MORSE, Mr James, Head, Industry and Policy, Macquarie Technology Group ZALLOUA, Mr Marcel, Joint Chief Executive Officer and Co-Founder, ResetData CHAIR: Welcome. I understand that information around parliamentary privilege and the protection of witnesses giving evidence. We'd now like to invite you to make a short opening statement, at the conclusion of which members of the committee will be invited to ask questions. Mr Morse : Thank you for the invitation to appear today. Macquaire Technology Group is an ASX 200, majority Australian owned and controlled technology company. We provide cybersecurity, data centres, and cloud and communications services to government, hyperscale and enterprise customers. We host and protect the data of federal and state government departments, critical infrastructure providers including banking and finance, the defence industry, health care, higher education, research, food and groceries, transportation, energy, water and utilities—essentially all of the SOCI designated sectors. As a company, we have a long history of proactive and constructive engagement with governments and parliaments at the Commonwealth, state and territory levels, and we regularly participate in consultations relating to technology, national security and resilience in the interests of helping shape positive policy outcomes for Australia. Our objective is not to advocate simply in our own commercial interests but to help ensure Australia's regulatory settings support investment, innovation, national resilience and sovereign capability. Our submission to the inquiry focused primarily on the committee's consideration of sovereign AI capability, data sovereignty, cybersecurity, critical infrastructure resilience and the adequacy of Australia's governance and regulatory arrangements to meet the opportunities and challenges which AI capabilities provide. The central proposition of our submission is that Australia is currently defining AI sovereignty too narrowly. Much of the policy debate focuses on whether data is stored in Australia or whether AI infrastructure is physically located here. Whilst these matters are important, they do not necessarily tell us who ultimately owns, controls or governs AI capabilities we currently rely on. Our view is that the more important questions are: Who controls the model weights, software, credentials and intellectual property? Which jurisdictions govern those assets? Who can alter or withdraw access, and can Australia rely on that capability during a crisis? Access to AI is not the same thing as sovereign control over AI. Accordingly, Macquarie has recommended a risk based approach to AI sovereignty. We do not advocate blanket localisation requirements, nor do we argue that Australia should try to build every layer of the AI stack here domestically. Rather, we propose that government determine which AI use cases are sufficiently important to our national security, critical infrastructure, cybersecurity and essential government functions to warrant greater Australian control over that capability. For those high-consequence applications, Australia should consider requirements relating to governance, auditability, continuity of service, incident response, access to evidence and Australian jurisdictional control. We have also recommended that the Commonwealth use existing policy levers, including procurement, assurance frameworks and targeted regulation, to strengthen Australian capability in strategically important layers of the AI ecosystem. Where governments invest public funds in AI capability, we believe those investments should deliver both a strategic return for security and resilience and an economic return through Australian capability development, innovation and commercialisation. Finally, our submission argues that Australia does not face a choice between sovereignty and openness. We should actively welcome investment from global AI companies and continue to benefit from international partnerships. However, attracting foreign AI investment should not be confused with building sovereign Australian capability. The national challenge is to determine where trusted access is sufficient, where meaningful Australian participation is desirable and where Australian control is essential. Thank you, Chair. I look forward to the committee's questions. Mr Zalloua : Chair, Deputy Chair and members of the committee, thank you for having me here today. ResetData is an Australian-owned and -operated AI infrastructure company. We have had an Australia-first business where the infrastructure we develop is for Australian government defence and Australian enterprise, in its first instance. Australia is at a crossroads when it comes to our AI future. We all recognise the enormous opportunity AI presents for productivity, innovation, Australian businesses and our economy. But the question we need to ask is: where will the value created by AI ultimately end up? Will we build the capability here so that Australian businesses, workers and communities benefit, or will we increasingly rely on capability controlled offshore, with much of that economic value flowing overseas? At ResetData, we build AI factories. There is an important distinction between an AI factory and a data centre. A data centre provides the physical infrastructure to store and process information. An AI factory brings together the entire tech stack of AI—that is, the infrastructure, the compute and the platforms needed to turn data into intelligence, applications and, ultimately, economic value. That distinction matters when we talk about sovereignty. 'Sovereign AI' is a term that gets thrown around a lot, and too often the conversation starts and ends with where the data is stored. But Australian data sitting in an Australian data centre does not necessarily make it sovereign. We need to look across the AI stack at who owns and operates the infrastructure, who controls the compute, where the models are trained and run, who has access and where the intellectual property sits, and, ultimately, which jurisdiction applies. Australian data is valuable. But the greater opportunity lies in what we do with it: the models we develop, the applications we build, the intellectual property we create and the businesses that grow around it. This is the capability Australia needs to retain, if we are to harness the value of AI for Australia long-term. Global technology companies will absolutely remain part of that ecosystem. But we need meaningful Australian capability. This isn't about asking for another pot of government money. The Commonwealth is already spending on AI technology and infrastructure. It's about ensuring that the expenditure goes to Australian providers across the tech stack who pay tax in Australia. There are a few immediate actions that government can take. Government can use procurement to help Australian businesses scale and invest in infrastructure and skills. It can help create a platform that gives more Australian enterprises access to sovereign local AI services. As it regulates AI, it needs to create an environment that enables Australian businesses to compete and succeed. Australia has a choice: we can primarily consume and host AI capability developed elsewhere, or we can use this moment to build meaningful capability of our own. Ultimately, this is all about value—value for the Australian people, Australian companies and the Australian government. If we get it right, we believe it will be the start of a new era of prosperity. If we don't, it will be a prosperous era, but not for Australians. That is the potential we see, and that's why we are so passionate about it. CHAIR: Thank you very much. I take from both of your opening remarks, and the submissions that you've made, the importance of sovereign capability and actually understanding and articulating what that means and what it should mean. How do we look at the elements of the stack that need that ownership or that definition of sovereign capability? What, in your view, do we absolutely need control over, in order to ensure, I think, some of the security elements and defence elements? But what is it that we need, in your view, to ensure that we have control over? Mr Morse : I'll approach that question in two parts. The first part is maybe a question back to the committee, which is: what use cases of AI do you want to have control over? There is a discussion in government at the moment about what kinds of use cases, fundamentally, Australia needs to have a greater level of control over, which comes to your question. In those instances, we do need to be higher up the value chain, in terms of the intellectual ownership of model weights, of training capability of a model that has got those specific use cases that will meet that, whether it's national security or cyber security need. In that context, it's about moving up that value chain within the AI stack, beyond the physical infrastructure and more into the intellectual knowledge based inputs, into that capability, and building a model that is going to do those specific-use cases that government and the Australian public believe are necessary for Australia's interests. CHAIR: You mention in your submission that South Korea is providing a useful model for building capability in higher values of the stack. Can you expand on that, in terms of the recommendations that could be made to the government about better supporting access. Mr Morse : Sure. The South Korean example that I refer to in the submission relates to government methodologies through which they are uplifting to build sovereign AI in that country. Europe is another example. Let me talk a little bit about the different models that these governments have deployed. In Europe, they have got behind a pan-national champion, Mistral, and they are supporting, through direct contributions, preferential procurement, government contracts to uplift that capability and the ecosystem that sits around that capability. The South Korean model is to apply the same procurement and other government uplift techniques but more in partnership with local industry—in South Korea's case, five different consortiums that are working to develop their own models. They are, as I understand them to be, general-purpose models—that is, large language models. In each instance, you've got a different method. Europe is a much bigger jurisdiction and one with much more money, but it is getting behind a single national champion in its ecosystem and supply chain. South Korea has an economy comparable in size to the Australian economy, and they have five different options that their government is working in partnership with industry to uplift. Mr Zalloua : When we look at the sovereign AI pieces, government workloads will obviously require significant sovereign capability. But I'd like to discuss how we need to look at it through a much broader lens. When we look at it through a lens of Australian enterprise and Australian businesses, if you look at the market today in which they are creating innovation, applications and these sorts of things, if we as a nation are not providing them with a platform in which they can build this innovation locally and control the intellectual property that they are producing, we are opening up our businesses to global players taking that IP and leveraging it for their own success. That is extremely important to us because we are seeing, time and time again, through the large frontier labs and those sorts of organisations who are using Australian data, creating competing applications and bringing it back to our market—so, if we don't support the Australian business community by providing them with the sovereign compute capability, the applications and the AI services here locally, we are going to lose more and more of that IP to global players. That's where we see the largest opportunity. Government has its requirements; we should be extending those requirements to Australian businesses to give them the protection they need so they can succeed. That is because, ultimately, we need business to succeed in order for us to sustain our current life right. CHAIR: A big part of the debate is about the extent to which these big hyperscalers are interested in coming here and interested in training on Australian shores. What does that mean for us as a country, in terms of our ability to leverage off of that investment and help shape it on our own terms? I am keen for you to reflect on the extent to which that is necessary and on what benefit that would bring to our sovereign capability and our ability to build out and invest in an Australian owned AI ecosystem here. Mr Zalloua : I'm going to try and not be as controversial as I want to be. If we look at the investment they're making, there's obviously an economic impact that it has on our communities. But, if we look at the hyperscalers coming in and taking large leases with data centre players, the bottom of the stack—the energy and the data centre—is probably valued at, to be generous, 25 per cent of the economic uplift it provides. As a nation, are we happy to take that 25 per cent and leave 75 per cent on the table to be shifted offshore? For me as an Australian, and for my children, I definitely don't think that is reasonable. However, Australia does have the capability and the ability to provide an enhanced service to these providers. We should be encouraging them to actually take more of our services rather than coming in and just taking our energy and then building the stack and reselling it to our businesses at a much higher price. There's also an opportunity here where we can find a nice balance. There are Australian organisations that can provide similar sorts of services. The hardest thing in this AI world is the amount of capital required to deliver these complex buildings or AI factories. So how do Australian businesses have an opportunity to compete on the global stage if they don't have the right capital support? It doesn't necessarily need to be in terms of hard cash, but support from government. We're working with a government agency currently which is a program in which they will support and underpin a certain amount of compute, and they will provide that compute to their portfolio companies in a compute-for-equity exchange. That is a great example of how government can support us as an organisation in securing capacity for Australian businesses and providing them a platform to scale and grow here locally. That is an example of where we could actually start to compete against these bigger hyperscalers locally by having that support. We need to stand up and grow our businesses, and our peers are doing a great job in doing that with international offtake. We're really focused on Australia because we see a risk here that, if we use up all the energy and we provide it to all the hyperscalers, Australian businesses have only one place to get their compute from, and that is via the international firms at a much higher rate, whereas we can provide the same level of service and the same complex solutions locally at a much cheaper rate, which effectively allows them to compete globally. That's a very, very important balance we need to find as a nation. CHAIR: Do you think we have an opportunity to do that? Mr Zalloua : I definitely do. I think we are doing great work with some departments in starting to unlock some of those things. In Western Australia, for example, we're pulling together a global consortium of partners who will come and provide the skills and expertise locally to local businesses and to government and also allow it to operate. So they design, build and operate, and then we transfer the ownership back to the Australians to run that and manage that. So there is a way in which we can find global expertise, bring them into country, upskill our people and then transfer the ownership to Australians and keep that going. That is a fundamental way in which we can start to help build our ecosystem locally but also balance the requirement of serving global tokenisation through the large-scale developments. Mr Morse : Could I respond to that? I agree with almost everything Mr Zalloua has said. The only challenge I would offer is that this should not be presented as a binary choice: all of Australia or all of non-Australia. We believe very firmly that the great power in our opportunity now is the middle path, which is both. By determining those areas that are critical to achieving sovereign control in AI, that is how the Australian government can carve that middle path. Here is where we're going to be having Australian sovereign AI, and, from that, we can expand and we can grow into maybe more of the kinds of applications that Mr Zalloua is describing. Over here, we're going to be forging deeper partnerships with global hyperscalers and global AI companies. They have a vast need for compute. Australia can help to fulfil that, particularly in the interests of their expansions into Asia-Pacific. We should not lose that opportunity. The only challenge I would make to earlier comments is that I think you were talking mostly about the economic benefit of AI outputs. There's significant economic benefit to be achieved by the inputs to AI, data centre construction and supply chains that go into supporting and servicing data centres, particularly of the scale that would be required for AI training. We shouldn't lose that opportunity as well. Certainly when we talk about the investment dollars that are being proposed to come into Australia for that expanding kind of operation the international AI companies are interested in, it's really the downstream industries that would benefit from that. Upstream, the intellectual property and the kinds of application capabilities that I think we both agree on—I think that exists more squarely in the sovereign AI discussion, where, again, Australia can carve out its operational jurisdictional and governance controls in the national interests. CHAIR: In your view, what would the potential opportunity cost be if, for whatever reason, the hyperscalers choose to invest elsewhere? Mr Morse : What would that mean to Australia in terms of economic benefit lost? CHAIR: Economic, capability—what's your sense of the cost there? Mr Morse : It's too much to ignore. Mr Zalloua : I think I'm just going to challenge you on that. When I say 'to step up the stack', I mean we can sell space and power or we can sell compute as an AI factory. It's two very distinct things, and I mentioned that through my opening. A data centre operator is going to sell space and power, and they want to sell it to a Microsoft who is investment grade. They can recycle their cash, they can get the lowest point of interest, and they can make the most money over a long period of time. That is the capital game that we're in. But if we are providing Microsoft the space, the power and the compute, we are earning a far larger portion of that pie. To say that Microsoft wouldn't take it as a compute resource is wrong because they're doing huge amounts of deals with the likes of the IRENs of the world in Europe and these sorts of things. So what I'm saying is—yes, you can be a data centre or we can be a country that's producing AI factories which actually provide a huge economic benefit beyond just selling space and power. So there's a very distinct difference between the two, and I think Australia should be focused on getting as far up the stack as possible so we can capture as much of that economic benefit as possible. CHAIR: I think it was a previous witness who used the title 'the decision in front of us'. Do we become just takers or makers? I thought that was a good way to describe, potentially, the choice that's in front of us if we're keen to take it. Mr Morse : I think that's right. To clarify, in our business we do cybersecurity, cloud computing, architecture and AI as a service as well. So we are up that stack, if that makes sense. So we're a little bit unusual, as a data centre provider as well. We're even more unusual because we've been building data centres for over 20 years. That informs our perspective such that we recognise that value-add of actually moving up to building things through the compute. But that informs my perspective, as I mentioned earlier, on that middle path that I think Australia should be carving. CHAIR: Deputy Chair. Senator DUNIAM: If the aim is to get as far up the stack as possible, which I think is an admirable effort—other witnesses have characterised it in different ways, of getting our fair share as a country et cetera. I think I speak on behalf of the committee when I say no-one here is a subject-matter expert. We don't come from the area, so there are a lot of unknown unknowns. I'll just cut to the chase and ask both of you—what would be the top three things we could recommend to government about what they should do in order to achieve this outcome? Mr Zalloua : The first thing is how government can support Australian businesses in providing them local access to compute and AI services. Senator DUNIAM: Do you have a model that provides that? Mr Zalloua : Yes, we do, and I can share that with you. That is one of the first and foremost because, as a country, we need to really level up our organisations and enterprises during this transition phase. We need to get them upskilled. That's what's going to carry the economic value long term for our country. The second is providing organisations like ours the ability to make large investments in compute knowing that we have support from governments in securing some of those financial metrics that are required to do that so that we can apportion some of that to Australian businesses. When I say that—to spend a billion dollars on compute, the bankers are going to require an investment-grade customer. There are only so many investment-grade customers globally, but there are huge amounts of AI native organisations, Australian startups and Australian businesses that aren't investment grade. They need access to this compute. How do we give them access to this compute if we're hamstrung by the mechanics of the capital required to deliver such large-scale projects? Having government come in play and support some of those initiatives is going to be a really good way in which we can start to provide equitable compute across the country and across our businesses. They're probably the two main priorities for me. The third—I'll have to think about that. Senator DUNIAM: Take it on notice. Mr Zalloua : Okay. I'll take that one on notice. Mr Morse : For me, it would be, first, to establish the use cases—the applications, the end product, the end benefit—that Australia sees as critical that we have sovereign AI to serve. That means all of the value-added benefits—security, governance or otherwise—that the Australian public can have faith and trust in. If it were a national security or a cybersecurity output, for example, then government should identify that. 'Australia is going to build sovereign AI for this and this,' and however many use cases it determines is going to be in the national interest. Use cases create natural ecosystems. You will have businesses, capabilities and researchers who will gravitate to those particular use cases, so the minute that you identify them you create clusters of industry experts and industry capabilities. That's the first thing I would suggest. Then, to catalyse that, government needs to get serious about preferential procurement, just as every other country on the planet that is looking into sovereign AI is doing—that is, to allow Australian businesses within those ecosystems access to opportunities where government can be a better customer to Australian businesses and, through that procurement, catalyse the expansion and scale-up of these capabilities. Ecosystems for this kind of conversation are necessarily large and complex. It would probably include academia. It would probably include industry. It would almost certainly include use cases in a national security context that are specific to government. But I think we've got the ability to stand up and have conversations around those ecosystems and how we can catalyse those. Senator DUNIAM: Thank you. CHAIR: Senator Pocock. Senator DAVID POCOCK: Thank you very much for your time here. I've got some questions to start for ResetData. Following on from the last witnesses, in your submission, you talk about one of my favourite topics: Commonwealth government procurement. You talk about the need to direct more of that to sovereign Australian AI. How do we make that happen? We heard a bit about the tax asymmetry and this mindset in procurement that the US companies are the best. How do we change that? What are some really practical things we could implement in procurement? Mr Zalloua : I'll start by saying the artificial intelligence journey is just beginning, so there is an opportunity where government can support Australian AI native organisations to upskill and give them the necessary—let's just call it—contractual ability to then go and invest and scale here, locally. A perfect example would be, say, an Australian-model builder like Trellis. Having a Trellis do government work on a long-term contract with government allows them to accelerate the requirement they need to continually improve and innovate their product offering here, locally. If you're asking for an organisation like a Trellis to do that on the global stage, it becomes very difficult, and the first thing they're going to do is pick up their business and move to the US so they can get the right funding to be able to do that. We're at the beginning of this AI journey. It's a unique opportunity, where government can work in a way to help these businesses secure a pipeline of work and let them build the capability. It doesn't necessarily need to be large scale. It can just be a start. We're doing great work with a government agency. We're using Heidi Health as our foundational partner for the Australian AI growth program which ResetData has put together and will be rolling out across the country. That AI growth program is a way in which we get in this instance the partner, which is Innovation Victoria—are going to fund compute for equity. They are working now with Heidi Health and ResetData to bring them into our ecosystem and keep their compute local. They're supporting them by funding the compute component for them for equity. That is an example of a great partnership where government has played a role in not only supporting an Australian business and serving Australian doctors and global doctors but keeping that compute, that sovereignty and that IP here, locally. It allows us, as a business, to bring in a great Australian AI native organisation and allow them to scale with us locally. That is a perfect example of where government can step in and provide the framework and the platform for these businesses to succeed. It doesn't necessarily need to be an AI workload from government per se. That's just a perfect example of how it could work. Senator DAVID POCOCK: And, to try and get those sorts of arrangements going across the board, do you think that the government should be looking at a percentage spend that should go to sovereign AI companies? Mr Zalloua : Sure. As big a percentage as possible would be my recommendation. Yes, there would be a percentage spend, I would say. But the Australian government has got a responsibility to its constituents to ensure that whatever it's doing it is doing for the benefit of its communities, and Australian businesses need to be able to support that. Obviously, if there are specific workloads and they can't, then they can't. But, where we are today in this journey, the opportunity is not today; it's the next 10 years. The opportunity is that the decisions we make today will form what the next 10 years look like. So, from a government perspective, we should be thinking about artificial intelligence over a longer period of time, not just an election cycle period or whatever it is. This is something that's foundationally going to shift Australia from what it is today to something completely different, and we should be carrying the cards as to what that future looks like and what that vision needs to be. If government does the right thing here, it will set a vision of what Australia looks like in 10 or 15 years and start the foundations today. Senator DAVID POCOCK: This is to both of you, if you want to weigh in. We see very serious money going to the likes of Microsoft and Amazon Web Services from government for government services. If you look at Copilot, it's the first port of call—'Let's just roll Copilot out, no competitive thing'—because we're locked into the Microsoft ecosystem despite the US CLOUD Act and some of the clunkiness of various parts of the tech. But are these capabilities that Australian companies could actually provide? How should we be thinking about this when we have this sort of lock-in at a government level? Mr Morse : I think the magic exists in the definition of what we want sovereign AI for. You led with your question alluding to frameworks that might be deployed to make it happen or whether they're being deployed and they're not making it happen. Cybersecurity can be designated as a national security interest, which exempts Australia from its obligations under our existing 18 FTAs. Senator DAVID POCOCK: Do you think it should be? Mr Morse : Well, clearly it could be and it should be and it is. It's obviously a very topical concern at the moment. But if the government was to determine that cybersecurity for government and essential services and critical infrastructure had national security implications, through that determination, we could conduct preferential procurement, and that could be the middle path for sovereign Australian AI to emerge. Senator DAVID POCOCK: That's great. Do you have anything to add? Mr Zalloua : If you take a step back, we're going through a transition of your traditional cloud to an accelerated cloud environment, and how we utilise artificial intelligence in the way government operates or the way businesses operate today is still being defined, so the easiest thing to do is to stick with what is known. But there are large numbers of organisations that are being created and huge amounts of money being funded to get them behind this new accelerated way of working. I think the decisions need to be made not on what we're seeing artificial intelligence doing today; it's really understanding what its potential is tomorrow. If we understand what that looks like, decisions can be better informed as to where we should align ourselves. Copilot is a great starting point, but is Copilot going to be the future of artificial intelligence? No. So you know you can keep laying your bed with Microsoft for as long as you need to, but, ultimately, what artificial intelligence does is it gives you flexibility. It gives you optionality on how you deliver your workflows, and that optionality is something that is uniquely sovereign because the more you can build local—the more you can harness here locally—the more you're keeping your intelligence here locally. And if you wanted to export that intelligence, you could. Senator DAVID POCOCK: On that point, in terms of thinking about the future and how we do that, have any Australian companies been offered the opportunity to provide the Department of Finance with artificial intelligence capability that is configured to process data exclusively onshore as part of Australia's GovAI build-out? Mr Morse : I'm not aware of any, but I'm probably the wrong person in our organisation to ask. In relation to that question but also following on from the previous one, the opportunity for—getting back to my earlier point—carving out that middle ground is going to exist in the context of the kinds of value-added services that Mr Zalloua has been talking about, insofar as the tender opportunities that, certainly, I've been aware of in the last 12 to 24 months from a federal government perspective have been more concerned with the output of AI rather than the inputs to it. I think, in order to build some sort of sovereign AI capability—getting back to my earlier point—in some kind of middle path, we're going to have to take a little bit of a step back and understand that we're actually building something that's going to have future application. It's maybe not going to solve today's problem tomorrow, but it's going to be building for a future capability that could have multiple applications and multiple use cases. Again using sovereign cybersecurity AI as an example, that could be a capability that could be applied in lots of different environments to secure our networks and secure our data. But it is going to take a change in mindset. The procurement piece at the moment is about an application that's AI enabled that's going to help me do X or achieve productivity measure Y. We may need to take a bit of a step back if we are going to get serious about sovereign AI, and procure to build something that's going to have future application. Mr Zalloua : To answer your question, Senator, the most I've seen is an expression of interest from the Department of Finance. I believe that I can take that on notice. Senator DAVID POCOCK: Yes, if you can, that would be great. Thank you. CHAIR: Senator Dolega. Senator DOLEGA: I might ask a different line of questions from my colleagues. In your opening statement you mentioned the benefits and economic opportunities for communities from data centres. What are some of the benefits for communities of having a data centre located in their region? Mr Morse : I'm happy to take that. If we build data centres, it's not going to be the same for every data centre and every region. The community benefit and the community need in every region is going to be very different. In our case, our major campus is in an innovation district, the Macquarie Park Innovation District. It's the largest technology park in Australia. It contributes over $14 billion to the national GDP. It is co-located with just about every technology company that is resident here in Australia. Universities, institutions, schools, biomed, research institutions—you name it; everything's there in Macquarie Park. So the natural extension for us is to partner with many of those institutions, which we have done, and we're working hand in glove with Macquarie University to deliver future research outputs and capabilities. The broader resident community benefit—again, that plays out in our context in terms of a partnership with United Way, and we're doing regular activities with local communities in terms of charitable sponsorships, such as books for children in underprivileged families. We've got different partnerships with schools and graduate programs to create pathways for local residents to work in our organisation and in technology more broadly. That's the context in which that plays out where we are, in our business. It may not necessarily be the same for everybody. But I think our experience has been informed by our longevity in that region; we understand that region very well. Similarly, we would invite other data centre operators to explore the same opportunities relevant to them in the areas in which they build and construct. Senator DOLEGA: How do you make those decisions? What share—if you know—of the revenue or profit that that data centre or that technology centre may create do you put back into the community? Mr Morse : I think maybe there are two parts to your question there. The revenue that the data centre creates, given it's a hub for every critical infrastructure sector— Senator DOLEGA: Or for Macquarie. Mr Morse : Or Macquarie putting back a percentage of revenues or profits into—we're an ASX listed company so that would be challenging. But we are informed by the value that the broader community brings to us in the decisions that we make in assessing opportunities to partner with local institutions, local schools and universities and local businesses. Senator DOLEGA: How do you go about making those decisions? Do you consult with local council? Do you consult with community groups? Mr Morse : That's a great question. We have longstanding relationships with local councillors, local institutions and local businesses which inform that decision-making process. Then we have an executive leadership team that will make those assessments as they come up. Senator DOLEGA: For ResetData, as far as I understand, you have your own dataset. You operate your own centre, but you also then have arrangements with Macquarie. You've noted in your submission you employ 30 people? Mr Zalloua : Yes. Within our submission—just to clarify, when we're talking about the data centres, we're referring to an AI factory. Senator DOLEGA: Yes, the AI factory. Mr Zalloua : The reason why we refer to it as an AI factory is there are significant benefits that the community can get. There are the economic benefits from providing the intelligence to the organisations that are surrounding those communities and bringing those businesses up. They're employing more people. We also look at what we can provide back to the research agencies. For example, when we built our first facility in Melbourne, which is an AI factory, we built it in a commercial building. We repurposed the space. It's highly sustainable. But that factory there, we actually gave control of that entire factory to the University of New South Wales to do some research work. The Australian Institute for Machine Learning did some research work, and, more importantly, we ran a competition in which the Children's Medical Research Institute, which is here locally, won. They got $1 million worth of credits to run their research work which accelerated the protein. I'm not technical when it comes to the scientific analogies around that. But what essentially came out of it is they accelerated their research by two years. Senator DOLEGA: That's incredible. Mr Zalloua : If we as a community, providing these sorts of services, provide access to the likes of the Children's Medical Research Institute, and they accelerate the cure for cancer, I don't think anyone in the community is going to be complaining about what that economic impact is or what that benefit to the community is. I see that as the reason why we need to be further up the stack and the reason why we need to share that equitable compute across the community, because that's what's going to elevate. Senator DOLEGA: This is part of building that social licence. People might say, 'Oh, we don't want a data centre.' But having a data centre actually is creating these opportunities for communities. I'd like to just find out a little bit more about the jobs. Say, in an AI factory, what's a job title? What's an average worker who might be in there, whether or not they're a technician—how much money do they earn? What's the salary of somebody? If you can't tell me now, I'm happy for you to take it on notice. Mr Zalloua : I can take it on notice. For example, in our facility in Melbourne, which is only 650 square metres—a uniquely small, very high-powered facility—we've got probably seven staff members on rotation. Within that you have network engineers, system engineers and mission-critical engineers in there. You have your general manager. You also have your security across that facility. That's a very small facility. It employs seven people on rotation. Some of those staff members are earning in excess of $150,000 each. With where the market's going, I anticipate there's going to be a feeding frenzy on network engineers and software engineers that are going to require us to keep raising the cost of those specialist employees. Senator DOLEGA: What about contractors for maintenance and everything like that? Mr Zalloua : In our facility, we would have six ongoing maintenance contracts across different types of specialist trades. Again, it's small. It's 650 square metres. I'd imagine something like Macquarie would have 30, 40 times that operating. Senator DOLEGA: The reason I'm asking is that obviously I have a tendency to be very biased towards my home state of Tasmania, which is also the deputy chair's home state, and we have AI factories looking at being built, and there's significant community concern. So what I'm trying to get is for communities to understand whether there are well-paying jobs that can be filled by locals who can upskill to get involved, as part of that social licence. Previous witnesses have described AI factories as a gold rush at the moment. That gold is not just going to line the pockets of a wealthy few; it's actually going to go back into the communities. We're potentially going to cure cancer by donating compute to be able to do this. It's not just putting a sticker on a local football club, even though that is important. There are going to be greater flow-on effects for communities. I'm not sure I got you. Mr Morse : It's a really good point. I think that at the heart of your question is that there are going to be opportunities for people who are not PhDs in computer science. Senator DOLEGA: Yes. Mr Morse: There are significant opportunities, particularly in the construction and maintenance of data centres, which require a lot of people with TAFE accreditations. Electricians and plumbers are critical to the operation, maintenance and construction of data centres. A need has been identified—and we've been part of a cohort to identify this—for a bridging certificate to allow plumbers and sparkies to actually have accreditations in both, because there's a real burning need for people who have that dual capability to understand how water and energy intersect in the operation of data centres, and that's certainly something that we've been communicating to TAFE. I think that, through that, there's potentially a whole line of employment opportunities that can emerge through that kind of accreditation. Senator DOLEGA: Yes. I might just pass the call back. CHAIR: Mr Hastie. Mr HASTIE: Thanks very much for your time today. I think most of the committee here has covered the more granular aspects of your submission, but I just want to touch upon the big picture, because part of the process of this committee is educative for the Australian people. There are the paragraphs where you talk about your central proposition that Australia is currently defining AI sovereignty too narrowly, and then there are the more important questions that you pose. Who controls the model weights, software, credentials and intellectual property? Which jurisdictions govern those assets? The larger strategic frame is that there's a contest at the moment between the United States and China and, being a middle power, we've got to make choices about where we partner, where we import technology and the models that we use. Do you have a view on security and jurisdiction questions in relation to those two countries? Mr Morse : I think Australia needs to walk and chew gum at the same time. As I've mentioned, where we need to have sovereign AI, we should pursue it and, where we don't necessarily need that, we can continue to pursue and foster the relationships that we have currently with our security guarantor and international partners. It's not even a binary choice of just the US or China that's emerging. Europe has a very capable model that is growing in capability all of the time, and we mentioned that South Koreans and others are also pursuing similar capabilities, so these will emerge over time. The question is whether Australia wants to have its own capability that can compete at that level—perhaps not at the frontier level, or perhaps it can at the frontier level, but certainly a capable model that we can rely on and have sovereign control over. Mr HASTIE: As the private sector makes choices and builds that sovereign capability, it's a global marketplace, right? There's the United States. There's China. There's Europe. Choices have to be made. I've referenced it before, but the decision in 2018 or 2019 about our 5G core infrastructure—the decision on Huawei—seems quite simple compared to the challenge that we have now, where potentially you've got thousands if not millions of people making choices which have the same sort of implication on AI. I just want to make it clear that there's a lot at stake here for our security, whether it be in government, the private sector, critical infrastructure, banking or telecommunications and whatnot. I suspect that you guys, being at the cutting edge, are seeing that firsthand. Mr Morse : Absolutely. For transparency, we have significant work and partnership with federal and state government agencies. I mentioned that in my opening. We'll be guided by government on where those geopolitical and security concerns reside. Recent examples of where governments have withheld a capability because it's been at that bleeding edge of frontier capability are obviously well known, and the committee will be familiar with those. We've had concerns that those types of frontier capabilities could one day be determined to be ITAR—international trade and arms restriction—which would make them difficult to obtain for non-defence-related companies or, certainly, non-allied countries. I think that risk remains. It would be entirely justified for the US to do that. It's within their jurisdictional control to do that. I think that should inform our approach to that middle path: what does the Australian government have within its power and jurisdictional control to direct, to govern and to give us that balance of capability that we need to protect ourselves? Mr HASTIE: Is it fair to say that the private sector is looking to government for strategic direction around trusted vendors and preferred providers as well? Mr Morse : I think they've already provided that, so perhaps less so. It's a matter of public record. With the companies that the federal government is partnering with in the AI space—again, going back to my earlier comments—I think what's missing there is that third way, that Australian way. We would certainly advocate that we not only continue on the path that we are on but also add that third way so that Australian critical infrastructure organisations can have confidence in the choices that they're making. Mr HASTIE: If we were to, say, cut our submarine cables today, how much sovereign AI capability would we have? Mr Morse : We don't have any large language models or model capabilities that I'm aware of today, so we'd have to get cracking. Mr HASTIE: Foundational models? Mr Morse : No. Mr Zalloua : I'd suggest there is an open-source marketplace where you would find that there would be a fair amount of Australian businesses starting to explore what that open-source marketplace looks like and how they can leverage the open-weight models and start to create their own bespoke models that deliver very specific workloads. Do we have frontier-level models here locally? No. I think what you'll find is that a lot of these frontier models are starting to equal out and a lot of the open-source models are starting to catch up. What I envisage will likely happen is, once that equilibrium is hit, a lot of their frontier models will turn into applications that they're going to bring to market. It's at the application layer that we probably need to be most wary as a country. But, from a ResetData perspective, we've laid our bet. Our bet's on Australia and making sure that we align ourselves with the Australian allies moving forward, because we do see a risk when it comes to AI and the impact that will have on national security. That's a bet we've made. We're focused solely on Australia first, but there is significant risk around either/or taking their models in country. Mr HASTIE: Firstly, it's been argued that the through-line to Australian sovereign capability in AI is through US frontier labs. I'm interested in your view on that. Secondly, how are you going with workforce? How competitive is it out there to get Australians who are specialists in this field? Are you recruiting from, say, the United States, bringing people back, and what does that cost? Mr Morse : We don't build AI models at Macquarie, so I might defer to Michael on that point. Mr Zalloua : It's Marcel, but I'll forgive you this time! We're not building models per se. You would have seen Nvidia's purchase of Hugging Face. Hugging Face is a platform that has over a million open-source models. There are models coming out of South Korea. There are models coming out of other jurisdictions that can be brought into country. You can fine-tune them and retrain them. You can do lots of things. We've got enough resources in that capability around the Australian Institute for Machine Learning. Within ResetData, we've got our own centre of excellence with PhD students who can retrain some of these models with global support. Are we competing for talent with the US? No. We couldn't afford to. That's not happening. If anything, we're probably losing talent to the US and to other markets. But is there enough of a talent pool here locally? I think if the government supports that ecosystem and provides that sort of platform that allows the startup community to prosper and these sorts of things to prosper then, yes, we can develop these skills and we can develop them locally. Traditionally, Australia punches well above its weight. There's no reason we can't do that on the AI front. Mr Morse : I might offer a clarification, because I think we're using slightly different terminology. At Macquarie we don't build models from the ground up, but the agentic AI capability of building application-layer AI—we are doing that extensively in the cybersecurity space. In terms of your question on talent, Australia has deep talent in that capability, deep cybersecurity talent. Do we need more? We always need more, and the extant threat is certainly growing. I'd refer the committee to ASD reports, which will bear that out. For the last five years, we've seen the number of cybersecurity incidents grow exponentially, and, in many cases, they are AI enabled. So I think we have enough latent talent here in Australia—and, more importantly, experience—to be able to build out deeper, more expansive AI capability to support cybersecurity and national security interests. Absolutely, we do. And they're all over Australia. CHAIR: Thank you very much for providing the evidence you've given us today and for answering our questions. If you have anything further, please engage directly with the secretariat. Proceedings suspended from 12:06 to 12:51 .