources and moving towards renewables as the the source of energy. Based solely on the increase in use of ai, the power required by data centers on the Australian grid will rise from 5% of grid capacity to 15% of grid capacity.
That’s a threefold increase by 2030. Oh, well, 2030 isn’t far away. Uh, so it’s a very significant change. And so it, it started to think about all of the other [00:07:00] impacts that having data centers will have both. Directly the impact on land use and indirectly the increased amount of power that the data centers get rid of.
So broadly what you’ve got with a data center, and this is a another learning. If we looked at data centers five years ago, essentially what you find is that most of what a data center does in terms of power, is power. Uh, CPUs, processing units, computer or central processing units. In order to cool those units, you need air conditioning of some sort, and that could be genuine air or more likely, uh, cooled water that’s going across the CPUs and most of the [00:08:00] CPU manufacturers, which are people like Intel and AMD and increasingly Amazon, Google and others.
Their focus was to have chips, which used the least power possible for the most computing power. Why? Because then they would be chosen for use in data centers. When you look at AI, they used GPUs or graphical processing units, and those chips were originally used in things like games, computers, to accelerate the videos. So that there was no lag and you had fabulous rendering of first person shooters and things like that.
Those devices where the power computing power or graphical processing power of these is doubling every 18 months or so. They’re simply not designed [00:09:00] to minimize power use. They’re designed to maximize processing. Uh, capabilities and for now you even see it in the deals. So there’s been deals done between open AI and NVIDIA or AMD and open AI, uh, AWS and open AI.
And what are they measured in? Not processing units. Megawatts because what’s happening is you can say, oh, well, we know that GPUs require a certain amount of power, so we can specify what power is required by looking at. The energy requirement. And so immediately you’ve got a shift from how much processing will it do to, well, let’s forget about that.
And our, our quick way of measuring it is to talk about the power that’s required now [00:10:00] in an environment where. Actually, what we really need to do is reduce power requirements. That’s a significant impact, and it was something that has an impact on all of us, but it also has an impact on pollinators.
Jacy: So I wanna look at Australia specifically for a minute.
The country’s agriculture relies heavily on pollination, yet your wild pollinators face distinct threats, heat stress, wildfires. What does your analysis show about the vulnerabilities of native bees? And how could those risks ripple out to affect our food systems?
Yeah. And so what I tried to look at was, well.
In three different cases if there’s constrained growth of data centers. I, I was looking at global data centers and saying, well, Australia will probably follow a global trend. I’ll come back, um, a little bit later on and explain why actually Australia will go a bit faster than global [00:11:00] trend. So I had a constrained growth case, a base case, and a high growth.
Case, essentially what the, this carbon area, a Stingless bill, bill, it be ends up with, uh, heat stress. So the heat stress probably doesn’t sound significant in many countries. It’s on the forages when the temperature goes above 42 degrees Celsius. Now this particular bee lives in the northeast of Australia, 42 degree days are not uncommon.
They used to be very rare, but they’re becoming more common. And essentially, if in the high impact growth, you’d see the local population is. Dropping or the viability of the local population dropping by 1.1% each year, [00:12:00] that’s a significant impact. Similarly with the green carpenter bee, this is impacted by fire, basically.
Essentially habitat degradation and again. Because of the risk of habitat degradation, you’ve get got two effects. One is that the amount of unsuitable habitat. Could drop by 8% in a decade. That’s a very significant change, but it also changes the agricultural practices so that you tend to have a higher risk for the the bees.
In any case, and I know your podcast doesn’t focus on honeybees, but we actually in Australia have a very significant number. Of wild honeybees, and they’re critical for a couple of reasons. One, they’re really great pollinators, even [00:13:00] though they don’t live in hives and they don’t get moved to, uh, a particular region.
So the part of the agricultural process is shifting hives from place to place. They don’t have that, but what they do do is increase and improve the genetic makeup of. The farmed honeybees, and the key thing here is that not only do you have the reproductive health risk that honeybees have Australia up until literally three years ago, didn’t have varroa mite.
Now that we have varroa mite. Farmed bees are treated with myocyte, wild honeybees are not, and their ability to withstand the varroa mite decrease is really markedly with temperature. So there’s a set sort of secondary effect, [00:14:00] which is likely also to have a big impact on agriculture.
So you laid out a number of ideas, some policies, and some market based for making digital infrastructure more sustainable.
Which ones do you think we could act on right away, and how might they actually help protect pollinators from climate stress?
Rob: Okay. I think the one that makes a lot of sense is to say, well, any new data center should be sustainable. So how do you make a data center sustainable? Well, in the Australian context, it’s actually pretty easy in respect to power, A bit harder in one other respect, so.
A logic saying, if you’re gonna build a data center, build a big one, build a big data center you really want it to be somewhere
outside of the metropolitan areas. [00:15:00] So if you’re gonna do that, then having a local use of solar and wind with a very big battery has the potential for saying, well, there’s. Sustainability in terms of power requirements, we don’t actually need anything else. And in Australia where the sun shines a lot, the wind’s a bit less reliable.
But, uh, like everywhere in the world, batteries are cheap these days. That makes a lot of sense. There’s a second problem though, which is that data centers require water in order to cool. And the sustainability of water is more of a challenge in Australia. So why are there areas outside of the metropolitan areas where people don’t live in That’s great site for a data center?
And the answer is ’cause there’s not much water there. Um. But with when there is water, when it rains, [00:16:00] actually collecting that water and reusing water within your data center. And the ideal thing is to sell the warmed water to businesses that need heating. It’s a bit harder in Australia ’cause we don’t need heating very much of the time.
Um, but that also starts to get to a sustainable output. It also has a second effect, and this is a pure, again, a purely Australian one, which is running fiber to data centers. So getting the connectivity isn’t too hard, and there’s look of a number of projects at the moment, which are looking at very large solar, solar panel arrays backed by batteries.
To export energy to nearby countries like Indonesia and to Singapore. Well, if you’re going to do that, having data centers nearby is also a great thing to do [00:17:00] and can be done sustainably. So you start to get to, well, would something that Australia has that might be useful in the regions? To actually get data center operators to do these reforms.
And that is, well, I speak markets nearby and the logical thing to do reduce your power costs and you’ll be competitive against data center operators that haven’t got a sustainable outcome. So there is a market-based mechanism there. Problem with market-based mechanisms is they’re always really good in theory, and sometimes they don’t always work.
And so to have a regulatory or legal reinforcement of the market is not a bad idea to make these things happen.
Jacy: And is the technology available now to run these data centers and all the power that they need solely on solar and [00:18:00] wind as well as with kind of the water regeneration that you mentioned?
Rob: Yep. Um, all of those technologies you do need a battery are available now, and businesses like Google are, are actually have built and are building data centers, particularly in Arizona and the, the Arizona and New Mexico states, which are dry, windy, and have lots of sunshine. They’re already building them now, so this is not some, well, we’ll be able to meet the, the issue in 20 years time and contribute to net zero that way.
This is a no, do it now. And the economies of scale that flow from large data centers are also valuable. So Google isn’t. Necessarily doing it because it’s a, a great philanthropic organization. It’s doing it [00:19:00] because it’s financially worthwhile to do it. That’s a logical thing to do.
Jacy: Absolutely. So your report and kind of the outcomes did focus on Australia, but you did mention that you kind of looked at data centers globally.
Australia does have unique climate challenges, but do you think the findings can apply to other countries looking into developing sustainable digital infrastructure? You just mentioned Google and their work in the United States, Arizona, New Mexico.
Rob: I do, but it might be that the mix is different. I think the first thing to say is, well, the warming is global.
Yes, we have some particular climate issues here in Australia, but actually it’s the fire seasons getting longer and overlapping with Northern hemisphere fire seasons. These are purely global warming phenomena. So. We share these issues across the globe, [00:20:00] just happens that, uh, I’m sitting in Australia, but it doesn’t actually matter.
A data center here is gonna have impact globally and so are data centers around the world. So will. Is it immediately transport everywhere The sustainability part is. So it might be that you start to say, well, actually in a northern hemisphere, a European country, actually the value of that warmed water.
As part of heat as a service that flows and useful in Australia is really important in the economics in countries in Northern Europe. So you start to get to changes and it might also be that you say, oh, well actually, uh, solar isn’t as important as wind or tide in other countries. With that, getting that [00:21:00] balance right is something which ends up being localized, but some form of renewable energy source or multiple sources, plus a battery plus.
As much as you can, a closed loop water system and getting rid of the heat from that water system, either selling it or using it in some way or letting it go up into the atmosphere are different ways that different countries can deal with this. I think. The hardest part is that there is in the US at least, uh, a little bit of nimbyism, so not in my backyard for data centers and.
Being able to show that a data center being built is sustainable and won’t have a local impact on people within a community is a really important message. So I think it is something that [00:22:00] is able to be used in a variety of different countries.
Jacy: What we’ve heard is sobering. Fractional warming from AI driven data centers can cascade into lethal heat stress in stingless bees and habitat loss for those carpenter bees.
And what’s worse? AI workloads make data centers consume up to 10 times more energy than traditional ones. Australia’s bee dependent agriculture is already at risk and the choices we make about digital infrastructure ripple far beyond server halls and indeed Australia alone, the bee barometer reminds us that every click query and model run carries ecological weight.
Rob mentioned that the technology is in place now to keep data centers sustainable as demand skyrockets. Companies need to look for vendors based on verified sustainability. And regulators need to set smart grid rules. The whole report is fascinating, and there’s a link to it on the [00:23:00] website, the Bees Knees website.
Thanks to Rob for the eye-opening conversation and you for listening. Until next time, let’s keep our digital buzz from drowning out the bees.