AI Brings Back the Early Days of the Web for Cloud Infrastructure
AI is rebuilding the entire technology stack, and cloud infrastructure sits right at the center. For veterans of the early web, the moment feels familiar. Everything is new again, and almost anything seems possible.
Vultr CMO Kevin Cochrane examines that shift with Alan Shimel at KubeCon + CloudNativeCon in Salt Lake City.
Back to 1996
Cochrane and Shimel both launched their first companies in 1996. The Apache web server had just arrived. Open standards and open source drove innovation at every layer of the stack.
Cochrane co-founded Interwoven, which helped create web content management. Shimel built a web hosting business. Both say today’s AI wave matches that era in scale, or even exceeds it. Shimel admits he never expected to live through another shift this big.
From Hosting Roots to a Global Cloud
Vultr has similar beginnings. It started as a small hosting operation run by a handful of engineers. Heavy automation let that tiny team support demanding enterprise workloads.
Today, Vultr positions itself as a global cloud platform for developers. It offers GPUs from NVIDIA and AMD alongside core CPU compute. Cochrane notes that CPUs matter for general workloads and for agent sandboxes.
More Than Renting a GPU
Shimel argues that selling bare metal GPUs is a scarcity model that won’t last. Cochrane agrees. Teams also need networking, storage and CPU clusters next to their GPUs. Developers want the whole stack, including the cloud native tools that run on Kubernetes.
Vultr focuses on infrastructure as a service and leaves platform services to partners. Cochrane says that keeps the company focused on performance, resilience and cost.
An Open Ecosystem
Vultr is now a CNCF Platinum member. At KubeCon, it also launched the Vultr Open Source Alliance. Partners get documentation, one-click deployment and joint go-to-market campaigns.
Cochrane draws on his Interwoven days, when the company won by working with everyone. He believes platform players must support choice. Open, composable cloud infrastructure is how developers will build the AI-native stack.
Transcript
Hey everyone. This one here is one I've been waiting for, actually truth be told, I've been waiting to do this interview for two months, three months since I first met this man, but it was a virtual thing. And you know how virtual ones are, it's a Zoom.
For all I know, he's 5'2" or 6'10". Who the heck knows, right? Right.
Exactly. But here live in the flesh is Kevin Cochrane. Kevin is the CMO at Vultr.
Kevin, it's a pleasure. It's so great to see you in person after all this time- Thank you ... we've been trying to hook up.
And you're not 5'2"- I'm not 5'2", right ... or 6'10", you're in real life. Kevin, we're going to talk about Vultr.
We're going to talk about a lot, but the thing I remember in our last conversation was saying, Kevin's been through a lot of the same things I have. Mm-hmm. He's been around the block.
He's seen this industry up and down and all around. Yeah. Give people a sense of your journey.
Yeah. So I've been in tech for over 30 years. So my origin story, if you will, is I was co-founder of my first company back in 1996, June of 1996.
Me too. Right? Amazing.
I know. Exactly. It's so long ago.
The internet went- The internet just happened. Yeah, exactly. So I was literally a management consultant for two years, and then the Apache web server was released in April of 1996.
And so I basically said, "You know what? I think I actually should stop being a consultant. " So here's what happened.
This is true. I literally rented a U-Haul, drove it to my house after work. So I got off work at 7:00, got to the U-Haul, drove it to my house at 8:00.
I was living with our brother. " Packed up all my stuff in the U-Haul, left at midnight- Holy ... drove up I-5- That was ballsy ...
got to San Francisco. My friend Aaron Ross had an apartment in Russian Hill at Jackson and Fillmore. Sure.
Aaron went on to eventually do Salesforce- Yeah. Yeah ... the principal revenue guy.
" And they've got this amazing, smartest person, the most incredible person I've ever met in my entire life, Peng On. He was the founder of Interworld, and I just joined as the business hire. And then I started becoming a product manager, and we invented a category of software called web content management.
So here's the thing. Why do I tell that story? Is because we're living in that exact same moment- Right now ...
with AI right now. That's what I said in the book. Everything was new.
Everything was possible. And everything was possible. In all layers of the stack, you were having innovation, right?
Yeah. And it was all being driven by open standards and open source. It was all being driven, Apache Software Foundation, which was just new at the time.
And so it was a terribly exciting time where just every day you were having to learn. People were getting away from client-server applications. They were starting to build their first websites and web applications.
And application servers were just starting to become a thing. And remember, everyone actually had all of their servers in rooms. In the ISP on donut baker.
Right. Used to call them baker racks. They would wheel the wire frame.
Right. Exactly. I know.
I remember those days. And so then we had to start figuring out how to build worldwide distributed architectures for scaling web applications planet wide. It was a wild time.
So anyway, so you fast-forward and we're at that exact same moment. We are. I see it.
It's literally the exact same moment. This is why I wrote the book, because I lived through that fiber thing. Yeah.
We really do lead parallel lives. I was living in New York, in Long Island. I went to law school.
I was practicing law. Right. That's what you did.
Law school, doctor, management consultant, investment banker. There you go. Yeah, and here we are.
Right. Exactly. But it was before Apache came out.
I was playing around with my friends building websites on Netscape Server. Yeah. And then Apache came out and we said, "Wait a second.
" Right. It's open source. Yeah.
" Right. We're not going to design websites anymore. We're going to buy a good Intel 486 machine.
Right. And we're just going to store websites. And as we need more websites, I'll buy more hard drives.
Amazing. And that's what I thought I was, a digital landlord. That industry became web hosting.
Right. 95 a month. And remember what websites were then.
Right. They were brochures. They were brochures.
Right. For sure. But I couldn't build...
We were using one of those dot matrix printers or whatever to build out of QuickBooks. It took us all month to build 5,000 websites. That's crazy.
But that's how I-- Same exact thing. Same exact thing. Then I sold, and we went...
It's a whole story. Yeah. It's so funny because it's also the origin story of Vultr.
You literally could've been the founder of Vultr because Vultr was literally two guys in a basement in New Jersey doing a hosting business. Yes, that's what it was. And standing up racks when people were starting to build websites.
Yeah. And then they started getting a lot of... If I may, the origin story of Vultr is that these guys basically just built a very profitable hosting business from day one, and then they started taking on enterprise workloads, and then they started actually scaling the hosting platform across the United States.
And then they got a customer, Activision, which was building its worldwide online gaming franchise back around 2006. I remember. And so then they needed to scale globally.
Wait, that's a different company. And Activision, if you're playing "Call of Duty" online, that's powered by Vultr servers at the Vultr cloud. And at the end of the day, the latency really mattered.
So what these guys basically learned how to do is, it was just a handful of engineers. They learned how to build a super resilient platform and do so with a ton of automation because they didn't really hire anyone. And Activision, if you have an extra millisecond of latency, your character could literally die.
You could die by the time- Your character could die. My trigger's not working. Your trigger's not working, right.
So anyway, so then after scaling Activision for a large number of years, back in 2014, this group of engineers finally decided to launch a new cloud platform, Vultr. And it was a platform built by developers for developers to basically scale enterprise workloads with incredible resilience. And its claim to fame was very simple, which is better performance at a dramatically lower cost.
Now, how could Vultr offer dramatically lower cost? Super simple. Everything was super automated.
You're talking about a handful of engineers running a global platform. So it's a funny little story, which is it's the only platform in tech history that grew north of $100 million in revenue profitably- Wow ... without a single dime of outside capital and no sales, no marketing.
Crazy. Nothing. You got the right product.
Got the right product. But still to this day, the reason why I tell that story, it's like we're here at QCon. This is about developers learning new things and needing a platform to build upon.
And so that's just our DNA. Our DNA is open source, open standards, developers just building. Great, we have marketing now, so I'm marketing, and I talk a lot.
But at the end of the day, the platform is just the best performance, the lowest cost, and it's the platform that gives you the greatest resilience for enterprise workloads. And that's why we're so excited to be here today. Absolutely.
Empower developers to build. I get my hair standing up on my hand. Well, what hair I have left.
So I sold that hosting company- Yeah ... to what became an ASP, an application service provider. We were too early.
Yeah. Vultr had the timing right. Right.
We were hosting Lotus Notes, PeopleSoft, all that stuff. No cloud, no virtualization- Yeah ... no ubiquitous cheap bandwidth.
Yeah. And we went public, and then we went bankrupt because we were too early. Too early.
Well, I had left by then. Yeah. My stock didn't leave, though.
That's another story. But Vultr sounds like it was right at that right time. Right.
This thing all of a sudden was possible- Right ... to build that infrastructure. Right.
And it's an amazing thing now. And let's take a full circle. Back to where we are now, it's the same stuff happening.
It's the same- Exactly ... the table has been set again. And I'll be honest with you, I didn't think I would be working or living long enough to see this happen again.
I'm with you 100%. We went through various stages of disruptive innovation since the dawn of the web. Yeah.
We had mobile, we had social, we had just cloud in general. Yeah, cloud. Like they weren't the same magnitude.
Oh, not that kind of- They weren't the same magnitude. This is the same magnitude or bigger. And so that is what, for me, so exciting.
We're rethinking and we're rebuilding an entire new stack. And it's changing cloud infrastructure, right? Sure, it is.
At Vultr, we were a platform of choice for cloud native developers. But we were also first to market for bringing GPUs to market, starting first with the A100, by the way, in 2021 for AI. Now, we were a little early on that front, but still to this day, we are a global cloud platform, where our nearest equivalent is in Amazon Web Services.
Yep. And we offer the full choice of GPUs from both NVIDIA and also AMD, and of course, all your core cloud compute, your core CPU, which is very important not just for your general purpose compute, but it's also very important for your agent sandboxes. And why this is so important is because right now here at this moment, how do we help every developer here at KubeCon build that new open stack- Yeah ...
to build and deploy real AI native applications, and how do we help them learn how to do that all? That's a lot of what we're doing here right now, today. And I can see why.
I want to hit a couple of things in that, Kevin. First of all, look, no one's saying Vultr's AWS, but this is not a data center. So when I went to the ASP, we did 30 acquisitions in 36 months.
What? How did you do that? Yeah, we raised one a month.
I'll tell you sometime over drinks. Yeah. But one of the companies I'll never forget was a hosting company in New Jersey, run by two brothers.
Right. That almost was Vultr. No, it wasn't that.
You could've done another. But Brad Feld, who was in charge, says, "Go out to New Jersey," because I was in New York, "Check this out. " Right.
I go to their data center. " I said, "Why? " I don't know.
Right. " Oh, my God. Oh, my God.
But that being said, there's scale and there's scale, and that's the point I'm trying to make. Yeah. You're not this company in New Jersey anymore.
Right. You're not AWS, but give people a sense of the scale of Vultr. Yeah.
So we have 34 data center regions across all six continents. So from a latency perspective, we like to say we can reach 90% of the world's population in as little as two seconds, somewhere between two and 40 seconds. So really, really amazing performance just across the board.
And when we say 34 data centers, some people count the number of data center regions. Some people count the number of data centers. We have a lot more data centers than that.
There's multiple data centers per region. But yeah, we have about 300,000 customers. A lot of those are individual developers with accounts, small to medium business.
And then we have our big enterprise customers. Right. And those enterprise customers are your traditional Fortune 1000, but it's also all of your AI native startups, as well as your hyperscalers that leverage Vultr in some capacity for things like large scale GPU clusters.
Sure. But the scale of the company is, basically we support some of the largest GPU clusters, and the thing that makes us a little different is we're not just selling bare metal GPU clusters. No.
That's necessary, but it's not sufficient. No, but that's a scarcity model. Yeah.
That scarcity goes away, and that model's going to go away. Yeah, and you need a lot more than just a bare metal GPU cluster. Absolutely.
You need the CPU cluster next to it. You need all of the additional cloud services. You need all of the network.
You need all the storage. You need a heck of a lot more. You need an actual full cloud platform.
You need a real stack. Yeah. So- It's more than rent a GPU ...
it's more than just rent a GPU. And I think that's the important thing to remember here, because today, yeah, you got your developers who are playing, let's call it a setup- Right ... a sandbox, whatever.
But I think people want the whole enchilada. Right. They want the whole stack to run their AI empowered, including, by the way, this cloud native stack, right?
Oh, for sure. They all run on Kube and the cloud native stack. What are people asking for that you don't have yet?
So it's actually a little bit of a trick question because we purposely focus on being the best infrastructure as a service provider, period. Period. Right?
So what we don't do is we don't go up stack and attempt to build all of the services that we might- That's not your game ... eventually provide. That's not our game.
We do one thing and we do it better- You're not platform as a service. You're infrastructure as a service. We're not platform as a service.
We do one thing and we do it better than anyone else, better performance, most resilient, dramatically lowest cost structure, bar none. So they ask us for the services at the platform layer, and so we don't do that, but that is why we have an ecosystem. I was just going to say, that's a lot of opportunity.
That's the partner opportunity. That's why we're here at CNCF as a platinum member. It's because you go to our platform, there's 60 CNCF projects you can hot deploy in a second.
So we work with an open ecosystem, and we want to help people build best of breed stacks. That's our game, open, composable, best of breed stacks. And when they ask us for things that we don't have, it's like there's a reason for that, because there are people that do it better than us.
That's their job. And they will always do it better than us. And there's actually three of them, and you might want to evaluate all three.
All right. So we get a lot of vendors, developers, our audience runs the gamut. " Yeah.
"We could be someone who adds on- We love it ... " How do they do it? How do they reach out?
Just reach out. We like to say everyone's invited. com.
I would love to meet you. I'd love to have a conversation. Just be careful because if you talk to me about doing some sort of partnership, I'm going to wind up spinning up integrations and documentation and campaigns, and you better have a- That's the marketing Kate in you ...
yeah, like the tornado begins. But I'm thinking- There is a tornado effect. This is a big floor.
It's spread out. Yeah. But almost every single company here, every one of these little stands that I see- Yeah ...
is a potential partner here. They are. So we also launched here, at KubeCon, our open source alliance.
There you go. And our open source alliance is actually all of the people behind us right here, and they're now part of our ecosystem where we document them, they're hot deployable on our platform. We're doing integrated campaigns.
We're doing joint go to markets with as many people here as possible today. Last question. There's a way to win.
We can all come together. It's the whole thing. Every single one of us needs to all work together to build this ecosystem, to build this platform.
That's the secret sauce- That's right ... of CNCF. Yeah.
Right? I mentioned in the early days of the company Interwoven, that's exactly what we did, because it was a platform play just like infrastructure is, and the way we won at Interwoven was very simple. We built an ecosystem.
Everyone integrated with Interwoven, every single person, every single database company, every single applications service company. You name it, they all worked with us. " That's the kind of the play that we're doing right here with Vultr, which is like, we want to work with everyone because at the end of the day, if you're a platform play, you need to support choice.
Yeah. You got to be open. It's got to be open.
You got to be open. But that open is one. That open is one.
That's no longer, I think, open for debate, no pun intended. Right. Let me give you the off the bingo card question that I know everybody's asking you that.
Okay, it wasn't on my bingo card. No one's asking this yet. The off the bingo card question was not on my bingo card.
When will we see a... Oh, shoot. I forgot the name of the thing.
A quantum. Have you thought about Vultr Quantum? So- In our lifetime?
I think you will actually see it actually relatively soon for very specific workload types, for very specific compute workloads. Really? Oh, yeah.
I think it's actually nearer than you think. I think in terms of any sort of general purpose, I think that's like- Five, 10 years out ... decades in the future.
But I think already today you're starting to see that there are very specific use cases where it actually can work right now today. So yeah, I'm an optimist On all things- Careful ... at the end of the day.
The advice is careful. For sure. And you know what I'm really optimistic about?
What? I'm optimistic about- Here we go ... your book.
Thank you so much. This is literally, you told me about this when we first met two months ago. " And I actually set a little Google alert, and I actually, I see your LinkedIn post saying- Uh-huh ...
the book was coming. And so the first thing I did is the day it came out, is I got it on Amazon. And then, of course, I was traveling.
It took two weeks for me to get home- I get it ... to actually get the book. And then, this was my reading material on my flight to Chennai, India.
So I would be honored- I would be honored ... if you would sign my book. Before I sign, I got to tell you the truth.
I brought about two dozen books with me. People come here, they all want the book, guests. Yeah.
I sign them, I give them the book. Kevin is the first person who reached into his pocket and bought the damn book out his own money and brought it here for me to sign. Yep.
And I'm honored. Yeah. Thank you so much.
And I'm going to sign it right now. And as you're signing that book, I have two other things to share with you. Go.
So first and foremost, you're signing with that pen, and I think we can do a little bit better for your book signings- Yes, we can do that ... going forward. So you are going to be getting an honorary Vultr Montblanc pen.
Beautiful. For purposes of all your book signings. Montblanc author right-- Okay.
And second, you're going to be getting a large order of books because this book is going to be given as a Christmas present to every single person on my team, and it will be mandatory reading. Thank you so much. So we have about 60 people on the team.
That's a big order for me. And this book will also now be part of our CEO book club, which will go to all of our top clients courtesy of- I love it ... our team at JJ Cardwell.
Wow. Hold on one second. I can't write and talk at the same time, I apologize.
I can't even write anymore. Well... I don't even know-- Have you seen my penmanship?
I literally lost-- I used to have beautiful penmanship. I literally cannot write anymore. I pray people can read what I write.
Right. But that being said, man, thank you so much. No, thank you.
That is an honor. I am out of my mind. This book is absolutely amazing.
Thank you so-- Well- I love it ... for someone like you, we lived through that fiber thing. Yeah, we lived through this.
Yeah. Right? Now I just got to finish it.
I'm only halfway through. Oh, no. I'm so busy.
The second half is all AI. Okay. It's all what's happening right- I'm happy to hear it ...
right up to the-- I even have a chapter in there about NVIDIA and what can go wrong there. Yeah. I didn't foresee what, but there's a lot of great stuff in here.
I can't wait. I highly encourage it. I'm halfway through.
It's life-changing. Thank you so much. That's okay.
Sorry for the lie. No, no, this is not a children's- Okay, thank you. Okay.
I apologize to all the children out there. Kevin, a pleasure seeing you in person. It's great seeing you in person.
Dinner in Florida when we're both there again. When you're in Florida, we'll get out the Boca, we're going to go on the boat. We will talk.
We will talk. On my boat. We're going to dinner from there.
Great. Done. Done.
Kevin Cochrane. Oh, wait. Vultr.
V-U-L-T-R.