 an error connect right after your keynote right here. So what are you talking about? Well, so today I was speaking about what the next 10 years of infrastructure in the internet would look like in our opinion. Because you're doing a cloud company in an era where Amazon is giant and Google is giant. Which one is bigger? Amazon or Google? They're pretty big. Well, I mean, in terms of their cloud platform, Amazon is definitely the bigger one. So Amazon is not just selling stuff. They have this huge thing on the internet with the cloud. But does your cloud differentiate from theirs? Yeah, so Amazon Web Services has a product called EC2, which is Elastic Compute. And what that is is essentially a virtual public cloud. So what you get from them is a virtual machine in the end is really what it is. And what our cloud is, it's a bare metal cloud. So it's dedicated servers, looks and feels to the end user, very similar to Amazon Web Services. Where you request a server over an API, you get it in a few minutes, it's billed by the hour. But it's a fully dedicated server. So no hypervisor, no virtualization, no virtual machines, totally dedicated bare metal server. Now a lot of people will bring their own hypervisor or load their own virtualization software on the dedicated servers, but we don't do that by default. So people could buy some bare metal servers at your place and then provide a service like Amazon Web Services if they want. Yeah, absolutely. And we do have some sort of niche cloud providers that, yeah, exactly, get dedicated hardware from us. They put their own virtualization software on there and then essentially resell it. And maybe one of the initial innovators in providing ARM servers as a bare metal service. Yeah, absolutely. Especially in the data center, ARM has its roots in the embedded space, obviously. So generally in phones or other embedded devices, the Raspberry Pi is a very well-known one that runs using ARM chips. And so, yeah, we're the first cloud to provide an ARM server that's basically an enterprise-grade, data center-grade ARM chip alongside our Intel chips. And if this works out, if all the software's there, if all the work the NAR is doing, if it works out, everything, it could really explode, right? It could get really big, very quickly. Yeah, absolutely. I mean, I think once the software is solid and people have confidence that it's just gonna work out of the box on the ARM architecture, you're gonna see people that are just gonna be too attracted to the price and the performance and the features of these different ARM boards and chips that are coming out right now. Qualcomm, Cavium, both have an enterprise-grade, data center-grade board and chip. And the economics there and the price to performance just completely blows Intel out of the water. And so really, in my opinion, it's just a matter of a software adoption. And so, you're the founder of Packet or co-founder? I'm one of the co-founders, that's right. So there's three of you. So how do you sit down and get the idea? There was not so long ago, right? That's right, yeah. So we started the company in 2014 and I was running a managed hosting company. And I was actually a customer of Zach, who is our CEO at Packet. And he had sold his previous company, but we had done a lot of business together and got to know each other quite well. And so after he had sold that company, yeah, he approached me in, I think, late 2013 or early 2014 and said, look, I think I see that there's an opportunity in the market here for either people who are sort of in a traditional, either co-location or dedicated server environment or at scale in a public virtual cloud like Amazon that aren't being serviced well. So if you're in the traditional rack and stack or co-location environment, you're not really able to take advantage of the elasticity and the flexibility and the scalability of the cloud. But you're kind of stuck there because on the cloud side, especially at scale, while you have a lot more flexibility and you have a lot more of that scalability and on-demand programmability, it's very expensive. And it generally suffers from performance inconsistencies, generally poor network and various other problems. And so, but if you're in the cloud over there, you can't generally give up all of those nice things about the cloud to move to a dedicated server environment just for the economics. So we felt like if you could automate that hardware and get the price and economics and performance closer to a traditional dedicated hosting environment, but with all of the elasticity and programmability and ease of use of the virtual public clouds, that you'd really kind of like hit that sweet spot. And so that's what we've done and that's kind of was our bet there. So what's the hosting company you were working at before? Well, it was a small boutique company that was part of a web development agency that I was the owner of and that worked primarily with governments and nonprofits. And the company that's axled was a company called Voxel.net that was a dedicated and virtual hosting company that internet bought in 2010, I believe. And so there's a lot of work to be done in the future or you already achieved a lot. I mean, you have SoftBank and Dell, why are they so happy to work with you? Yeah, well, absolutely. Do you have something special going on? Yeah, absolutely. There is not a cloud like us that exists. So, especially if you look at the integrations that we have, so we're integrated with all of the kind of popular infrastructure orchestration software out there. So there's Terraform, Ansible, Docker, LibCloud. We have a bunch of different API clients and all the kind of popular programming languages out there. And so to be able to get bare metal servers at the end of an API like that using that software is not something you can get anywhere else. And so it was actually a really nice fit as we were closing our funding round with SoftBank right around that same time they acquired ARM. And we had already been talking internally about bringing an ARM server to market because we're very bullish on it. And so that was actually a great fit and it ended up, you know, we ended up bringing that Cavium ThunderX chip to market in October of last year. That's been received very well. And now actually work in collaboration with ARM on the Works on ARM project, which you can see at worksonarm.com. And the goal of that project is to foster software adoption for ARM and to kind of bring multi-architecture support to a lot of the foundational software that's out there in the world running most of the internet in the hopes that once the software gets there, the adoption of ARM in general for internet-based software is gonna take off. So you have 50,000 deploys right now every month that's pretty high activity, right? So who those kind of customers that are deploying on your, what kind of people really need what you have? Yeah, they, you know, they range from Fortune 500s to SaaS platforms, to gaming companies, some IoT companies. A lot of people are using us for their CI CD pipeline. We have a bunch of basically continuous integration platforms and image building platforms and testing platforms that use our hardware for their build systems. So it's a wide variety of things. You know, we have people that are using everything from VMware and then KVM virtualization to orchestrating containers with Kubernetes to running containers directly on the bare metal. So it's a pretty good mix. And they wouldn't be able to do all that stuff on a traditional, virtualized kind of hosting service. Well, some of those things absolutely not wouldn't work. You'd be essentially running a hypervisor within a hypervisor and there are certain workloads that you can't, there are certain software that won't run inside of a hypervisor at all that need direct access to the hardware. And for the workloads that you could run in a virtual environment, like you can certainly run containers and orchestrate them into virtual public cloud like Amazon or Google or Azure without any problems. You're just paying a little bit of a hypervisor tax because you're running that container within a VM which is running inside of a hypervisor on the bare metal, whereas with us, you can run the container directly on the hardware. Can you mention a little bit what you were doing in the past in the political campaigns? I'm sure, yeah. Because it was like, you were part of tech teams, right? Yeah, that's right. So I was a web developer for Howard Dean's presidential campaign in 2003, 2004. Like making PHP kind of stuff on the web? Yeah, that's exactly right. Yep, yeah. So I was part of the state web development team and so I was building everything from ride board, ride sharing volunteer coordination software to door knocking software to content aggregation. So you did ride sharing in 2004, like six years before Uber? Yeah, not exactly the same thing. Okay. But yeah, helping people hitch rides with each other to events and that sort of thing. Because there's a lot of tech infrastructure under these huge things that are very famous, like those presidential campaigns. They have some tech going on that's quite complex or needs to have Burmese servers or what? Yeah, they do. Yeah, there was a whole ecosystem of software and platforms that is totally just focused on organizing events and campaigns, absolutely. And it ranges from very small out of the box, I'm running for school board and want to pay $10 a year for it all the way up to the hundreds of millions of dollars that are spent on the big presidentials. And when people are a customer in your system, is it easy for them to load balance and manage huge traffic peaks? They just buy more automatically and just make sure it works and stuff? Yeah, absolutely. Or spread around different data centers? Yep, yeah, absolutely. We don't have a hosted load balancer service but if you're tooled up in a microservices framework and you're using Kubernetes already, yeah, absolutely. You can request more services on demand. We are a global platform, so we have, I think, 15 locations now. So we definitely have people that are doing multiple region facility deployments for various reasons, either just straight up load balancing or DR. How long does it take to deploy an extra capacity? Yeah, I mean... You might get out of stock. If people need a lot, suddenly. We definitely have our limits, so that does happen sometimes, but we spend a lot of time doing the capacity planning and capacity management to try and keep out ahead of that. We have a spot market, which basically lets you get servers on an open priced market, but allows us to basically claw them back at any time. So those are for workloads that obviously are okay with all of a sudden a server disappearing, but that allows us to basically have what would otherwise be unused capacity, but sell it essentially below market, and then if we have demand somewhere else that's basically at our list rates, we can claw those resources back and allocate them to paying customers so that there's some levers that we have in the platform to kind of help with that capacity. And do you have unlimited bandwidth? You have really good fast internet. There is no such thing as unlimited bandwidth, but I mean, yeah, we run our own fiber backbone, and so you pay for the bandwidth that you use, but yeah, we don't rate limit it at all, so most of the higher end servers that we have have dual 10 gigabit Nix in them, so you get a 20 gig line rate network port, and you're free to saturate that as much as you want. And bandwidth is not so expensive? It's, I wouldn't say we're on the lower. And it keeps going lower and lower? Yeah, yeah, I mean, I wouldn't say, the more you use, the cheaper it gets. Yeah, I'll put it that way, and I wouldn't say that we have the cheapest bandwidth in the world, it's definitely like, we pay for peering, we're very aggressive on the network side, so you're paying for good bandwidth, but we're also not the most expensive in the world either. All right, so looking forward to what's gonna happen in the future, hopefully the ARM server is gonna rule the world, and you're gonna be ruling that world too. Yeah, absolutely, it's gonna be the year of ARM in the data center.