Megaport Connect Cloud Networking Summit

Beyond Networking: Building the Infrastructure Platform for AI – Full Transcript

Alexis Bertholf: I think that one of the themes that has come up throughout the day is just how much the requirements for modern infrastructure are changing. Over the past several years, Megaport has evolved quite a bit as a company, and I’d like to spend some time talking about where we’ve been, why we’ve expanded beyond networking, and what we’re building next. So for our last session, please welcome the CEO of Megaport, Michael Reid.

Michael Reid: We made it, or we nearly made it. We’ve got a little bit to go, I guess, not much more. I’m standing between you and some drinks, which is a dangerous spot to be standing at any moment in time. What I’m going to do is, I think a picture is a thousand words and a demo is a thousand PowerPoints. So instead of me presenting a thousand PowerPoints, we’re going to see if we can get a demo up on screen.

Maybe. Maybe. Who knows? Live demos, am I right? We didn’t pray to the demo gods. So together we’re going to do the rain dance or the demo dance or whatever it may be. Anyway, if you keep doing that, I’ll keep telling you about Megaport until this thing appears.

So we’ve been through a bit of change. For those of you who’ve been following us for some time, you would have known us for networking globally. We’re the largest connectivity provider on planet Earth, in 30-something countries around the world. I don’t even know if that number’s correct or updated. And we’ve continued to add more and more countries as we go.

We started in Brisbane, Australia, and we actually started inside a data center called NEXTDC. It was actually our founder, Bevan Slattery, who created that company, and he had this problem where he was trying to solve this issue of connecting from a data center to this thing called the cloud. That was sort of 14 years ago, and Megaport was born to basically connect you from a data center to a cloud, and to do that in seconds. And to do that, I think we learned 60 seconds is how quickly you can connect from that data center to a cloud.

Now, at that point in time, we had a single salesperson, who’s standing over there. His name’s Matty. Matty was actually busy, I think, running a holiday resort called Peppers, or Salt and Pepper, or what is it? Spices. There you go. Spices. He got bored one day and Bevan convinced him to come and sell this thing that didn’t exist. And so he was literally out there trying to explain to someone a thing that didn’t exist, which became Megaport’s network connectivity story.

And from that moment in time, and Matty’s still with us actually, we had to fly Matty to the United States, because AWS realized that getting connectivity to the cloud became super critical, and being able to spin up private connectivity was the blocker between getting enterprises to move and migrate into these cloud platforms. So we were pulled from Brisbane, Australia, all the way to the United States. And today, something like, I don’t know what the number is, but maybe 90% of our revenue comes from the United States. We’re distributed, again, across 30 different countries.

Now, I’ve been in the role about three and a half years, and three and a half years ago we were purely a cloud connectivity company. In the last three and a half years, we’ve done some incredible leaps forward. In fact, most recently we’ve done some pretty insane leaps forward. We’ve gone from, I think we had 250 staff inside the business three and a half years ago. I think today we have 800 staff globally. So it’s pretty phenomenal growth if you look at the human side.

On the revenue side, we’ve got a pro forma annual recurring revenue inside the business of, I think it’s 1.1 or $1.2 billion, you’d have to check the chart, but I don’t have it. So I’m working off zero information here, just off the top of my head. But we’ve crossed about the 1.1, $1.2 billion in annual recurring revenue, from where we were about three and a half years ago at about $150 million. I don’t know what number that is, someone can do some maths. It’s incredible growth.

And that’s done purely because we’ve got incredible product, incredible humans building incredible product, and we constantly take feedback from our customers around what are the problems that customers are facing, and how can we solve that. And if you look at it from our perspective, we think about automation every morning. We get out of bed and spoon-feed ourselves automation. Every single piece of our infrastructure, delivered across 1,100 or 1,300 data centers around the world, is 100% automated.

In fact, it’s funny, when we look at hiring a network engineer into the organization, everyone turns up and starts trying to log into the CLI or log into a device. I think we’ve got something like 3,005 Cisco devices sitting in those 1,300 locations connecting all of this fiber together. We’ve got 6,000 pieces of fiber, and I think that number continues to grow. And most of that backbone is 400 gig today. So think about that: 400 gig wrapped around planet Earth in 30 different countries, resiliently stitched between every single data center, with high-speed internet in every single location. And you can’t log into a single router anywhere in any of those locations. It’s all deployed on demand.

So what I’ll do is, instead of continuing to try and articulate it, I’ll show you what this thing looks like. There you go. You’ve got to dance on your feet in this role constantly.

Now, for those of you who are joining us online, I’m sure we have some investors on there. They’re used to me doing these demos. I’m probably the only CEO on planet Earth that will ever, during an earnings call, do a live demo. And so far, we’ve never had a fail. So have a quick look at the platform that I’m showing you here on the screen.

Well, firstly, you do realize this is planet Earth? Are you familiar with that? Obviously, this is the spherical version. I’m not going to tell you who, but there’s someone in the room, and they’re sitting at the front of the room here, over this way, and they believe that the Earth is flat. So we actually don’t discriminate here at Megaport. We are very open to all of your views, especially if someone had a viewpoint that the Earth was flat. I’m kind of a spherical dude.

So anyway, you’ll see we’ve opened up all different locations. About a year and a half ago we opened up in Brazil, and now we have 11 different locations all across Brazil. You can see we’re all across Europe, in all these different locations. And if you cross down here, you’ll know that we opened up 44 different data centers in India last year alone, which is pretty wild. India was actually one of our largest requested locations. And it’s not because of Indian customers looking to get out to the world, it was actually most of our global customers who were looking to connect into India.

If you look at our customer base, obviously we’re all across Australia, New Zealand, you name it. But then if we come back into the United States, you’ll see this. I think there’s 400 or 500 physical data centers that we live in. Think about it. Every single dot represents a data center. We put resilient network infrastructure there, usually some compute, sometimes some storage, and sometimes some pretty large GPU clusters. So we’ve had to physically go out to every single one of those sites and stitch together, like I said, 6,000 pieces of fiber.

And all of it is just run via a portal. And the outcome, what I’ll show you we can do inside this platform, is actually a result of all the things that you guys have asked us to deliver. So let’s just zoom into a location here. You can start to see here is a data center. There’s NTT Global Data Centers. There’s QTS Portland. There’s Flexential. There’s EdgeConneX, another EdgeConneX. You can keep zooming in over here. Must be a data center alley over here. Then there’s Digital Realty and another Flexential. So you can imagine this very dense data center location; we have all these spots.

Now, the original portion of this business was to go into this data center, and you would say, I have some CPU or some compute infrastructure. I have applications running in there and I need to get a connection. And so our story was, instead of waiting 3, 6, 9, 18 months if you want to get a 100 gig connection, you can just click on this particular location, and we’ve got options. Create a Port, create a Cloud Router, or create a Virtual Edge. The Virtual Edge can spin up Cisco, Palo Alto, Fortinet, you name it. SD-WANs, etcetera, a whole range of products in there, firewalls, you name it.

Over here on the Port side, it’s really simple. What we’re going to provision is a 100 gig connection, and we’re going to bring that up. There’s a whole heap of stuff here, but I’m just going to go and give this thing a name. We’ve chosen the Portland data center. It’s Digital Realty, and all I’m going to click on is add a Port. You’ll see that should go blue. Okay, there we go. The demo has worked. We have deployed a 100 gig physical piece of infrastructure in that location, so you now have connectivity to your CPU and to your application.

Now we take that connection, we’re going to grab that Port, and we’ve got to go somewhere. So we go and add a connection. And what you can see is we have these different options to connect to. To a cloud: AWS, Azure, GCP, Oracle, you name it. To a private data center, so some other data center that you have. To Megaport’s internet platform. Something that most people didn’t even realize that we do is we provide high-speed internet up to 100 gig, delivered in 60 seconds at ridiculously low cost rates, with full hyper resiliency in this thing.

The guys can take you through how that works. Megaport builds and runs internet exchanges all around the world. I think we have 40, 60, I don’t know what the number is anymore. Matty will tell me. We’ve also got our Megaport Marketplace, which is where you have the likes of Cloudflare and all the different companies that we connect to. And then we’ve got service keys, where you can enter into a private customer.

So here’s something that’s really interesting. We have 600 financial services customers on the platform. Imagine that you needed to connect from one financial services company, say one bank, to another bank, and do that privately. Maybe you’re acquiring each other, maybe you needed to do some private, super secure connectivity; you can click and connect. And lastly, we’re super excited to announce that we’ve got Megaport Object Storage. We do both VAST and we have Wasabi platforms built in, and we build them in globally. I’ll show you a little demo around that as well.

But let’s just roll through and click to the cloud. So we’re going to say, all right, here are all the different cloud connectivity options. There’s Latitude; we’ll talk about that more in a moment. There’s Alibaba Cloud. There’s AWS, Azure, GCP, DigitalOcean, IBM Cloud, Oracle, SAP, and there’s our friends Wasabi there. And you can actually on-ramp yourself into any of these platforms. I’ll show you more about that later.

But let’s just choose Amazon Web Services. We’ll click a hosted connection. We’re going to choose a location that’s a little bit further away. Now, what you can see is Megaport’s the largest connected to AWS, Azure, and GCP on planet Earth. We’ve got more than 300 different cloud on-ramps globally, prebuilt with infrastructure connected into them. And in fact, this week you would have seen that we announced a huge partnership with AWS for their… I think someone will have to explain it. On-ramp, or what are they called? AWS Interconnect. AWS Interconnect. It’s a pretty obvious name. I guess I should remember that.

So if you scroll down, every one of these locations represents a cloud on-ramp that we have previously built into, and you can scroll through and actually see where the furthest location is, in milliseconds. So you can see each one’s got a latency. The furthest away from there is Mumbai, 205 milliseconds. Well, that’s a long way, isn’t it? Let’s scroll all the way back up to the top.

And again, every one of these is prebuilt, so I feel sorry for our network engineers. If you notice, there’s a red and a blue path. What that means is we’ve built physical resiliency between it, so physically separate network infrastructure. The only thing you need to think about is, I have a red network all the way through, I have a blue network all the way through, and you have 100% fully resilient infrastructure on all sides. If there’s ever maintenance done, it never gets done on a red and a blue zone at the same time. Kind of obvious stuff.

But that’s all you need to do. Now all you need is a name, so you give it a little name. You can choose a speed. You can scroll down; we can go up to 25 gig. We’ll do 1 gig in this instance. You can choose month-to-month, yearly, whatever you want to do. Click Next. All you need to know is your Amazon Web Services account ID, and the magic behind this is the platform will stitch it into your AWS platform. So once you finish the connection, it will appear inside your AWS portal, and it’s magic.

So we’ll scroll down, we click add a VXC, and now what you’ll see is that connection is live. So I guess faster than 30 seconds, I don’t know, 60 seconds, whatever it is. So there you go. There’s a connection to AWS, provisioned across 100 gig connectivity, and you’ve done that in 60 seconds.

But we’re going to speed it up and go elsewhere. Let’s add another connection. Let’s say you want to connect data center to data center. Here in the US, you can actually connect at 100 gig in 60 seconds anywhere in the United States, and I think it’s $2,099 or something like that. So someone will have to confirm that number for me. But it’s incredibly low cost, but hyper resilient. So you can run across thousands and thousands of different paths. You actually get access to that.

In theory, if you dragged an anchor through the middle of the United States and you were connecting from the East Coast to the West Coast, we would just route you around the other side of the world. Let’s hope no one’s going to drag an anchor through the middle of the US. Now, for those of you who are listening, my accent is from Australia. So let’s choose NEXTDC, based in Brisbane. So we’ll do a Brisbane location. It’s 148 milliseconds away. This is where the company was founded.

So why am I doing this? I have my application running in this particular data center in the United States, and I want to replicate that application into, say, Brisbane, Australia, on private 100 gig connectivity that runs between them. And I want to spin that up on demand, instantly. So let’s just do a couple of those pieces. Click through; there’s a whole heap of technical stuff you can add in here. I’ll see if it just lets me finish it. Oh, there you go.

So now what we’ve done is we’ve provisioned across multiple subsea paths. By the way, we have at least three, and in some cases seven, subsea paths to most locations. And then you can see that connection is now delivered. So you can deliver 100 gig, wrap the world in 60 seconds, and continue to do that. It’s quite crazy what you can do with the platform.

Now, the last thing I’ll show you is I will add a connection to the internet. It’s this crazy thing called the internet. We can see that 4 milliseconds away, we have an internet on-ramp in Digital Realty. That’s pretty close. We click Next. We give it a rate limit and a speed, whatever you want. Choose month-to-month, terms, etcetera.

And then all you need to do is add DDoS protection. This is a product that we launched this month. So what you can do is actually click on that and add the VXC connection, VXC on-ramp DDoS protection, and that’s $30 USD per month. Click, add that, and away you go. You’ve got full DDoS and internet.

So for those of you who know, we acquired a company in November last year, December last year, a company called Latitude. If you think about Megaport, we’ve been connecting folks from one data center to another, from a data center to a cloud, and between clouds. And we do it in 60 seconds, high speed, super resilient, all on demand.

And so our mission was always automating physical infrastructure. It was originally network infrastructure in data centers. And then we started building out compute infrastructure. If you look at our Virtual Edge platforms or our Cloud Router platforms, they actually spin up software routers and firewalls in physical servers that we started deploying in hundreds of locations around the world.

And then what we realized was that our customers were really connecting from one site to another, but not really between sites. We’re really connecting from one server in a data center to a server in another data center, or from one server in a data center to a server in a cloud. And then those applications would run across. And so what you see on this screen is servers. So now Megaport has compute, network, and also storage, delivered in less than 60 seconds around the world in 30 different countries.

And so we’ve spent a lot of time over the past 10 months integrating the business, which is Latitude, for our compute stack. And just like you saw on the previous demo with the network side, you saw Create a Port. Over here, you just see this button that says Create a Server. And so we can click on here and we can create this server.

And now what you’ve got is bare metal. For those of you who are familiar with what bare metal means, it means a full physical server, and you get every component of it. If you think of a cloud, they’ll deliver you a virtual environment that lives on many, many different servers. You don’t have access to the physical machine. So we provide a very different service to what AWS, etcetera, is providing. It’s not competitive. It’s actually an extension of what they’re doing. And it even allows you to deliver AWS into that premise infrastructure as well. So this is truly a value-added stack.

So you get to deliver what is bare metal, so a physical machine, and then you can say, all right, well, where do I want that machine? I want that in North America, in South America. You can see all these different locations. In Europe, in Asia Pacific, etcetera, etcetera. Now, if you look at Ashburn, let’s take Ashburn, which you’re familiar with. It’s the home of the clouds. So we can scroll down and we can start to see the different types of physical machines that we have pre-provisioned in this location.

If you’ve all been into a data center and you’ve walked down that hall with all those blinky lights, imagine what we’ve built is thousands and thousands and thousands of physical servers. I think there’s 15,000, I don’t even know what the number is anymore, and every day it seems to change. Down a hall, say 60% to 70% or 80% of them are currently utilized. But there’s this big pool that’s available. So we’ve pre-provisioned all these different types of servers.

And you can see this is an RS metal extra large. It’s AMD 9554, 64 cores, 3.1 gigahertz, this many blah blah blah, RAM, etcetera, etcetera, $5 an hour. So I’m going to click and deploy this particular server and have total control over it, which means I can max this thing out. And this is one of the reasons that our revenue has gone vertical, and I’ll show you a slide that goes with that. The AI customers that we have on the platform are using compute so fast that we actually can’t keep up with the deployment, let alone getting it to them. So it’s one of the reasons they love us.

So if we scroll down, something really interesting: imagine all these physical servers sitting down this data center. On each server, we can actually deploy an operating system. We can actually pre-deploy an operating system dependent upon what the natural usage is. So we’re constantly scanning what other customers are using, and we’re pre-provisioning a physical machine with an operating system.

And the best way to think about that is, if you’ve got an iPhone, you pulled it out of the packet, the beautiful Apple packet, and imagine that it didn’t have an operating system on it, and then it took you two hours to go and download it. What we’re doing is pre-provisioning. So here is an example. We’ve pre-provisioned Ubuntu, and you can see there it deploys in 5 seconds. You can deploy all sorts of SSH, user data. You’ve got all these sorts of crazy technical things you can do.

We can make it hourly. This is truly hourly. At the end of the hour, you can stop paying. So you can just pay $10 and then turn it off. And I can click deploy. And so what will happen now is you can see that that server is now deploying. We’re in Ashburn, and look at that, less than 5 seconds, it’s already on.

So that is kind of crazy. We have deployed a very expensive, high-end, huge amount of memory physical device sitting in Ashburn that you have complete control over. You can actually click into this. You’ll see this, and I’ll delete it shortly, because that is a live IP address. Last time I did this, our shareholders started pinging it to make sure I wasn’t doing some fake demonstration. One thing you’ll learn with me is everything’s real. And half the time I’m learning this stuff as I go, because they keep updating the platforms on me and I just never keep up.

So you can see there’s the IP address. You can ping it. There’s all sorts of information, and you can scroll down to the bottom. You can add virtual machines on the top, SSH keys. You can add user data. You’ve got AI inference you can stick on top. Kubernetes clusters can be deployed on top of that platform. High-speed network, you can go and spin that through. You can add storage platforms straight away to it.

Now what I’m going to do is delete this server. And the reason I’m going to delete this server is I actually genuinely want to sell this server to someone else other than myself. Compute’s pretty hot at the moment. So I’ll show you. That’s me deleting the server. You can’t fat-finger it. Server’s gone. So that’s it. What happens now is that physical machine gets wiped three times, so it’s a really safe and secure way of removing all of the data, and it gets added back to the pool, and you can procure it moving forward.

That’s my ta-da moment. No one’s excited for it. People are meant to be punching the air and saying that’s the greatest thing I’ve ever seen. People are probably pretty bored. They’re like, just get me a drink.

All right, so a few other things that we launched, and this is very early. We haven’t yet launched AI inference, so I won’t demo that. I’ll just hint at it. And we’ve got Kubernetes clusters sitting across the platforms in preview. And we’ve got this cool thing called storage. Now, we’ve gone and launched object storage, block storage, file storage, Postgres. That wasn’t there last time I demoed; I didn’t even know what that means.

But let’s just go to Object Storage. And it’s very simple. You can come in here and click a button, and here it is: Create a Bucket. So let’s go and create a bucket. Now you have two options. Super high performance, and you can see that it’s, whatever that is, 0.024 cents per gigabyte. It’s very cheap, apparently. I think it gets expensive as you add more and more things. But that is incredibly high performance, on demand, and that is the VAST platform that you heard about today.

And then you’ve got our standard storage, optimized for what you would think of as backups. It’s very much an S3 equivalent in the clouds. You’ve got it at $8.49 per terabit per month, or terabyte, I don’t know which one it is. You can pick your locations. By the way, this is the Wasabi platform that we’ve gone and automated into the business.

You can get up to 100 gig of free connectivity, private connectivity, deployed with this solution in country, deployed in seconds, and then for $8.49 you get to have that component there. There’s not a single cent of egress. No cost for ingress, obviously. There’s no API taxes. That is the entire amount that you pay. So that’s an awesome platform.

You can see all the different locations. We’ve gone to Europe, Paris, Frankfurt, Milan, Amsterdam, Sydney, Osaka, Tokyo, Singapore, you name it, all those different pieces. And I’ll deploy an example of this in, say, Ashburn.

Now, what we’ve done from a cool perspective is, if you think about it, if anyone’s had an experience with crypto-locking or getting hacked… It’s usually grandma that calls up and says, all my files have been locked and I need to pay this thing called Bitcoin to someone. Well, that’s your personal experience. There’s lots of companies that have had the same experience. There’s some incredibly large companies that you would have known that have been completely crypto-locked.

Now, one of the problems when you get crypto-locked is, and this is an interesting stat, 70% of the time this occurs. 70% of the time, you call your IT person. Your IT person says, what do they say? Don’t worry, I’ve got a backup. 70% of the time, they’ve actually already encrypted the backup, then they hit your front end. And when that happens, it’s game over, Red Rover.

So the good news is, if you want to prevent that from happening with us, you just click this button. Just one button could save your life, or at least save your job. You click, choose compliance or governance. We’ll say governance, and we’ll add 30 days, for example. And what we have here is we’re creating a bucket of storage that can scale and expand with the company. You’ll never pay more than $8.49 per terabyte, or bit, which one was it? Byte. It’s in terabytes.

And we can add a 30-day immutable backup, which means you cannot change that backup for 30 days, which means if someone tries to crypto-lock you, they can’t do anything. There’s all the detail. You copy and paste that endpoint in. You point your software at it. Done. You couldn’t really get any more simple than that. You go and add connectivity into it, and you’re away.

So what have I shown you? I’ve shown you the world’s best network company, literally the world’s largest global, you name it, everything delivered via software on demand. We’ve shown you how we do that across both private connectivity to AWS and a physical Port. We’ve shown you connectivity across subsea all around the world at 100 gig. We’ve connected you at 100 gig to an internet provider. We’ve added DDoS to that platform. We’ve gone and spun up physical compute in Ashburn. I’ve gone and put a storage bucket on top of that, and I can keep going.

The last one that we wanted to share, which is growing at a rate of knots, is the other component, which is where you create a server: you’ve got bare metal and GPU. And so one of our challenges here is we are literally sold out every 5 minutes. We’re rolling out probably, I think, a couple of billion dollars’ worth of GPU inference factories across the world, and they sell out in 5 seconds. So we are desperately trying to continue to keep up with that.

So what I’m going to do is, I think I can switch back to my presentation, if I did have a presentation, and just work through these pieces. Time-wise, how am I going? How long have I got?

Alexis Bertholf: As long as you want.

Michael Reid: Oh my god, I was going to keep going. No, I’m going to try and keep this as quick as possible, because I know you’ve been here for a long time. A picture says a thousand words, and a demo is a thousand PowerPoints.

So if you look at the company, we’ve come a long way and we’ve changed dramatically. We’ve moved from what was just a cloud connectivity company to data center to data center, data center to cloud, cloud to cloud, virtual firewalls, virtual routers, subsea connectivity, internet providing, DDoS, cybersecurity, and that’s in the last 3.5 years. Last year we then added compute, and then this year we added storage.

And if you look at the three key pillars at the top, they enable these pieces at the bottom. AI: we are producing AI inference clusters all around the world. So we are becoming an incredible provider of inference, which is tokens as a service, the tokens that you utilize every time you click into, say, ChatGPT. When you ask it something, when you’re punching out tokens, that’s us.

On the security side, we obviously secure this entire platform, and we’ve added all sorts of security products across it: firewalls and encryption platforms, DDoS, you name it. You should expect us to continue innovating. And we’ve also added services. We’ve had a lot of requests from customers. Believe it or not, as simple as what I showed you is, we’ve had lots of customers say, I wish someone could manage this entire device. We work with our partners, and we also deliver services ourselves.

We’ve done pretty well. We’ve got 4,000 enterprise customers in that space. We’re across 30 countries, 37,000 services. Those services are changing every second, and not a single human ever touches it. We’ve got something like 170-plus terabits, terabytes, I don’t even know the right word. We’ve got 1,100, I think it’s 1,300 data centers, 20,000 different connections. The on-ramps are over 300, 410 service providers. I think we’re up to 15,000 CPUs and GPUs, and you name it.

We’re incredibly diverse. We’re all around all these different locations. We’ve got 330,000 actual kilometers of fiber routes. I don’t know what that is in miles; it’s a lot. We’ve got 3,300 network devices, and you can see us continue to grow. You should expect us to continue to expand, to continue to invest in locations.

So from a strategy perspective, from a CEO side, what are we trying to do? And I tell our shareholders this all the time. Our job is to add more rings and expand rings. That’s it. There’s actually nothing else that we need to do as a company. It just turns out adding and expanding rings is not easy.

So what does that mean? This slide represents total addressable market, what we would call TAM. When I first entered the business, we really did cloud connectivity and a little bit of virtual edge, so the two right in the very center. I don’t know whether it’s going to work at home, but this is what we delivered. And then every time we add innovation, our team goes and builds incredible product, usually at the request of our customers telling us, I wish you did this, I wish you did that. Pretty much all of our products have literally come from someone saying, I wish you did.

The first one was Global WAN. Then we added data center interconnect. This particular product has gone through the roof. If you think about it, a connection to the cloud might be $100. A connection from the East Coast of the US to the West Coast might be $2,000. A connection from Brazil to Singapore subsea at 100 gig is like a million dollars a year. And so what you’ve seen is all the services that we’re adding here, hyper resilient and far better priced than what you see on the market.

Adding internet: now, most people just assumed we’re an internet company, but adding internet has been an acceleration of our business like you’ve never seen. We added a thing called NAT Gateway, which literally reduces your cost of network address translation to the cloud by something like 70%. Now, we’ve got customers that are spending something like $50 million a year translating an address from a public address to a private address, and we can reduce that by 70%. Security: we went and added all these different security products. You should expect us to continue to invest in that space.

And remember, we can build, innovate, and invest. Investing means you can acquire companies that make sense in this portfolio. Everything I’ve shown you so far came from innovation in-house. We’ve got something like 90 engineers inside the business, and all they do is build software. By the way, they’ve 10x’d their software since Claude. We can talk about that over a beer. And then if you look at the CPU and GPU space, that was our acquisition of Latitude. And then storage became the last product, which was the innovation we’ve added since.

And you should expect us to add rings and expand rings. So you can see what adding rings does. What is expanding a ring? Well, expanding a ring is where you don’t innovate a new product, but you add a new country, a new location, or more speed. When I first joined, the fastest we could deliver you was 10 gig anywhere in the world. To go from 10 gig to 100, you needed to upgrade every single connection to 400 gig resiliency, across subsea fiber, you name it, around planet Earth. That is not easy. That took us about 18 to 24 months to get all of those pipes upgraded.

But by doing that, we didn’t add another product. We just massively increased the speed, which increased the total addressable market, which means we can actually go and deliver more services to you as customers. So what you should expect from us forever is new DCs, new markets, more capacity, and adding more and more CPU, GPU, you name it, investing in that space. You should expect us to continue to innovate. You should expect to see me continuing to invest in world-class talent and phenomenal engineers, and you should see the rate of innovation that we’ve had and actually believe that we can deliver on that. So please keep telling us the cool stuff you’d like us to add.

DDoS came from one of our French customers, who was like, we really want you to add DDoS for us. Well, of course we can. We’re one of the most globally diverse networks on planet Earth. We can actually scrub DDoS in every single location and actually take a feed, versus getting smashed in a single location. So it’s actually more powerful than what most carriers on planet Earth can do. And so we continue to do that.

The principles by which we live and breathe are interesting. If you ever wanted to build Megaport, I’m giving you the answers to build Megaport. It’s just not easy. The first and most important one is automation. Every single thing that we release, every single thing that we do, cannot be touched by a human. It has to be delivered via software. And to do that is incredibly difficult.

And there’s not another telco on planet Earth that’s been able to do it, because you have to build from the ground up. The second you buy another network is the second that your automation instantaneously fails. And most telcos are local; they’re based in one country. We’re in 30, and we have one piece of infrastructure, one platform across all of that. If you can automate something, it means you can make it instantaneous, or 60 seconds at least. If you can do that for one thing, it means you can hit massive global scale. We don’t actually care how many sites we add, how many machines. Again, it’s all run by software.

Now, if you can build it, you can make it incredibly resilient, more resilient than what you could do by yourself. How do you do that? We have dual paths out of every data center, in some cases six paths. Particularly in India, you need six different paths because of all the outages that you will have. In the United States, most of the time it’s dual path, and when you go to South America you’ll see multiple paths, sometimes four paths. So we physically build that diversity for you.

And then we go and start to build diverse paths even into internet. There’s like three internet providers in everything that we blend. So if an entire internet provider went down, you still have resiliency. If we make it super flexible, what does that mean? You can deploy it in seconds. You can actually click. You can change it. You can pay per month. You can pay per minute, in fact, in most cases.

Then you can add the self-service element, and that’s what I just showed you. If you can make that really, really easy, and you can add the best support, and then ensure that the pricing is incredibly disruptive, and somehow remain profitable, you have the ingredients for an incredible company. And so we’ve been super blessed. We’ve been able to grow this company at a rate of knots. We’re a publicly traded company. We forecast this year’s growth to the market, which means you put your left pinky on the table and they chop it off if you miss it. We’re over 100% revenue growth this year. Crazy.

All right, so there’s the journey. We started with network. We added compute. We added storage. We’re adding AI inferencing, and this has been a huge rollout. These are NVIDIA chips that are deployed across many, many of our data centers. I’m not even sure of the number of data centers and the number of megawatts that we’ve been buying, but it is a lot. There’s billions of dollars of investment going into that, and that’s equating to some pretty crazy revenue.

So if you look at that journey that I was trying to articulate before I had my slides, see how much better it is with a slide deck versus just trying to tell you? I think I joined the business somewhere around here, and what you saw was the network business continues to grow. We added Latitude, so we got the compute. We added the compute piece. So this is just an update today. And these are the future contracts that we’ll deliver by the end of the year, or whatever it is, committed ARR, which takes us up to 1-point-something billion dollars of annual recurring revenue. So you can see that orange literally is contracts done in the last four months. Crazy.

I’ve never been in an environment that’s growing this fast. I’ve never seen anything come close to the speed at which we’re growing, but the speed at which our customers are growing too. We have a customer that’s huge that announced two days ago that they grew 15x. 15x, and they’re raising money at a rate of knots. This space is really, really interesting.

We’ve moved from this world of AI training. I think many of you know about all these giant chips: you build out gigawatts of factories, and you train this model. And training was all the rage for the last, say, four or five years, or four years, or whatever it’s been since we’ve gotten into this AI race. But the money is made at inference. And inference is every time you access the model.

So if you’re creating a model, you need to train. Once you’ve created the model, you actually make no money out of training the model. It’s very expensive. It’s only when you take that model to users, like Claude, for example, if you’re doing coding. Every time you enter into Claude, it uses tokens. Those tokens are the thing that creates AI and gives you, a user, an experience. And AI inference is what that is.

Now, the difference is training is giant clusters that are super dense in one location. It could be a gigawatt in one site. Every single GPU connects to every other single GPU. Inference is actually about low latency, and it’s about distributing. So instead of having a gigawatt campus, you could do a 2 megawatt facility. That’s 2%, 2% of the size of a gigawatt campus for the training, and you can start to deliver inference.

And you would want a company that’s located in lots and lots of data centers. I know one that is networked together. I know one of them that can deliver you inference on demand. So this crazy position that we landed in was the exact perfect location. We’re in 150 different data center providers’ locations. So we’re in 1,300 physical locations, but 150 providers. We have contracts with them. And so when we went to expand the last four or five months, we’ve grown pretty fast.

This is a chart to show that there’s a lot of AI coming, if you’re not aware of it. It’s going through the roof, probably pretty obvious right now. The puzzle is this. The puzzle for AI is three pieces. It’s power and space, if you’ve all heard about it. And so there’s a lot of controversy and a lot of chaos going around data centers at the moment. But this is never going to go away.

Shell, so it could even be this room here, literally something to keep the rain off your stuff, and then to power it. The next one, you need to connect it, and the last one is obviously you need to stick a lot of NVIDIA GPUs into that. And so that is literally what you need to build the inference cloud. And then you can service all sorts of customers.

And if you look at the compute stack that we’ve gone and rolled out, people are like, who is going to buy physical compute? Bare metal? Who is going to buy that on demand? Well, it turns out a lot of companies. AI is probably the fastest-growing space, and that thing is a vertical line going straight to infinity, and I don’t know when that’s going to stop. At the moment, probably the limiting factor is power, and the second limiting factor is capital, literally money in the world. But it’s not demand, so that’s pretty crazy.

You’ve got Web3 and blockchain applications that need super high performance, very low latency physical infrastructure distributed all around the world. You’ve got enterprise applications. Anything in your enterprise that needs super high performance, it can’t run on a virtual machine in the cloud. And if you do, it’s either ridiculously expensive because it’s not optimized for it, or you can’t get the right performance out of it. SaaS platforms: again, if you need performance, you need low latency, this is the perfect piece.

Gaming is the obvious one. My son loves his Fortnite and some other new one he’s playing; I don’t know what it’s called. But anyway, if you look at his platform, he will tell you, Dad, Dad, you’re too far away from wherever we’re meant to be, or something. God knows. Latency is a massive issue for that space.

Ad tech. If you’ve ever jumped on a laptop and someone’s trying to show you an ad, the speed at which they show you an ad, the speed at which they work out that I am a male, I’m X age, I have two kids, I don’t know how it figures that out, and then it shows me some advert. The faster it does that, the better. And then streaming is the last one. That’s the CPU side of the market. Probably the fastest growing is AI at the moment. Everything else continues.

And on the GPU side, you’ve got inference, fine-tuning, and obviously training, and you all know how fast that thing’s growing. When it comes to storage, storage is a really important part of this puzzle. The three pillars that I talked about before: basically every application on planet Earth just uses three things. It uses compute. If you look at this laptop, there is a CPU inside there. There is a Wi-Fi network card, and there is a hard drive.

Every laptop on planet Earth, every single device that is deployed to give us an application, is its compute, its network, and its storage. That’s what Megaport delivers: compute, network, storage, on demand, anywhere on planet Earth. And the storage element is when you look at object, block, and file. Storage on demand, delivered in seconds. If you get into the storage realm, there’s different types of storage. You can go super technical on it, and the boys did a great job before.

But standard object storage, really simple. $8.99, or 49? I’d have made it 99. 49. Change that to 49. See, this is me randomly at night, just entering things in. I’m wrong. Terabits per month, is that the right B? Should it be a lower B?

We’ve then got high-performance object storage. This is the majority of the storage that stitches into the AI clusters, the inference clusters that we’re providing, or the training clusters. This is VAST. This stuff is incredibly fast. It’s crazy, crazy fast, the performance that you get off the back of these. It’s custom built and designed for that platform.

This is a company that’s just gone to the moon. I think, last time I checked, they’re a $30 billion valuation private company, and they’ve only been around for a few years. So they’ve done incredibly well. We’ve then got high-performance block storage, which is growing like a weed inside the business, because that stitches onto the side of any CPU. And then you’ve got high-performance file storage, which is your traditional file storage that you would look to use.

So every year I get in front of the sales team and I would say to them, all right, we’re on a mission. So rewind. Imagine I joined the company and I said to everyone, we’re $100 million of ARR. And I got on stage and I said to them, we’re going to get to $1 billion of ARR. And everyone looked at me and thought, whatever, Australian rando. And I was like, well, we’ve got to put out some lofty goals that we can all head to and try and achieve.

And I thought, well, maybe over five years we might get close to it and we’ll get excited. Well, this year we crossed a billion dollars in ARR. So the problem is, being a publicly traded company, it’s not just about getting to the destination. As soon as you get there, you have to extend it. And so I just thought, well, we’ll 5x that, because I didn’t really think we’d get the first one in the first place. And so now we have this vision for $5 billion. I don’t know. We’ll make up a number; we’ll call it 10 probably next year.

The rate things are growing is off the charts. This is more of just an aspirational view of where the company is going, to give you some perspective. So I threw this to the sales team. But what’s probably more important is how you take a company there. And whether it’s $5 billion or whether it’s 2 or whatever, the importance is not the dollar. It’s actually what we’re trying to do. We’re trying to actually deliver incredible product that solves your problems, and do it in a way that saves you money, is more resilient, and you actually get an incredible experience from it. And funny enough, that is product.

And so the number one most important thing that we get up every morning and think about is product. How can we continue to innovate? How do we add more and more product to it? Now, interestingly, with AI, we have this blessing where we can actually move about 10x faster than we could ever move, because we can code at a speed that’s probably 10x faster. What we’ve got is something like 90 software developers inside the business that wake up every day. We just went and 10x’d the capacity of what they’re doing, which is mind-boggling.

And just for a note, publicly, AI is meant to be this thing that’s going to take everyone’s jobs. Everyone’s freaking out about it. The most disruptive thing on planet Earth that AI has disrupted today, who knows what that is? Software engineering. It’s the first thing, the most disruptive thing. And it happened at Christmas last year. In January, people came back with Opus 5 and realized that they could code 10x faster.

Are there less software engineers being hired today, or more? More. More. Let that sink in. I’m a massive optimist in this space, but you just have to look at that. There are more software engineers being hired today, and in theory that was the one thing that AI certainly disrupted. Now, luckily, we’ve had more and more things that it’s coming in to disrupt, but it’s not going to take your jobs. You’re just going to hire more. We humans have this terrible trait of just turning the treadmill up and running faster.

Let me give you the best example. When the calculator was invented, we as humanity could have taken a three-day weekend. We literally could have just said, that’s it. We’ve got the calculator. We don’t need to work as hard. We’re having Friday off. Did we do that? No. We just worked harder. God damn it, humanity. We actually have this opportunity right now to have a three-day, four-day weekend. You know what we’re going to do? Unfortunately, run harder. God damn.

Anyway, this is a different discussion that we can have. But we’re just going to move way faster. Companies grow way faster. Companies, instead of growing at one times a year, are growing 15 times a year. That’s all that ends up happening. You’ll see us continue to do new products and new components, and we’re going to be flexible.

But what is the one thing that I explained out of that? I said we need to work on product. What is actually more important to actually get that outcome? Does anyone know? Do you think it’s AI? What did I say I’m hiring more of? Product engineers? Are they humans? Yeah. So my number one job is hiring and attracting world-class humans into this business, keeping them inside the company, enabling them with AI, and producing wicked products that we can take back to you. So really, the number one priority of this company is people. It’s keeping great talent inside the business.

And the second thing, once you’ve got a wicked product, is that we need to tell you about it. And that’s really, really hard. It’s getting incredibly difficult because AI is becoming so much slop inside your email and what have you. There’s been a lot of piss-taking around customer success before, I heard. Customer success is a really important part of our business. When I first joined the company, we had one customer success person. I think we have 50 globally today.

And we’ve got Bennett sitting down the front here. He’s not customer success. This guy is hardened. He’s our frontline seller. So his job is to go in and talk to every one of you and our customers, and actually bring the new products to market and figure out how they fit inside that space. And we’ve got solution architects in the room, someone at the back I can’t see. We’ve got some wicked solution architects that will come in and actually design the solution with you, work with you, make sure the commercials are right, structure it in a way, put together a managed package for you, and deliver you this beautiful thing that is incredible.

What does that result in? From our business, if you think from a CEO perspective, if I can get all those things right, it just feeds that last piece of the puzzle. It means we have new customers arriving. Alexis is one of our greatest go-to-market leads. Every time she does a post, there’s 10 billion people that like it and then start buying our product. And so, getting the message out there to bring new logos, having customer success people, so those new logos go to Bennett, all of them. Bennett’s like, yes, we’re good.

So new logos would land with the new frontline sellers. We would expand and love and protect that base; that’s our customer success team. Their job is to make sure that you’re using the product to its full, and that you’re aware of all the other cool stuff that we’re doing. And if we bring cool stuff to you, we should be able to retain you as a company. We should be able to expand you. And the last one is constantly looking at strategic, crazy big opportunities where we partner together.

And that’s it for a company, and it finishes off where we started, which is compute, network, and storage, delivering AI, cybersecurity, and services, all wrapped in. That’s where the business is. It’s come a long way, having lots of fun. But without you as customers, without someone saying I wish you did this, or I can’t believe you don’t do that, or this is driving me crazy, or this takes too long, we wouldn’t be where we are. So please keep giving us the feedback. Every single time you tell us something cool, it just becomes a product on the list, and then we’ll keep innovating. I’m pretty sure that I am done, and I’m done early.