Interview with Josh Joshi, Executive Chairman at AtlasEdge
Josh Joshi, Executive Chairman at AtlasEdge, analyzes the impact of AI and developments in the fast-growing digital infrastructure field.
Sep 14, 2023
As part of our interview series featuring senior leaders with extensive experience in the debt markets, we are delighted to speak with Josh Joshi, Executive Chairman of AtlasEdge and an Operating Partner at DigitalBridge. He has over 20 years of experience building value in the digital infrastructure sector and has held senior positions at several multinational companies.
Prior to joining DigitalBridge, Josh served for over a decade as CFO of Interxion, a leading provider of carrier- and cloud-neutral data center solutions across EMEA markets, which was acquired by Digital Realty in 2019. He also previously served as CFO of TeleCity plc, a pan European carrier-neutral data center business, and co-founded and served as CFO of Storm Telecommunications Limited, a private-equity-backed U.S. and pan-European voice, data and network service provider. Earlier in his career, he spent eight years in professional practice as an accountant, primarily with Arthur Andersen.
Josh holds a bachelor’s degree in civil engineering from Imperial College, London and is a Fellow of the Institute of Chartered Accountants in England and Wales.
Josh, thank you so much for taking the time to speak with us today. I hope we’ll have the chance to examine several topics related to corporate finance, AI, and the rapidly growing data center industry.
To start us off, can you please describe your role as executive chairman at AtlasEdge? I suppose the role of executive chairman might be somewhat comparable to the executive producer of a film. It’s not always immediately apparent to those outside the industry what it is they do – and the role can also vary quite a bit, depending on the film or company.
The best way to term my role as executive chairman is a general fixer.
However, before reaching this point, I spent a lot of time in the early days helping shape the company's direction and assemble what, in our case, has turned out to be an extraordinary team. As AtlasEdge grows and progresses, I apply challenges to the team, and we work our way together through the various issues of the day and any roadblocks that might arise when tackling those challenges. The key to the EC role involves acting as both an insider and an outsider, stepping in and engaging in a dynamic way when needed, and at other times serving as a sounding board on more strategic, high-level issues.
Finally, it might help if I quickly contrast the EC and chairman functions, since the two often get confused. The latter is chiefly concerned with board matters, while the former plays a much more hands-on role.
You just touched upon the strategic aspect of the position. Can you perhaps go into detail about how strategy applies to the data center world? Data centers are generally shoehorned into the infrastructure sector. How or where does strategy come into play in this context?
Well, what's interesting is that there are two components to data center businesses. One component is strategic and another which is more concerned with civil engineering and the sorts of issues we might typically associate with infrastructure.
Strategy in the data center context is focused on understanding developments in the highly dynamic marketplace that our customers operate in. And our customers encompass the big platform companies, network companies, content delivery networks, as well as enterprises trying to move to the cloud and incorporate AI. We really need to understand each customer so we can provide them with a proper, curated physical environment.
Meanwhile, the civil engineering aspect focuses on building power and cooling infrastructure. We have, for example, HVAC engineers - extremely talented people - that have backgrounds in running nuclear facilities.
So, on the one hand, I pay a lot of attention to the strategic aspect of understanding what customers are looking for. However, across the board, I also need to make sure we're staying nimble, allocating capital correctly, installing efficient processes, and helping us to avoid pitfalls that could take place in any business, whether infrastructure or non-infrastructure.
I'm curious about the transition from your prior CFO roles.
As executive chairman, you’re no longer confined to one function, like finance. Instead, you’re exposed to the full set of corporate divisions, including marketing, sales, and HR, among others.
Was there a certain complexity, or difficulty, about the work involved in these non-finance areas that you gained a new appreciation for?
Historically, CFOs are often the ones challenging every area of the business – in relation to costs or hitting budgets. So, for me, it’s vitally important that CFOs have a good understanding of what they’re challenging - including critical assignments like talking to and engaging with customers.
In my experience working at start-ups with smaller teams – I’ve been fortunate to have had a hand in strategy that involved sales, HR, legal, and other departments. And this background has helped me appreciate that understanding what customers need is critical. It’s taken a while to develop both the appreciation for customer interaction and the aptitude to be able to carry it out. And it certainly wasn’t something I learned overnight but had to learn gradually as I became a more senior CFO and then spent more time here at AtlasEdge.
I’d like to turn to the process of raising funds in the capital markets. You’ve helped shepherd through more than a few bond issuances in your career. Can you tell us about some of the challenges you faced?
Well, in my previous life at Interxion as CFO, we were repeat issuers of bonds. I had left it to better minds than myself to deal with the more transactional details of bond issuance. But the bit that I found really hard – and where I devoted most of my energy – was in educating and spending time with credit investors to help them understand the investment thesis behind data centers.
When I started, data centers weren't considered on their unique merits, or as a separate investment category. They were put in a box and lumped together with shopping centers and other components of the real estate sector.
I believe we issued the very first European data center bond. And so, while they were certainly poorly understood at the time, that lack of understanding persists today to a degree. Although awareness of the market dynamics that support data centers is now growing, many people don’t fully appreciate how fundamentally extraordinary they are – and how they differ from traditional real estate and infrastructure plays.
Data centers provide investors access to an investment that is at the very foundation of online activity. It allows you to be agnostic regarding the popularity or success of any of the particular apps that operate above the base layer, and instead benefit from the general – and massive – growth profile of all things internet.
It must be getting a bit easier?
Well, something important has changed and it’s made my job – making the case for data centers – far easier.
The industry – and the firms I’ve been involved with – now have a track record to point to, and we can demonstrate remarkably steadfast returns, in line with the uninterrupted growth in online activity.
Through multiple crises and severe credit issues that have wracked the European and global markets, data centers have continued to deliver consistent, positive results. Just looking at Interxion, we sequentially increased revenue and EBITDA every quarter without fail during my eleven years there. That it was never shaken by wider market turmoil that affected many of the customers who used our facilities speaks to the unique combination of growth and stability that data center investments can provide.
I’m hoping you could give us an insider’s view of the finance function and specifically how you think AI might affect internal finance operations. Every department in the enterprise will need to wrestle with AI. What are some of the issues particular to finance? Will the impact vary across the sub-divisions within finance?
I'm really excited by what AI can do and the potential ways we can harness it within a finance environment. It will have an impact across the board in audit, and all the way through to tax and treasury. However, because many of those areas are process-oriented, I think we can probably already make out the contours of the benefits, or efficiencies, we can expect to realize.
In contrast, FP&A is the area I’m really interested in. It’s where I think AI has greater potential to move the needle in terms of augmenting our strategic thinking and allowing us to approach things differently.
Having said that, we must stay mindful of the dangers or pitfalls AI could open up. One thing I learned early on as an accountant / CFO is to never believe my own spreadsheets; it’s also necessary to step back and apply common sense. We mustn’t defer to AI as the solution to every problem. It won’t be. We will lose something valuable if human intervention and critical thinking are put to the side.
In terms of AI’s wider impact on AtlasEdge beyond the finance function, can you describe your mood: more apprehensive or excited?
Excited off the charts.
There is a tsunami of demand on its way to data centers.
While some might argue that AI will compete with data centers, I believe the two are symbiotic and will work together.
Let's go back to first principles and break down what we – as a data center business – are trying to do and what AI is doing.
We ingest inputs from our customers and use silicon and transistors to generate outputs.
All those bits of silicon need homes - and those homes are data centers.
AI is creating this massive demand for compute architecture – and critically, that architecture needs to be on, operating effectively on a 24/7 basis.
The only home for AI to reside; the only plumbing or engine that will make it work is in a data center.
So, there’s a symbiotic relationship that is going to be very interesting to watch over the next 5 to 10 years.
Of course, we’ve seen this movie before. Back around 2008-11, the concept of the cloud started to emerge in full force. Everyone was thinking about the cloud and what they were going to do with it. As I mentioned earlier, during my time at Interxion, we issued the very first data center bond deal in Europe in 2011 and it was carried out to fund a cloud deployment. Since then, cloud computing has exploded - but I see AI exhibiting exponentially greater demand.
Well, you’re on the frontlines. Can you give us a bit more perspective on that? Why do you think AI will so markedly outstrip the cloud in terms of the demand it places on distributed computing?
There are more than a few reasons, but I’ll point to two quickly.
First, as enterprises move to the cloud, it doesn’t necessarily create net new demand for compute architecture. The cloud is about taking existing computer workload that runs in an enterprise’s computer room and transferring it to the cloud. The impetus here is not about creating new workloads, but transferring mostly existing workloads to create greater efficiency. In other words, an outsourcing of workloads from an ‘in house’ data centre to a cloud scale data centre.
But with AI coming down the track, we’re seeing something that's quite different. What makes it so exciting is that it’s not simply transferring or cannibalizing existing frameworks. It represents completely net new demand to the compute ecosystem.
Furthermore, AI brings with it far more taxing and time-sensitive demands. When we think of increased computing loads today, we probably think of real-time media and communications, such as video streaming and online group meetings. But, even under the stresses incurred by the whole world going remote in the early days of the pandemic, the existing architecture held up and handled things pretty well. Everyone was able to rely on the in-place infrastructure to run their businesses. The infrastructure of today more than adequately fulfills that need for real-time communication and the overall way people use the internet today.
But AI changes all of that.
It's no longer humans talking to humans through a computer interface. In that context, latency up to 500 milliseconds would be acceptable.
Now it's compute talking to compute and here, suddenly, the response times are increasingly measured in single digit milliseconds.
So, these two factors – net new demand coupled with more urgent technical requirements - will drive an exponential increase in the need for compute architecture.
You touched earlier upon the view held by some that AI poses a threat to the data center business. Can you talk about that in more detail? What specifically is the risk these people foresee and why don’t you share their view?
Well, let’s use an infrastructure analogy. Let’s say you have an airport. And you then treble the number of planes that can fly into and out of the airport and you also make the planes more intelligent. Those planes, however intelligent and well-manufactured they are, are still going to need to land and take off at an airport.
Now, coming back to data centers: AI can help manage data center architecture more effectively. But, at the end of the day, you still need a physical structure in place.
Having said that, there are people who argue that AI will one day be able to design a silicon chip that doesn't use as much power as the chips of today. Therefore, a data center that once used a silicon chip that needed a kilowatt of power to manage – let’s say - a terabit of data, might, post-AI, only need a watt of power to manage the same terabit of data.
Now, if you're a data center operator that's just in the business of selling kilowatts, you have a problem.
But if you're a data center business that concerns itself with figuring out the required combination of kilowatts, network type, and response times for the workloads that you're managing, then you have something that’s valuable with or without AI.
And that value compounds when you build an ecosystem around the AI operators, cloud operators, and enterprise and network operators located in the same building.
The value of that is not measured in kilowatts but in the strength of the community that you’ve assembled in the same environment.
The evidence suggests that AI will play a critical role in data center management and developing the data centers of the future, but it will not supplant them.
Governments are just now starting to consider AI from a regulatory perspective. Media attention mostly focuses on the intellectual property and ethical angles. As an executive at a firm providing the plumbing for AI, which regulatory issues do you think will come to the forefront?
This is an important question. The single biggest issue in the data center industry today is on the regulatory front. We talked earlier about the tsunami of demand coming for AI. Well, that demand will bring with it a host of questions related to energy consumption and sustainability. Europe is leading the way in terms of thinking through these issues, partly because we have less secure sources of energy than the US does. There’s also less physical space available so finding locations for data centers is difficult. There simply aren’t undeveloped twenty acre lots out there on top of which we can put a 400-megawatt data center in Europe today.
I think the data center world in Europe will soon reach a critical juncture as it tries to figure out how to manage the demand fast coming its way. Everybody – and that includes governments as well as those they represent - wants to see the benefits of AI. Everybody wants to see the benefits of technology. But we must understand the real cost of that in terms of the underlying architecture required to support it.
As AI consumes more energy, all the parties here – industry, governments, and society – are going to need to come together to address some crucial sustainability-related challenges.
And we in the data center industry are certainly going to need to make significant strides towards becoming more energy efficient.
The market dynamics for digital infrastructure - high demand, low supply - are quite favorable now.
But let's envision a less favorable environment: what levers could a data center business pull to still deliver decent performance? What separates the skilled from the lucky when the tide rolls out.
This question also touches on several important areas. And it all goes back to customer knowledge and allocating capital appropriately considering that knowledge.
I’ll reiterate what I said in the beginning: you've got to understand what your customer is looking for. You need to be honest and rigorous about that. No one data center is going to be suitable for all customers. You need to balance the three parameters of location, price, and performance to determine the suitability of a data center to an application. You need to figure out where you stand in that equation and then position yourself accordingly. That strategic thought process - that rigor - needs to happen.
Because if you don't do that, then you won’t know the reasons your customers are in your data center and why it is they want to pay you. And believe me, it’s not because you have a sleek, attractive-looking set-up.
And so allocating capital in a thoughtful way that is led by your customer is critical. In fact, it’s the only way that I know how to deliver decent performance in variable environments.
You've worked for both startups and established companies. People often regard startups as hyper-efficient organisms that can accomplish goals faster; where personnel are given blissful freedom to break conventions and cut through red tape. Meanwhile, established companies are seen as hidebound, slow-to-adapt sloths.
You have experience in both settings.
To give established companies their due, are there strengths they possess that you don’t think receive proper recognition? Do they have some positive qualities that startups could or should perhaps emulate?
Well, I’ve been lucky to work in a variety of environments. That includes startups, larger entities like Arthur Andersen, and Interxion, which wasn’t a start-up when I joined, but still at a very early stage in its development.
But before I answer your question directly, I should say that successful firms at any stage share a lot in common. What’s essential – and what good firms succeed at regardless of size - is getting the culture right. There are certainly plenty of instances of startups, as well as larger firms, with toxic environments.
With that being said, a problem that seems more specific to startups is a sort of overweening vision or overconfidence. A startup can become convinced of its forward-thinking golden vision in a way that leads them to fail to consider alternative paths or, just as damaging, to overpromise to customers. Startups can have a drive towards perfection that can be admirable in some contexts. But just as often, the perfect can be the enemy of the good, and promising outcomes are ignored in favor of grandiose futures with less chance of success.
Meanwhile, the reality is this: If you can figure out how to be thoughtful and accountable; to be fair about the way you engage with your employees and customers - and you simply do what you say you're going to do - altogether, that can have remarkable impact and you can drive growth in a manner that makes a lot more sense.
Forthrightness – humility about what’s possible - allows you to build sustainable, long-term relationships and to deliver more assured, quality results to customers.
I’d like to ask about a different type of meaningful relationship. On a personal level: can you tell us about any mentors that helped guide you on your way up and any specific advice they may have given you?
There were two people that made an especially big difference.
The first is someone I came to know during the boom-and-bust cycle that shook the telecom industry at the turn of the century. It was 2002 – and tech overall was undergoing a major slump. I was serving as the CFO of a telecoms company that filed for bankruptcy. I was in my early 30s. And I thought that my career was finished.
My mentor then was someone called Bruno d’Avanzo - an incredible talent in the telecoms industry. And he proverbially slapped me across the face and essentially said, "Wear this as a badge of honor. These experiences where you go through immense difficulty are important. It’s a crucible of war and it’s these hard moments that will enable growth. You’ll eventually come out on the other side – and what you learn from this experience will be tremendously valuable; more valuable than any MBA.”
And then our investors gave me some money to be able to take a year off, which is incredible given that they had just lost a lot of money as a result of my poor strategy.
During that year off, I had time to reflect, and I came to appreciate that guidance as the best advice I had ever received from a mentor.
Now the other mentor helped initiate me into the data center industry. I can't lay claim to any great epiphany when it came to understanding how to build great data center businesses. Instead, I was educated by a gentleman by the name of Dave Ruberg, who was an industry titan in the data center world.
He could be really difficult to work with, and for. He fired me three times over the eleven years I worked with him. It was that kind of a relationship. Still, he was a huge talent and was, in my opinion, one of the greatest minds in the data center industry.
I was very lucky to learn from him, and though I can't say we always saw eye to eye, the lessons he imparted were invaluable.
I understand you also sit on the board of a medical school. Can you tell us a bit about the school? I’m also curious if you have advice for other private sector professionals who might be considering or about to take on board roles for the first time. Coming from the private sector, how can they best serve the non-profit institutions they want to support?
The medical school is in London. It's the UK's only independent Medical and Allied Health university.
To help frame this, when I retired in 2018 (before subsequently unretiring!), I needed to stop and consider what I would focus on. I did have some caring responsibilities for an ill family member at the time. But aside from that, I was thinking a lot about education in general, and increasing social mobility for people without access to decent education.
I was already supporting a school and a monastery in Nepal, which I continue to do.
So, as I thought about what I wanted to do, it became a natural progression to join the board of an education institution that was providing medical services to a close family member of mine and to give support to a field that is unfortunately in crisis right now in the UK.
Now, in terms of guidance I could give to other professionals regarding pro bono non-profit work, I think there are two things that that you can offer.
The first is your skills. And the second is your passion. Ultimately, your passion for a cause or effort is what will enable you to engage and ask the right questions in the boardroom, which is where you’ll find yourself seated.
Now the hardest thing for me - and perhaps for many CFOs, because we're detail-oriented – is to understand that you’re chiefly there to ask constructive questions; not necessarily provide granular solutions. Coming from outside the non-profit sector, you also need to appreciate that there are likely different ways to approach and solve a problem.
That seems like sound advice – and a particular challenge for someone whose everyday role involves serving as a fixer.
I guess that also helps bring our discussion full circle. Josh, thank you again for speaking with us and for sharing your insights regarding an industry playing an essential role in the development of AI; the set of technologies that is very much central to everything we are working on at CredCore.
You’re welcome. It was a pleasure on my end, as well.