Interviews > ExtraMile Video series

So, it is a fair question because we have a lot of companies. I often kind of think, to be clear, not that I’m comparing myself to Elon Musk, but you have to look at Elon Musk as well, right? He’s the most standout example of someone who’s somehow able to do multiple things all at once. And I think the answer is actually because I think this is actually a very good question. The answer is it depends on the situation. So my current role, which is managing a large group of assets, requires actually having a bunch of very, very capable CEOs running different companies and then overseeing and supporting those CEOs. And then there’s obviously a CEO role, which is actually managing all of that. And that’s my role. Really, in this instance, the thing that makes the difference is those individual CEOs.

And I don’t think it actually think that would be true if I was in a standalone company. It would be the same thing. So, for instance, if I was CEO of one AI company, then what I would really care about is scale.

And I get that scale from selecting very, very strong leaders below me who can then manage. And I think it really comes to the same thing at the end of the day. I’m supporting and selecting and working with a set of leaders that are in their own right, determined to deliver and on the hook to deliver specific outcomes for those businesses.

And if they do their job right, then everything works. It’s actually in the selection of those people and the management of those people that makes the difference. The expectations, making sure everybody’s aligned on what’s supposed to happen and also being there when they need me.

And actually, if you take that in this context that I am right now, where CEO of a family office and multiple businesses, and then you take it to just say a standalone company, it really comes to the same thing. So, I actually think what we’re talking about is leadership and management at its core, whether it’s this construct or another one. And for me, being really explicit about choosing people, culture is really important.

Like you can get all of the structure right and have a terrible culture that doesn’t allow you to move forward. And that’s really going to be based on trust and then expectation management. It’s whether I’m the chairman of a company, I’m working with a CEO or I’m the CEO of a company, I’m working with the head of department, it comes to the same thing. What do I expect of you? What do you expect of me?

I tend to find if we get those basics right, everything seems to work as long as you’ve got the right people. But ultimately, this is the challenge of scale. All of this is a long-winded way of saying scale.

And I learned that very early on in my career that the people who get it wrong is where they try to keep things themselves. What you really need to do is to let everybody do everything for you. That’s the ultimate scale.

And scale is one of the most powerful competitive advantages you can have.

Host: Yeah, that’s a strategic approach, I must say, especially when you are operating across different responsibilities. So why did you choose TensorX to join as its Executive Chairman? And what specific customer problem convinced you that Europe needed a sovereign AI infrastructure company like TensorX?

Tim: Well, you know, the way that we’ve built companies over the last few years is we usually tend to look at adjacencies of one company that we’ve already built. And then we see what are we learning from that company where there’s maybe a product, a service, a capability that doesn’t exist that would make that company work better. And so, this is a very, very black and white, very clear example of that.

So, I mentioned earlier, you know, Apex, I am Executive Chair of that as well. CEO Usman Khan is a gifted technologist. And we’ve been working together for six years, building ultimately a suite of AI capabilities for capital markets.

We sell into large regulated financial institutions. And about a year ago, maybe more than a year ago, we started thinking about sovereignty as a concept because we’re working in regulated industries. We know what procurement looks like.

We know that what matters to regulated financial services players is usually sovereignty is very, very top of the list in terms of data protection and control. And it’s usually a matter of, again, regulated or policy driven reality. Like it’s non-negotiable that you have to do this.

And we thought kind of before this sovereign AI wave started to happen in 2026, we started thinking about it in 2025. And we realized that this was going to happen, that we were selling our Alice product into financial services players. And these players are using this product.

But then the question would quickly become, well, what if I use this here? Am I sending my data to a public domain? Am I, is my data being trained to be used to train a model?

Do I have full ownership and control of that data? And we knew we could see this trend coming a year ago. And so really TensorX was born of the knowledge that that trend was coming.

And so here we are fast forward to today. We now have deployed GPUs and we’re selling inference as a service. And indeed, it is being used by Apex, which is then being used by those end clients.

So, we proved the concept over that year. We built it, we made it work. And that’s really the most powerful way to build another company is because you know you need it because you built something else that knows you need it.

And so, we, in a way, we were lucky. I think we were prescient to think ahead. But also, we were lucky to kind of be already at the coalface of procurement and decision making with large financials that we knew that this was going to come up.

And so, we solved it ourselves. And sure enough, one of the biggest trends in 2026 is sovereign. But hey, we got ahead of it. So that’s how we ended up being in TensorX.

Host: Yeah, interesting. And that brings us to the challenge enterprises face when they actually try to adopt AI. So, for regulated companies, security makes AI adoption lengthy. So where should leaders differentiate between necessary controls and unnecessary friction?

Tim: Yeah, look, I think the filter play out. We’re in an early phase of a global transformation. Having been involved in other transformations, I’ve been very much in the digital assets space for the last decade.

And that’s been an interesting evolution to see it now become very institutionalized. And the biggest asset, the biggest players in the world are now engaging. But it takes time.

But this is different. That was almost a little bit more niche-y in terms of how it’s approached. This is so broad and so pervasive in terms of how it’s going to affect us all as people, as organizations and as governments.

We’re very early in a very large-scale shift. What I think is really interesting is that about AI is it’s not widget sales. You’re not buying something off the shelf. It’s not high volume, low margin sales. This is low volume, high margin. This is we need to engage and understand how AI is going to help your business.

And I think the big light bulb moment for me was hearing that McKinsey was almost bought by Anthropic. It’s a really interesting data point for me of why would that be true. And there’s a paper actually that was done, I believe it was by Sequoia back in March, which was really around what do the next trillion-dollar companies look like.

And basically they’re saying that it’s more likely to be a services business wrapped in an AI business. And so, you ask yourself the question, why would Anthropic nearly buy McKinsey? And why are Sequoia one of the most impressive PE firms in the world?

Why are they saying this? And you realize that the reason that everyone’s talking in these terms is because this isn’t widget sales. This is actually organizational change.

If you’re any sized company, you’re likely to have organizational questions about expenditure, about your resources, and maybe even about precisely how you do your business. You’re not changing things on the margin. You’re actually changing things a little bit more significantly than that.

And that requires a different conversation. That requires harder decisions to be made. It requires stakeholders to understand what’s going on.

And sometimes it’s going to require some bold moves. It’s going to have to require leaders to just say, right, I’m going to take the leap and I’m going to do this, even though I might not necessarily completely understand where this is all going. And I think when you put it like that, you know that a lot of people are going to be hesitant.

You just know that that’s human nature. You know, the classic adoption cycle of the pioneers, the early adopters, the early majority, the late majority, like that is going to play out. So, there’s most people are going to be like the half of the world is going to be late.

And so, what that tells us, I think, is that engagement in sovereignty as the key pillar of that is going to be really important. And therefore, understanding what sovereignty means is going to be really important. And I don’t think there’s an agreed definition of sovereignty.

Right. I think you could ask different people what sovereignty means and they might say different things. And again, the truth is it’s not one size fits all.

If I’m a health care company, I might need sovereignty might mean something else than an education firm that might need for a financial services firm that might need for a manufacturing firm. It really depends. And also, it depends on local regulations, which we’ll talk about as well.

The idea of sovereignty, the danger of sovereignty, I think, and this really relates to adoption is that people aren’t quite sure what it means. And I know for a fact that there are companies out there who are claiming sovereignty. And I would say some of the largest companies in the space are claiming sovereignty.

But actually, it’s not verifiable. And that wouldn’t be good enough for certain regulatory purposes. Right.

And so, to me, this is the key is what is sovereignty? What does it mean to you? How do you get it?

The fact that that that, you know, it’s not commoditized. That’s the thing. It’s not commoditized.

And so therein lies both the challenge, because it means it’s going to have a lot of work over years to do this. But for us, therein lies the opportunity. That’s the key for us.

Host: Yeah, that was truly fascinating to know. So, moving ahead, data sovereignty sounds simple, but is actually complex to understand and deploy. What has been the hardest to make work in real world data sovereignty deployments?

Tim: Yeah, I think the challenge. So actually, let’s put it this way. The core idea of ensuring that no data and left on the system, zero data retention, and ensuring that the journey from the beginning to the end is validated and verifiable.

That’s the challenging bit, because you can say that it is, but then you have to be able to show that it is right. That’s the key. You can’t just say it.

You have to show it. I think we’re still at the early stages of evolving into a very clear understanding, as we were talking about just a minute ago, about what sovereignty means. So, if you’re not sure what sovereignty means, then it’s different things to different people.

I think we’ve gone about it a little bit like the way that Apple did. Apple’s got, I think, a good reputation in terms of data and privacy, and it’s been laid that out. And it’s really done it by being quite sort of almost brutal to the extent of like, well, we definitely don’t even see or understand your data.

And I think that’s the angle that we’ve gone for. We can show you with certainty end to end. And I think the challenging bit is just being able to show that it’s analytics.

It’s the dashboards and the verifiability that really matter. Actually, doing it end to end, we can do. It’s the part that’s, can we show it to be true?

I think that’s the part that’s the challenge. And I think with time, there will be certifications. There will be certain things that companies need certainty.

If I have that rubber stamp, I know it’s sovereign. And at some point, we hope, actually, this is where regulation and government directives, et cetera, can be very useful, because it creates a standardized language that we can all say, OK, well, it applies to that. That’s relevant.

OK, now I can put that on my product. It’s XYZ compliant. The summary is it’s less about the actually making it cases. It’s how do you show it? I think that’s the challenge.

Host: So, it’s not simply about where data sits. It’s about the entire infrastructure around it.

Tim: Yeah. And I think the other thing is it can’t be, we’ve seen this a lot when you look at it in the context of the AI application that we might deliver. But then you have to look beyond that.

Like, what are you integrating to? Are there sensitivities that are being introduced because you’re integrating to someone else? We’ve seen situations where the actual AI solution itself is compliant and it is sovereign.

But then because it was connected to something else, and that wasn’t anything to do with us. So, this back to that core point of this is organizational change. It’s re-architecturing sometimes. And that takes time and effort.

Host: And that was truly a holistic perspective. So next up, TensorX announced an 8-million-euro seed round investment in NVIDIA. Blackwell GPUs in June 2026. How will this step advance TensorX’s sovereign AI inference platform and secure Europe’s AI infrastructure?

Tim: Well, when you think about it, I like to really simplify things as much as possible. And for us, ultimately, this is, can you get machines? Can you get GPUs?

Can you deploy them? Can you operate them? Can you operate them at capacity?

And then can you tune the models on top to be delivering inference? So, you could just sell GPUs as a service. You could just sell that.

What we’re focused on really is the inference layer. We’re really an inference cloud at this point. And so that’s the two things we need to do.

We need to be able to buy and deploy the machines. And we need to be able to tune the models and have them reliably be used by others. And in all states of the world, for a Neo cloud, which is what we are, like the purchasing of the GPUs is the starting point.

What we’ve learned though, and this is, you know, there’s an interesting shift that you’ve seen that maybe not everybody’s aware of. If you’re in the digital asset space, you’re probably aware of this, but if you’re not, you might not be. A lot of the big data centers that were set up over the last 10 years for Bitcoin mining and crypto mining, which are really big installations, very material, significantly sized facilities all over the world that have moved around.

When we were working on that, and in my previous roles, we’ve been looking at those things, you realize that, you know, it’s not as simple as just buying the GPUs and then just plugging them in, right? Like this is actually a very specialist hardware engineering to be able to take it, optimize it, make sure it works, know how to debug it, then tune it, and then make sure it connects and to enter the network and work flawlessly like that is actually, that’s the IP that’s hard. So, the critical step we’ve now made is that we’ve bought our initial GPUs.

We’ve received them from Dell. We love working with Dell as a partner. We’ve installed them and we’ve gone through the hard work of actually making sure they work.

And now they’re up there. And that was really the big critical starting point for us is that we could do that. Now it turns into a different problem because we’ve actually, and you know, we will keep learning and we will keep, you know, as models become more complex, as the space evolves, we’ll have to keep evolving with it.

But the core idea of taking a GPU, putting it in place and making it work, we can do that. Now the problem actually switches to how many GPUs can you get? And so that shifts the problem to a hardware and a sort of an operational execution problem.

And it’s a financing problem. Actually, the world, the large hyperscalers out there are all borrowing or raising huge amounts of money, right? That side of the market has got plenty of funding.

And usually if I’m a big hyperscaler cloud, then I probably sold all my capacity before I’ve even built it. Some large corporate. Okay, that’s easy to finance. I have a contract to take the off take. It’s a very nice business if you can get it. We’re not in that space.

We actually have to build a client base and then go and buy more machines. And there’s actually I think one of the critical differentiators between the cloud providers over the coming weeks, months and years is going to be their ability to raise capital. It actually becomes a capex problem.

It’s like, can I raise it? Can I get a margin on what I’m borrowing to buy the machines? Can I make sure that they’re deployed at capacity and that they’re generating a yield? This is the steady state we’re going to get to. So, we’re very excited because we made that first investment. We’re now we’re up and running.

We’re a real company. We’re generating revenue and generate, by the way, revenue is moving really, really quickly, which is what we want. Now we’ve got a new problem, which is how do we finance the next 100 millions of GPUs?

Host: Yeah, that gives a clearer picture of why hardware investment matters beyond raw computing power. So, moving further, TensorX uses zero data retention as a core principle. What technical or operational decisions were necessary to make this promise credible?

Tim: Well, I think so. The good news is we’ve got and we’ll talk about it in a moment. We’ve got in certain jurisdictions, regulations and direct that again, allow us to have a common language, but that’s relatively new.

And when we started talking about this before, we realized that we had to engage with our clients on this topic first. Like, what do they expect? And it was really interesting to talk to them and get feedback from our clients.

Like when we first started talking about sovereignty, win an AI context, it was very uncertain what we meant. What are we really talking about? Why is it important?

And then you explain it in the context of any financial services company, but the one thing they can’t do is have data leave the place, right? That’s kind of a very core element. You get fined for that stuff in regulated financial services business.

Suddenly that brings out the sharpened view of, oh, wow. OK, so if I send my data to Claude, that’s probably being used to train a model. Oh, but I’ve signed something with Claude Enterprise that says that they’re not going to do that.

But can you verify that? And is that going to be sufficient for your regulator? Or is it just the fact that it’s contractually said that like that?

Suddenly they realize, oh, that’s not going to be sufficient. I’m going to need some sort of validation. And again, this really comes back to what we were talking about before.

It’s one thing to talk about sovereignty. It’s another thing to prove that you’re doing it. And I think for us, the critical element is first define sovereignty in every state, because it’s never the same twice.

And to show how you’re going to confirm that it’s true, that is indeed sovereign. And again, this is the challenge for us is consistently being able to show that in a very powerful way. That’s been our journey there. Talk to the client, understand the sovereignty, show that it’s doing that. Wash, rinse, repeat.

Host: Yeah, and I think that’s why transparency becomes more relevant as European AI regulations continues to evolve. Speaking of that regulatory ecosystem, Article 50 of the EU AI Act has now started applying its transparency obligations across AI providers and deployers. How do you think this will change how AI is being used and implemented across business operations?

Tim: Again, I’ll mention the digital asset space, which I’ve been in for 11 years. And at the beginning, one of the biggest challenges that we would talk about was technology. And does it work?

Well, actually, very quickly, over a period of a few years, it was demonstrable that it does work. So, the technology resilience wasn’t so much of a problem. And I think we’ve reached that point already in AI.

There’s a decent understanding of it might be early, but it looks real and it looks productive. And it doesn’t look like the whole thing is hyped. So, we’ve now got, we don’t need to sit and worry, does the tech work or not, right?

That’s the sort of first question. Then you come to a second question, which is usually around regulation. And in the absence of that, especially in regulated industries, it’s really difficult because who knows the real answer.

Like you want something referenceable, again, this common language. And what we’ve now seen in pretty much every jurisdiction in the digital asset space is there’s some sort of either coming or it’s already there. Like look at Clarity in the US if you follow this space.

Like these are very big, big pieces of legislation that have been flowing through various jurisdictions. Great. We are now at an earlier stage of that here with AI.

Like not every jurisdiction has really developed and formalized its approach. It’s not surprising because it’s to keep up. I think the EU 50 Act, that is a very important piece of the puzzle. At least it gives us something government driven that we can talk to, right? Okay. This is now in force.

August 2nd, this isn’t perspective. This is actually now something that anyone using AI is supposed to look at this and make sure that they are adhering to what this is about. And this has pretty important pieces of the puzzle around what you can do and what you can’t do.

And I think that’s really to be welcomed. The thing is, it’s only one step. So, the upside is that it is a step. And the downside is that it’s only one step in one jurisdiction. Like we really need this in all of them. So, I think regulation is going to be a constant theme for a couple of years.

And then it will start to subside, which is what happened in Digilat. Everyone will start to feel comfortable about definitions and direction of travel. And then the one thing we’ll end up with, once tech is established and regulation is established, it’s all going to be about mindset.

It’s going to be about people’s ability to engage in it. But right now we’re still in that regulatory. And Article 50, I think is a very welcome step. Very incomplete though. It really doesn’t answer everything. But at least it gives us something we can talk to about our clients and we can get very specific.

And we’ll take the more of that, the better.

Host: Indeed, new guidelines could have a real impact on how companies design AI systems, you know, from the beginning. So lastly, while excelling in EU sovereign AI deployments, does TensorX have any plans to expand across the globe in the coming years? And which challenges do you think are unavoidable for global expansion?

Tim: Yeah, that’s a great question. And in fact, it really is kind of the aggregate of all the questions we’ve asked in a way. It’s like, well, if you believe in sovereignty as being important, the core ability that you’re selling is indeed sovereignty, then it can’t be limited to one jurisdiction.

It has to be every jurisdiction is going to need it. The EU is a good place to start because of where it happens to be in its evolution and where it happens to be in terms of where clearly the EU is ahead of the rest of the world. It’s already been quite a high standard for things like GDPR and any data related legislation, which makes it kind of a gold standard.

If you can do it here, you can probably do it anywhere. And so that was our logic. Can we deliver that in Europe where there’s a clear demand, but also, I think better specificity, but also, I think a very clear indication.

Once you’ve got that, you’re in a strong position to do it elsewhere. So, we’re definitely going to go to other jurisdictions. I think the US is a non-negotiable, but it’s the largest, deepest capital market in the world.

And therefore, it’s going to have the largest demand, and it’s going to have a lot of people are going to need sovereignty one way or another. And I just don’t think it accrues to the largest players. I think the scope for niche players to work in different industries.

But then we have to think about other jurisdictions too. The Middle East, notwithstanding the challenges around the military action around there, like pre that, I think we were walking towards very interesting engagements in the Middle East. We actually have some Middle East clients as well.

So, I think that’s an area that we would want to also move to. And then you’ve got the other sort of the two wings. You’ve got the Asian markets, which are all quite idiosyncratic.

You’ve got the South American market. So, I think we will look at all of the above. I think we’ll tend to work with partners when we go to these new jurisdictions.

We now have the IP to be able to take some machines, do everything we’re doing, who are we going to work with in that jurisdiction? But I would anticipate us going into the US and the Middle East in a matter of a relatively short period of time and definitely going to go global for sure. And we’ll also think about other capabilities that we need to bring in because I think the more full stack view of the world is going to be the thing that makes the difference.

Host: Yeah, perhaps that is what makes global sovereign AI particularly interesting as a market. So, thank you, Tim. This has been a fascinating conversation.

We have covered leadership, sovereign infrastructure, security regulation, and where AI infrastructure is heading. So, thank you for joining us. It was indeed a pleasure to host you. Thank you.

Tim: Thank you. I really appreciate the time, Rittika.

Host: And to our audience, thank you for being a part of this episode of ExtraMile by SecureITWorld. We’ll be back with another expert assessing leading tech and cybersecurity trends. Until then, stay informed and stay vigilant.