Behind the Prompt: Data Centres, Water, and Australia's Energy Crunch with Tim Prosser

In this episode, digital sustainability strategist Tim Prosser returns to The Green Fix to unpack the physical footprint of data centres — the invisible infrastructure behind every AI query, cloud service, and streaming session.

Tim breaks down what corporate sustainability professionals need to know right now: why data centres could consume 11% of Australia's electricity by 2035, what the CEFC's reliability gap warning means for energy prices (and your mortgage), and why annual offsets are no longer cutting it. He explains the shift to 24/7 carbon-free energy matching, the emerging water risks that have Sydney, Melbourne, and Hobart mayors sounding the alarm, and how direct-to-chip liquid cooling could halve data centre water use.

Plus: six metrics to see through greenwash, procurement moves ASX-listed companies can make today, and why your sustainability team needs to be talking to IT — yesterday.

Whether you're reporting under Australia's climate disclosure framework, managing ESG for a company with cloud infrastructure, or just trying to understand where the grid is heading, this one's essential listening.

Find Tim Prosser at sustainabilitydigital.com

Resources

• CEFC (Clean Energy Finance Corporation) "Data centre growth and the energy transition"
• C40 Cities Network
• "Sustainable IT Playbook for Technology Leaders"  By: Niklas Sundberg, Richard Pastore

Transcript

Why Data Centres Matter Now

Dan Leverington 0:02

We're thrilled to have Mr. Tim Prosser joining us again on the Green Fix, and we specifically want to be speaking to him today about data centres, an industry that's been building very, very quickly across the world and specifically Australia. When it comes to digital sustainability, sustainability teams are really starting to ask what impact do the data centres have on companies from a sustainability, climate, water, and nature perspective. And so we thought we'd go straight to the source and uh have Tim come back and join us.

Tim Prosser 0:41

Good to be back. Good to be with you, Loretta and Dan, and good to be back for my green fix. Absolutely.

The Invisible Energy Behind AI

Dan Leverington 0:46

So we want to we want to start off um really talking about what is this invisible impact, if you like. So most corporate leaders associate emissions with supply chain or travel. How can we help our listener visualize the physical surge of energy triggered every time their teams are using generative AI?

Australia’s Grid Reliability Warning

Tim Prosser 1:06

A common question, Dan, I get this several times a day at the moment. Understanding technology-related emissions really requires a systems lens on supply chains. So for listeners, it's important to recognize that every digital interaction, a Google search, a like or a comment, emails, photos, videos, and particularly prompts to AI activates a substantial physical behind-the-scenes infrastructure. Um the same dynamics apply to uh AI outputs, especially when generating AI images and videos. To grasp the scale, consider that a single 100 megawatt data center uses power and fresh potable drinking water at levels comparable to up to 150,000 homes, so quite a significant amount. Billions of daily AI computations drive rising power and cooling demands, making data centers among the grid's most rapidly growing energy consumers today. Projections specifically in Australia illustrate the broader implications. By 2035, data sector the data technology sector could account for as much as 11% of the nation's total electricity consumption. And that's a massive, massive increase from around 1% to 2% today. So this covert yet accelerating electricity draw represents a new industrial load with energy intensity comparable to heavy manufacturing, such as a large aluminium smelter, which we might already be familiar with. So digital activity translates to tangible emissions and resource use through data center operations, network infrastructure, and cooling and such. The environmental footprint scales with AI adoption. So that definitely includes the generation of images, or particularly includes video image generation and complex computations. So strategic planning, energy efficiency, renewable integration in data centers are essential levers to manage this emerging demand.

Loreto Gutierrez 3:06

Tim, that's really fascinating. And I'd love to bring this into the context of the projection AI is projected to add roughly $600 billion to Australia's GDP. That's an expectation for 2030. But it also has the potential of driving up power prices by 26% in New South Wales at least. So how do you think Australia's um current grid is being prepared for that trade-off?

Hourly Clean Energy Matching

Tim Prosser 3:31

A great and very timely question, Loretta. So recent findings from the end of 2025 from a study conducted by the Clean Energy Finance Corporation, the CEFC, which is Australia's specialist climate investor under the Australian federal government, shed light on the data centre boom and its implications for the nation's energy landscape. The research indicates that the growth of data centres is reshaping Australia's future and projects a reliability gap in New South Wales by 2033-34 if only current committed projects proceed alongside the committed coal retirements. This serves as a wake-up call to us all. We can drop a link into the research in the show notes if you like. Some of the key takeaways for listeners, the rapid expansion of data centers is placing a material load on the electricity grid, creating medium-term reliability pressures. Listeners will be familiar with Australia's national renewable energy target of 82% renewables by 2030, but many renewable projects are behind schedule, which has necessitated the extending of the life of several fossil fuel generators to fill that shortfall. In practical terms, to support data center expansion plans, the CEFC recommends an incremental add-on of 3.2 gigawatts, that's quite a large number, of new renewable generation beyond existing commitments and 1.9 gigawatts of storage capacity all by 2035. The underlying premise is clear. Achieving these targets requires not only a rapid deployment of generation and storage, but also the timely execution of transmission and grid integration projects. Coordinated planning and implementation across generation, storage, and transmission will be essential to close the reliability gap and stabilize prices. And this will help support a resilient energy system for Australia. This is a topic we're going to have to monitor very closely over the short term. The rapid build-out of data centers is projected to significantly increase electricity demand, which, when combined with rising network costs and the removal of government subsidies, will drive up energy prices. Since energy is a major component of what economists refer to as the consumer price index, these higher prices will sustain elevated inflation, thereby compelling the Reserve Bank to keep interest rates high to bring inflation back within target. So for anyone with a mortgage, take note.

Dan Leverington 6:09

Yeah, or anyone who doesn't have renewable energy powering their home or business. I guess the traditional response from most businesses has been, well, if we've got an emissions problem, then why don't we just grab offsets? And so as we move into this new world of data centers being powered by renewable energy, why is it that matching renewable generation in real time is becoming the new gold standard?

Tim Prosser 6:33

This is really getting to the nub of the threading the needle challenge we we have here, Dan. About 97% of emissions reporting frameworks align with the greenhouse gas protocol. This protocol defines the standards for scope to emissions reporting, as we all know. These standards are currently undergoing review in part to strengthen integrity and transparency with potential implementation as early as 2027. We're already in 26, so wow, it's coming. Many current net zero strategies rely on annual offsets, as you mentioned, which means data centers continue to consume vast amounts of fossil fuel electricity from the local grid. By contrast, real-time or 24-7 hourly matching aligns electricity consumption with clean generation every hour. This approach reduces dependence on expensive fossil fuel peaking plants and mitigates grid price volatility. Leading organizations, including Google and Microsoft, are advancing towards 24-7 decarbonization with a target horizon of around 2030. This shift moves towards genuine physical decarbonisation of operations. Time-stamped certificates are being developed to support this standard, particularly in Europe. These certificates incentivize battery storage and grid flexibility, which are essential to truly greening the grid.

Loreto Gutierrez 7:58

And as a follow-on from that, on what you're mentioning, I'd love to understand how data centers can go from being these passive power consumers to almost acting as virtual power plants.

Tim Prosser 8:12

Great question, Loretto. We're moving away from passive concrete boxes in industrial estates to a new beginning. So a sustainable data center as an active value-generating grid asset. We're not there yet, but to achieve this, data centers must evolve through large-scale energy storage that helps coordinate with the power system so it becomes an active participant in the community and the grid. So data centers can store excess renewable energy in large-scale batteries, solar soakers, if you like, and release it during peak demand, thereby stabilizing the grid rather than contributing to peak stress in the evening. Now moving on to how we use AI to optimize the operation. Advanced AI can help shift non-urgent workloads to periods of high renewable output and orchestrate participation in virtual power plants or VPPs. This enables facilities to generate additional revenues by providing essential, what the industry referred to as frequency control, so balancing the grid services and to help lower peak demand load, again, particularly in the evening, and operate as large, flexible energy hubs. Now, just to give an illustrative case study, your listeners might be familiar with Google DeepMind. Today, one of its core applications in data center operations uses AI to optimise cooling for overall energy use, reducing both power consumption, in some cases as much as 40%, and operating costs while preserving the reliability of the data center.

Dan Leverington 9:49

Which is pretty phenomenal results in a relatively short period of time. On the flip side, there's uh a few Australian mayors who are increasingly raising alarms about water scarcity issues. And we know that uh traditional data center air cooling is very water intensive. How critical is the shift to direct-to-chip liquid cooling?

Tim Prosser 10:11

Yeah, you're right at the bleeding edge of the uh the industry conversation at the moment, Dana. A brief context for your listeners. In late 2025, the mayors of Melbourne, Sydney, and Hobart join forces with other global city leaders to launch a forward-looking initiative under the C40 Cities Network ahead of COP 30 in Bilem. Some of the key goals of the initiative are to establish shared standards, transparency and robust environmental protections and industry collaboration to align data center development with the Paris Agreement. The shift towards liquid cooling is increasingly critical in terms of efficiency and reliability. Now to the water and energy considerations. The city of Sydney has highlighted a potential risk. Data centers could consume up to 2% of the city's drinking water by 2030, absent of any additional water efficient strategies. So bear in mind we rely on the desalination plant in New South Wales and Sydney. 15% of our water comes from the desalination plant. So if you link that, that's a problem. Traditional evaporative cooling, while energy efficient, relies on substantial drinking potable water streams, raising sustainability concerns in water-scarce urban contexts. And Australia is recognized as the driest inhabited continent on earth. And innovative directors chip liquid cooling presents a compelling alternative with estimated water use reductions in the order of about 50%, enabling higher density cooling without demanding fresh water resources.

Loreto Gutierrez 11:48

That's really Tim. How accessible are these advanced cooling technologies in Australia?

Tim Prosser 11:55

They're coming often by necessity. Some of the infrastructure that is installed in order to enable AI necessitates better cooling methods. IT equipment becomes very, well, it it becomes very risky to operate when it doesn't work properly. And if it's not cooled, it may not work. So some of the traditional cooling methods are not allowing it to operate effectively. So it's just not an incremental option. It's something that's been driven by the technology.

Loreto Gutierrez 12:21

That actually gets us thinking a little bit about asset lifecycle, right? So how long can these assets be in use, remaining um relevant for the current technology demands? What can we expect from a lifecycle perspective when things are changing so quickly?

Tim Prosser 12:36

It's a very current question, one that I know the industry is motivated by. Let's call it that. We're talking about constructing assets which should have a lifetime of at least 20 years. And I don't need to tell you or your listeners that the AI horizon, you know, six months feels like affinity. So the challenge at the moment is there's a lot of infrastructure being commissioned or built into some of these data centers that can cater for any eventuality. So there's a lot of additional stuff going in, which we know probably won't get used. Yeah, it's really tricky. And particularly when you're looking at the difference between AI training versus AI inference. We don't really have so much focus on AI training in Australia. It's mostly in the US or overseas. But we're about to see a significant investment wave in AI inference in Australia. I know a number of data center operators have already partnered with, say, OpenAI, to have sovereign-based locally hosted versions of that to then incorporate Interim into their secure enterprise environment. So that's that's in play at the moment, but it's it's a tricky one.

Loreto Gutierrez 13:40

And then from an investor perspective, is there a risk the demand might plateau or might decrease? And for the investors, how can they distinguish between a sustainable long-term asset and potentially what will could become a stranded asset?

Tim Prosser 13:55

Yeah. Um I think the first response is if if we don't think we're invested in data centers, listen again. Um, and that's how I'm going to frame my answer. Investment in data centers is a global momentum driver. It's driving a lot of value creation in every aspect of every industry and every economy. And it's frequently embedded within specialist commercial property portfolios. Traditionally, these assets have provided long-term dependable returns. However, the AI era is redefining that risk return landscape. For many listeners, exposure to data center assets may be embedded already in your superannuation portfolios via commercial property holdings or international technology investments and the like. AI-enabled data center asset class still has ground to cover before the risk and return stability are fully established. Nonetheless, industry observers suggest AI-enabled data centers can deliver substantial lifetime value with potential earnings in the vicinity of $10 to $20 billion per gigawatt of capacity over their lifespan, though quite a big number. In the Australian context, investment should prioritize assets that demonstrate resilience beyond AI pipe. Sustainable data centers differentiate through energy efficiency and design. So they pursue low power usage effectiveness or PUE in the industry, targets to implement end-of-use efficiency measures across the facility lifecycle. They also invest in renewable energy procurement and they secure long-term, verifiable renewable energy supply. They also help with grid responsiveness capabilities. So they enhance flexibility with integrated energy storage, batteries, virtual power plants to monetize those grid services and create additional revenue streams, even during periods of lower compute demand. And then finally, diversified tenancy, and this is probably a really important point. Avoid over-reliance on single high-density AI workloads, so all your eggs in one basket. Diversified tenant mix with strong ESG commitments helps mitigate the risk. Conversely, assets lacking verifiable green credentials or reliant on single high-density AI demand profiles. So think about AI factory AI training factories for one customer with one chipset or technology solution may confront stranded asset risk. So the um a bit of an industry note, um, I was on a Bloomberg NEF session last week. So while we discussed the data center investments, it's worth noting that Bloomberg New Energy Finance reported global investment in data centers at 441 billion US dollars in 2025, with projections to rise to around 712 billion per year by 2028.

Dan Leverington 16:46

Yeah, fascinating. One of the areas that we're also very interested in is how can they be utilized for community benefit as well. And so we've seen in Paris there's examples of using data centers to uh heat community swimming pools. What are you seeing as some of the most viable waste heat opportunities for Australian facilities, particularly in urban areas like Sydney or Melbourne?

Tim Prosser 17:10

Yes. Um in the 2024 Paris Olympics, a data center won. There's over 11,000 in the world, in the city recycled its waste heat to help warm the water in the Olympic Aquatics Centre by converting excess thermal energy into hot water, heating the pool and surrounding buildings. District heating networks are less common in Australia than in Europe. We live in a warmer climate, but select commercial hubs present opportunities for Australia to think about this. In Sydney, the Western Sydney erytropolis, near the yet-to-be-opened WSI airport, there's a strong candidate for channeling industrial waste heat into co-located manufacturing and light agricultural facilities. In Melbourne, cooler climate further enhances the potential for residential district heating, achievable through collaborations between data centers and energy service companies to manage the infrastructure or by supplying heat directly to near nearby industrial users, converting that waste heat byproduct into a revenue stream.

Loreto Gutierrez 18:13

So what are some of the um trends that you're seeing in regards to location of data centers in Australia? Where are you seeing data centers being located or planned for the future?

Tim Prosser 18:25

90-95% of the large-scale high availability data center services are based in Sydney and Melbourne. What's interesting about those locations is both of those locations are heavily reliant today, still on both the grid and fossil fuel generations behind that. So a lot of our cloud compute is quite dirty on a global scale today. People don't like me saying that, but that's true. We we need to clean that up quickly by choosing some cleaner locations and providing cleaner alternatives in Australia so people can make suitable actions based on that knowledge.

Loreto Gutierrez 18:59

So then um I guess there's a bit of a tension between locating a data center in a renewable energy zone and being close to the workforce in less than ideal grid mix situation.

Tim Prosser 19:16

Yeah, this is another threading the needle question, Loreto. The placement of data centres is shaped by three core basic factors: um, reliable access to power, water, and fibre, those three things. In practice, supply resilience constrains the viable locations, so where those three services intersect and multiple sources of those intersect. So as the high availability or resilience requirements come on our technology, they need multiple backup sources of those three inputs. So policymakers should pursue locational coordination, tailoring incentives to the data center functions. For your listeners' benefit, not every data center is the same. There are different um they host different services. So, for example, co-location, cloud services, AI training, AI inference, that there are others. To your earlier point around architecting and knowing exactly what you're going to be using it for helps with the sustainable outcomes by reducing that redundant infrastructure we we spoke about earlier. Um low latency co-location services, that's maybe a bit of a bit of a technical term, but um will benefit from urban proximities near the users, the consumers of this technology, um, to minimize latency and avoid streaming disruptions. So remember how frustrating in the middle of a cliffhanger in a Netflix binge you get a buffering situation? Not great. So conversely, hyperscale facilities supporting AI training or crypto mining, for example, are less distance sensitive and are well positioned in renewable energy zones, like, for example, the Hunter Central Coast region in New South Wales and there are others in Australia. So leveraging the REZs can help reduce the transmission losses of getting that renewable energy to city urban centres. To address workforce and fibre connectivity barriers in the regions, because they're not populated and they don't have great fibre connectivity, maybe. Governments could establish dedicated hubs that offer expedited planning processes and favorable land incentives. I'd keep an eye on that point because the federal government are about to release some policy guidelines in exactly this space in the next month or so. So keep an eye out for that. So some of the key takeaways for sustainability-minded stakeholders. Align data center sighting with resilient, diversified power and fiber sources to bolster reliability and reduce environmental risk. Differentiate incentives by data center function. And maybe this is something that this community can help with. So when we know what the solution is going to be used for, we can better design the solution, and that helps optimize energy. Use and support renewable integration. Consider REZs for large-scale energy-intensive workloads to maximize renewable access and minimize transition loss. I've yet to see that really effectively done in Australia, but people are certainly asking that question. An investment in regional planning hubs to accelerate green infrastructure deployment and address grid constraints. And finally, monitor and mitigate curtailment by strengthening grid interconnections. So this is quite a technical engineering kind of answer, but accelerating the storage and demand response solutions in our deployments. Thank you, Tim.

Metrics That Prove Real Progress

Dan Leverington 22:36

I feel like we all know a lot more about data centers and the role that sustainability are going to be playing over the foreseeable future. There's absolutely a lot of gaps that need to be filled in order for this to stay aligned with Australia's and a lot of other countries. A couple of other questions that we'd like to ask you about based on the previous conversations that we've had with you. So you've long advocated for green ops within tech teams and moving from slogans to data-driven accountability. What specific metrics should sustainability professionals look for to ensure that their supply chain risk exposure isn't increasing due to the increasing usage of data centers?

Tim Prosser 23:18

Yes, absolutely, Dan. I'm very glad to hear you've been paying attention. To assess true risk, look well beyond the glossy annual net zero claims and demand metrics that demonstrate real operational resilience and tangible decarbonisation. And this probably isn't limited to technology or data centers. This is probably a broad offering across the sustainability field. So listeners might want to consider several, if not all, of the six suggestions. 24-7 carbon-free energy or CFE. Does energy use align with clean generation on an hourly basis? Or does the organization rely on offsets we mentioned earlier while burning fossil fuels at night? Additionally, are they underwriting new emphasis on new renewable energy projects via PPAs or similar mechanisms? Or are they merely purchasing certificates often for infrastructure built many, many years ago without funding new capacity? They could also consider demand flexibility. So can IT workloads be shifted or can on-site batteries be employed to avoid peak grid pricing and reduce the strain on the system and also the high carbon intensity energy periods? Fourthly, water usage effectiveness, or WUE is the metric. Is water consumption minimized relative to rising scarcity and industry benchmarks? And a point that's probably not unfamiliar to listeners, activity-based location metrics. Apply location-specific metrics across all products and services rather than relying on aggregated figures. And my favorite, curiosity-driven verification. So ask probing questions until verifiable facts and robust data emerge. A point that really frustrates many on the receiving end.

Procurement Moves That Cut Emissions

Loreto Gutierrez 25:08

Yeah, thanks, Tim. And just on that, actually, I know we talked about this when we talked about digital sustainability and decision making in the IT procurement process. Is there anything specific that you would call out to Australian ASX listed companies if they do want to reduce their carbon footprint today? Are there any recommendations on what they can implement in their IT procurement when it comes to data centers?

Tim Prosser 25:35

Yeah, uh that's a big question. There are many depends to that question, which you know, what industry they're in, what what the strategic ambition is around sustainability, lots of um precursors. But firstly, a bit of an insight that you and I'm sure listeners are aware of, but it helps set the foundation. Approximately 80% of group one entities currently required to disclose their non-financial climate information in Australia starting this year are not ASX listed. Most of these reporting entities are privately held or listed overseas, which has implications for governance and reporting standards. So it's it's a much broader challenge than just the ASX. Now to your question, the the most consequential shift in the obligation for data center and cloud providers is to align with the Australian sustainable finance taxonomy. This entails operating without on-site fossil fuel combustion, so no diesel generators. I I'm seeing planning applications for large data centers at the moment going with 800 plus diesel generators right now, maximizing energy efficiency, so targeting low power usage effectiveness, sourcing 100%. That's genuine, 100% renewable energy on a 24-7 basis, not based on annual offsets which wash out the integrity of that metric. Some strategic procurement guidance. So prioritized power purchase agreements that fund new renewable energy projects, um, additionality, we call it, and aim for 24-7 carbon-free energy matching rather than relying on non-local standard um energy uh annual offsets. So no longer is it um allowed to buy uh a rec that's produced in Iceland for something that was built 30 years ago. That's not the game of the future. So by concentrating on new local renewable energy generation and continuous carbon-free energy, um, digital infrastructure investments will actively contribute towards towards Australia's energy transition rather than merely absorb existing clean grid capacity. So really important point. Um a bit of context and implications for this. This this framework reinforces accountable climate disclosures beyond the ASX listed population, and specifically for data centers and cloud providers, the transition to 24-7 carbon-free operation represents both a compliance obligation and an opportunity to advance decarbonization.

Dan Leverington 28:00

Tim, this has been so helpful, so insightful. We'd really love for you to give us a summary of the takeaway that you want our listener to have, and then a couple of calls to action that they can follow up, having listened to this interview.

Tim Prosser 28:14

Yeah, thanks, Dan. I'll be brief. Be sure to visualize the digital emissions. I talk about digital sustainability and green ops, very passionate about that. Push for rapid, renewable, and grid-friendly tech. Adopt 24-7 decarbonization initiatives and projects, and invest in efficient cooling and storage and assess long-term asset risks. And that's about it, really, Dan. I think there's if people have lasted this far in the podcast, I think that we've got them on board. But really go and have a chat to your IT team and bring them along for this journey. Don't do this in isolation. This this is a big challenge.

Dan Leverington 28:52

I couldn't have summed it up better. We love having you on the podcast.

Loreto Gutierrez 28:55

I just want to acknowledge as well that this is a highly technical topic. It's daunting, it's confusing. And to add to that, it's changing every single day. So I can completely empathize with people in industry, whether they are sustainability leaders, ESG practitioners that are in charge of the measurement or even employees of data center companies, would be so overwhelmed with this topic. So I just want to thank you because you are doing so much to bring the uh information, the knowledge, but also decoding it for a lot of people. If people wanna, you know, ask you for help. So are you open for chats about the topic?

Tim Prosser 29:36

Sure, absolutely. Um I would say embarking on this space, it's like an overnight success 20 years in the making. You know, I I've worked in the technology industry for more than 20 years, specifically in sustainability and climate tech the last three or four years. I'm seeing a lot of challenge and opportunity in equal measure. I'm certainly open to, and I have a calendar session for no commitment kind of initial chats with people. Um, go to my set website, sustainabilitydigital.com. But I would highly recommend listeners that are in the sustainability domain that don't yet have a technology offsider to make sure they connect with their technology team. I keep repeating that in my previous two podcasts with you guys. I would highly recommend again the book recommendation I made last um episode. Again, for those listening, Sustainable IT Playbook for Technology Leaders. It does have a technology-specific audience in mind, so it's less tailored to sustainability practitioners. But if you can be the referral into your technology team for that book, I wouldn't say it's a grab and go and read on the beach because it's about three kilos. It's a really heavy book, but it has some fantastic case studies, probably 25, 30 global case studies of companies that have gone through this journey, tried and tested various um sustainability programs. There are a whole bunch of open source frameworks and tools and things in the book. I'm I'm partway through reading it myself. It's 850 pages mine, but I am very confident the more I'm reading this book that you just got to read it.

Loreto Gutierrez 31:06

Yeah, thanks, Tim. We really appreciate you coming on the podcast. Uh we love chatting with you. And look, third time on the podcast may not be the last. So um again, keep doing the great work that you're doing. And thanks again for coming.

Tim Prosser 31:20

Thanks, Loretta. Thanks, Dan. Great to be here. Thanks, Dev.

Loreto Gutierrez 31:24

This was Green Fix with your hosts Loretta Gutierrez and Dan Levington. You can get your Green Fix every two weeks on Apple Podcasts, Spotify, or Pocket Cast.