Podcast Archive - StorageReview.com - Podcast #123: Cloud-Based GPUs For AI

Episode Date: September 26, 2023

This podcast features a conversation between Brian and Jeffrey Gregor (he prefers Gregor).  Gregor… The post Podcast #123: Cloud-Based GPUs For AI appeared first on StorageReview.com.

Hey guys, thanks for checking out the podcast again. We've got a great conversation with a company called OVHcloud. They may not be the biggest cloud provider in the world, but they are very well established and one that you should definitely know about. We just did a paper on their GPU instances. We'll talk about that in terms of what that means in AI development, most importantly, and also how you can get started with OVH, what their services are, and all sorts of other things. So Jeffrey Greger, or Greger as I'm told you prefer to go by, is here from OVH. Thanks, Greger, for doing this with me. Yeah, great to be here. Very happy to talk to you and whatever live audience we have on the internet. Yeah, absolutely. So thanks for setting that up, Gregor. We are streaming this

live to our Discord audience. So if you've got questions, feel free to post them in the chat. Jordan's moderating and we'll make sure to get the good ones to me if you guys come up with such a thing. And for you guys catching this on iTunes or whatever your podcast tool is or on YouTube for the video, we do have a Discord highly active where you can participate with our team in real time and a couple hundred other friends that are into IT topics. And that is linked to from the top right corner of storageofyou.com and I'll put a link to it in the description as well. Less on me, more on you, Gregor. I set this up as OVH is a cloud that we've known about, of course, for many, many years. The GPU project was the first time that we'd worked together on a review. So I'm thrilled to have

gotten that done and for what we can do with you guys in the future. But set up for the audience, who is OVHcloud and what's your target? If you can do that in a few minutes or less, that's a big question. Yeah, sure. So yeah, so I expect that this is a mostly or maybe exclusively US audience, or maybe it's international. But as you mentioned at the top, OVH Cloud used to be known only as OVH, but we've been OVH Cloud for several years, is, depending on which survey you look at, the seventh or eighth largest cloud provider in the world. And we're the only one that's in the top 10 that's headquartered in Europe. So I'm the general manager and president of OVHcloud US. We're a Delawarean corporate LLC. So we're our own

separate company, but we're a wholly owned subsidiary of OVHcloud in France, and they've been doing business since 1999. And like I said, globally, they're the seventh or eighth largest cloud provider in the world and the only one that's not a U.S. or Chinese company in the top 10. We've been operating the U.S. company since – well, we started selling in 2018. We actually started it. It was incorporated late 2016, started hiring selling in 2018. We actually started it. It was incorporated late 2016, started hiring employees in 2017. And both OVHcloud in France

and OVHcloud US are in themselves global companies. So about a third of my customers are from outside of the US and we serve US residential customers and we serve anyone in the world who wants to buy our servers or services here. And we also can deliver services in any of OVHcloud's global data centers.

So North America, US, Canada, Europe, in the UK, Germany, and France, India, Singapore, and Australia. And you might have seen some news recently, we just made an acquisition of a company called Gridscale out of Germany, and that's gonna allow us to put our services in a lot more edge locations that we'll be rolling out soon. Last thing, maybe just to close out this question is,

so what you need to know about OVHcloud is we're a pure play cloud provider. All we do is cloud, but unlike other competitors who are mostly either focused on public cloud or private cloud or some niche folks who do just bare metal, we do all the clouds and we always have. So we're very strong in bare metal. And that, you know, was our basis from, you know, many years ago. We also offer private cloud, which is, you know, dedicated resources on, you know, industry standard sort of stacks, like we've been long term partners with VMware, we're partners with Nutanix. So if you need dedicated resources, for both, you know, physical and logical isolation, bare metal or a well-known

stack on VMware or Nutanix. And then we also run our own public cloud. Again, I'm not 100% sure, but it either is or at some point it was the largest operation of OpenStack in the world. And we're really focused on public cloud because that's where we know the market is growing very fast and it's very easy for people to get in and use public cloud. We're also building some key platform as a service offerings like databases as a service

and AI and machine learning tools on top of it. So that's what you need to know about OVHcloud. Okay. Well, you did well. tools on top of it. So that's what you need to know about OVHcloud. Okay, well you did well. I said in a few minutes and you hit that target. So I get the scale and that's important. I'm glad you set that up. When we think about your big competitors, your big global competitors, and I know you probably don't want to name them and go head to head, but when we think about AWS, Azure, GCP, the common ones here, at least in the US, many of them have made their focus to have a little more targeted about where you make those investments to make sure that you're able to implement support and that obviously customers want to pay for those. When you think about some of these new services like the GPU offering, which we'll get into here in a little bit, what does OVH have to do there that may be different from your massive competitors

or maybe even different from your much smaller, you know, regional CSPs or whatever that might kind of dip a toe into this kind of thing as well? Sure. Well, I think, you know, even Google and Microsoft and the other hyperscalers, you know, realized and admitted, you know, years ago that AWS had, you know, such a large start. And I don't think anybody is really trying to match them on item by item. And we're certainly not. So we are trying to give our customers all of the right building blocks that they need. And we certainly support multi-cloud. We know the research that's out there, both from talking to analysts and folks like you

and our customers and others that, you know, most companies are using at least two cloud providers, if not three. And so we think we have a place in that multi-cloud environment. Again, we're strong maybe in some areas where others aren't, particularly on bare metal or some of our private cloud offerings. And then on the public cloud side, yeah, we're not trying to give every microservice and everything that the hyperscalers or even some other providers might have. We're trying to give the key building blocks plus the main platform as a service offerings that we think our customers in the market need.

And like I said, those are mostly around databases of service and of course, AI and machine learning and notebooks and things that support data lakes and various different types of storage offerings is the sweet spot for us right now. So we're not trying to be everything to everybody, but I think we certainly fit in this ecosystem. And another key aspect of how OVHcloud operates is that we don't lock our customers in, even though we have offerings, like I said,

with VMware and Nutanix, and we offer Microsoft Windows licenses on our bare metal servers and different things. Those are sort of the de facto industry standards, but we really support open source software. And even through acquisitions that we've done, our founder, Octav Klava, has a vision to generally productize but then open source some of that software that we get from companies that we acquire.

So our customers are free to come and go and use us for the workloads that are appropriate for them. And if they need to put other workloads elsewhere, we recognize that it's a multi-cloud world. Well, I mean, you're not wrong about that, of course. And your strength in Europe has got to be nice for your U.S.-based customers to know that if they're with you here in the U.S., that you're as strong or stronger maybe in, I don't know, say, France or like you mentioned for the HQ. What do you guys think about in terms of footprint of where these data centers are? Are you expanding that through new investment or is that largely through acquisition? It's generally through new investments. So OVHcloud, we're a build-it-here company and we're very

vertically integrated. So another thing to know about OVHcloud is again, unlike most other providers, we design and build our own servers. So we sort the components, but it's OVH cloud designs. We do everything from bending the sheet metal to putting the racks together, building the servers, cabling them. And then we own and operate our data centers around the world. And I gave you an overview of the locations before. But in the coming years and months, we're planning to expand rapidly. So at some point in the future, we're going to add a central U.S. data center, probably in Texas or somewhere in the central region. We

have a data center in Virginia and we have a data center in Oregon. So we've got the east and west coast of the US covered. We've had a data center in Montreal in Canada for about 10 years, a little over 10 years. And early next year, we're launching a new data center in Toronto. So we'll have four North American data centers. We just opened a data center in India, which I mentioned earlier in the

call, and we're looking to expand our Singapore and our Australian data centers, as well as some additional data centers in Europe. And also, like I mentioned earlier, with this acquisition we've just done of the German company called Gridscale, they're going to allow us to get out into other locations where we already have points of presence. So we've got many points of presence around the world where we don't necessarily have a data center full of servers that we're operating. But that will very quickly allow us to expand our reach, you know, first by tens of locations and then eventually by hundreds. Well, I mean, I'm a hardware nerd, so I kind of stopped listening

after you said you bend your own metal. I mean, that's somewhat unheard of anymore, right? Because you've got the OCP movement, of course, you know, that Meta and Azure and some of the enterprise guys are supporting, which is one thing, but that doesn't necessarily mean they're making what you want to consume on the hardware side. Then you've got a lot of the ISG, the server guys that are trying to build solutions exactly for your scale of hyperscale to give you pre-packaged solutions. And obviously,

those aren't working for you either economically or feature set. Tell me a little bit more. I mean, I realize you may not be the metal bending guy, but tell me a little bit more about that program and what you're doing there and why you feel like that gives you guys a competitive advantage. Sure, it's a couple of things. So it allows us to control the whole part of the supply chain from when we get the components through, like I said, when we build it and deliver it to our data center. It also allows us to come up with the configurations that our customers need in the market. So we can design these to have various different configurations of RAM and disks, from spinning disks for a lot of slower storage up to the latest NVMe drives. So we offer a wide variety of configurations. And the other reason that we do it is we've

been operating since around 2000 or the early 2000s, our own proprietary water cooling. So instead of air conditioning all of our data centers, we use water cooling directly to the chips and other components on the board. So it's very efficient. Most of it operates at ambient temperature. Although in the summer months, we do have supplemental cooling, but it allows us to be very environmentally friendly and have some industry-leading power usage effectiveness,

water usage effectiveness, and even carbon usage effectiveness, PUE, WE, and CUE. We've got commitments to be a leader in that area and also with this vertical integration and cyclic internal

lifecycle for our servers, we also give multiple lives to these servers. So every 18 months or so, we refresh our business and enterprise class of servers and come out with the latest CPUs and fastest RAM and drives, et cetera. But then we don't just get the old ones and we have a commitment to not put you know you waste to landfills and in the next coming years we repurpose those existing servers maybe we remove some of the features adjust some of the you know

feeds and speeds but we you know reprice them and we keep offering them you know in different configurations for several years down the line because we know that there's a huge market out there and not everybody needs the latest and greatest for every application or workload or not everyone can afford it. So if you're an individual just trying to learn to be a sysadmin and you need to get your hands on a bare metal server so you can use IPMI and do some other things to get yourself set up, then you can do it. Or again, depending on the workload, we've actually had a lot of customers come to us asking for four core servers. And hey, it's not the 1990s anymore. Like most of the servers have tons of cores, right? But, you know, some people want

four core machines because of, you know, databases and other software that, you know, is very punitive on charging you for every core. And so there's still some use cases out there for that, even if it's a several year old server,

you know, somebody's got to use for it. I'm laughing a little bit because we're seeing eight cores or more on DPUs now, which are net add-ins, which is kind of funny. But you're right. And you said early on in the conversation that OVH didn't want to penalize customers with long contracts and give you some flexibility. What you didn't say, but you started to say there, though, too, is that in my experience with OVH, you guys tend to be much more cost effective than other solutions, right? And maybe some of that's because of the flexibility

to come in and out without having to contract a big chunk of credits that you burn against, which is a common modality. So I don't want to position you guys as always low cost because I'm sure customers will spend plenty of money with you, but you do have a lot of low cost options, which I think from the, like you said, the experience from training, from experience, from getting hands-on with these things, if you can do it for a couple bucks here and there,

you probably don't care if the system's a little bit older. It doesn't make any difference. You're still learning how to be a practitioner of some of these services and setting up, you know, you're in Kubernetes or database, as you said, or whatever you want. It's a great way to have access to a pretty robust playground.

Yeah, again, that's the view of our founder, Oktav Klaba, like one of his sayings, and his sayings and one of the sayings at OVH Cloud is innovation for freedom. And he really wants to let everyone in the world build their freedom in the cloud. And so we can offer something at the appropriate price point for everyone, you know, along the spectrum, that's a benefit to us and a benefit to them. And yeah, I certainly don't position OVH cloud as, you know, cheap or low priced, although we do have very competitively priced offerings, and some of them are low, know we look at at look at it as delivering high value and high performance for uh price and that you know uh again allows us to address this you know long tail of the market and also allow some of our partners to use our infrastructure

with their value-added services on top of it and uh you know offer it to folks who might not be able to pay enterprise sort of rates. And just on that point, our customer base ranges from people who literally do pay us $1 a month for a very small VPS up to enterprise customers who pay us $1 million a month and everywhere in between. Sure. But right now in the US, our sweet spot is, you know, SMBs and smaller enterprise customers who pay us, you know, say between, you know, 50 and 200k per month. But, you know, anywhere in between, we certainly service lots of customers that pay us 10,000, 1000. And like I said,

lots of individuals, you of individuals that maybe just have a VPS or an older bare metal server that maybe pay us $50 a month. Well, there's two things there. When you're talking about building your own servers and being able to adopt the latest technologies, Gen 5 Flash, high-speed interconnects, which in your business is probably extremely critical when you get into talking about shared storage and things like that. That's at one end of the high performance scale. And also, as you're mentioning, the fact of the matter is that most SMBs, SMEs, robo deployments, anything edgy really doesn't need that much performance.

What they need is reliability. They need uptime. They need cost effectiveness. They need that much performance. What they need is reliability. They need uptime. They need cost effectiveness. They need ease of use. They need a bunch of other stuff that the cloud can deliver, doesn't always, but can deliver. And it seems like that you're finding a sweet spot right there. Well, maybe a couple of different sweet spots, depending on the persona of your customer. Yeah, I think it's the right price for you, as you're saying, what the market or the customers need. And also, I think we've got a very compelling story

for helping people move to the cloud. And that's one of the things that I think we position ourselves well against the hyperscalers and some other ones because years ago the cloud was what AWS said it was and a lot of people said, oh, well we need to move the cloud but they didn't really understand

that moving to the public cloud requires you to do some refactoring and changing how you operate your workloads and then especially even more so with Kubernetes and containerization. But with our broad range of products, people who want to get the benefits of the cloud, which are predictable pricing, great performance, monthly subscription payments instead of upfront, capital expenditures every couple years,

etc. You can do that with a lift and shift to our bare metal cloud. You can do it if you're used to running VMware or Nutanix on-prem or in a colo, you know, zero skills gap, because we know there's a lot of skills gap in the market. And for companies who do want to modernize their apps or build modern apps and there's a you know a lack of skills out in the market so as I like to you know joke and say you know you there's plenty of these you know like me graybeard sysadmins who know their way around linux box or a Windows server and you know you can take that first step and get a lot of the benefits from the cloud even if you're not ready to modernize your apps and then you're already in the cloud

in an ecosystem that you understand that you're not locked into and then you can work on moving over to public cloud or Kubernetes. And even on that point, Kubernetes, we offer, you can run Kubernetes on everything we have. We've got, you know, a fully CNCF certified, you know, managed Kubernetes service on our public cloud. Of course, you can get Kubernetes on Bare Metal if you want to install it yourself, any flavor. And then there's Tanzu on VMware and Nutanix has their own version of Kubernetes. So

if you want Kubernetes, you can do it again, public cloud, private cloud on data. Bunch of different ways, yeah. Whatever skill set you have. Right. Hey, so I also want to go back to what you're talking about on the sustainability side. Obviously being liquid cooled gives you some advantages for all of the efficiency that you were talking about. But reusing the servers for as long as possible is something that's become a hot button lately, maybe a little more from get storage from going in the shredder to being securely erased and repurposed and so that it can be used again and have future life. But the same can be said for GPUs, for servers, for other technology that

just because it ages out and may not be as economically efficient or fiscally as efficient as whatever new servers you're putting in there, shredding that just seems like a tremendous disservice to, not to be hyperbolic, but to the planet, right? I mean, that doesn't make a lot of sense to me. So I'm glad to see you guys trying to get more life out of these systems. Can you talk at all about what you do at the very end of life of these systems? Is there any recycling program or something a little more harmonious, I'm hoping,

than just chucking it in a e-waste pile? Oh yeah, like I said, we have a commitment in the next couple years to not send any of this to e-waste. So yeah, we certainly try and, yeah, again, use it and get the maximum amount of life out of it and then recycle as much as we can. And I do just want to make one other point is that this isn't something that we've just jumped on the bandwagon because sustainability and green and, you know, sort of greenwashing and all that is, you know, the hot topic of the day. This has been in the company's DNA since the beginning and

you know like I said we've been doing the water cooling since you know the early 2000s and we didn't even know how good we were at this because again like it's just how we're used to operating. The parent company in France went public a few years ago, about 18 months to 20 months ago on the Euronext exchange in Paris. And when they were doing the due diligence for that and all the road shows and everything, that's when they realized, the investors and the banks and everyone were telling them, do you guys realize what a great story you have here? This is just like what you do and this is what everybody's like trying to

get to and you've just been doing it the whole time you've existed. So I think that's a great part of our story. No, it's certainly compelling. Of course, when we talk to enterprise customers that are figuring out how to make their data centers more efficient, especially when it comes to GPU servers, which as you know, can be thousands of watts just in their own box. It's daunting. So the HPC guys, educational institutions have been on this for some time now, but the enterprise, I don't know how many more generations of air-cooled, pure air-cooled servers, and especially GPU servers there will be. It can't be too much more until something has to be done there, whether it's a closed radiated loop, whether it's a liquid door,

whether it's full immersion or full liquid loop. I mean, there's a lot of choices, but it will come to it at a certain point. I've got a question here around security and I want to ask you here too, in terms of you're building all of your hardware is one thing, securing it is another. Can you talk a little bit about your security approach? And then I've got a follow up for you. Okay, sure.

I mean, we're certified to all the standards, SOC and ISO, and we've got for all of our product offerings here at OVHcloud US, we're certified at the infrastructure level what we deliver to our customers, certainly not to the applications that they're running on them. But what we deliver, you know, PCI DSS and HIPAA compliance. Of course, we've got, you know, our own physical security protocols. So yeah, just like any data center, we've got all the physical and logical security to keep your data safe. And I mean, at this point, having the appropriate security and certification so customers can be confident in the safety of their data is table stakes. Nobody's gonna be in the business

if they can't deliver that at this point. Well, one of our guys in Discord, I think echoes your sentiment and says, "'Since OVH is known to handle DDoS attacks exceptionally well, I wanted to ask how the network can basically handle everything and what's the process of patching new vectors? I mean, you talked a little bit about it right there. I was unaware that you guys have such a great reputation when it comes to something as specific as DDoS attacks, but clearly that's a focus for you guys. And I mean, it's a focus for everyone, but it's a fair question. Like, how do you keep your eye on the

ball from enabling all the services we've been talking about and then layering on these additional and emerging and continually changing cyber attack issues at the same time? Yeah, well, again, it's not a plant, but whoever brought that up, I'll just make a few words about it. But yeah, that's another feature that we have. It's a homegrown system, our anti-DDoS solution. And I think you can look it up on the internet. It was maybe three or four years ago,

somewhere in that timeframe. Yeah, our anti-DDoS solution has withstood some of the largest attacks on the internet and protected our customers. So that's something we're proud of and something that comes with all of the servers that we deliver to our customers. And then, yeah, of course, internally for all of our systems that we use to support and deliver the services to our customers, we're always vigilant and have lots of very good internal controls to keep those safe and make sure everything is patched. And then really,

because we're an infrastructure as a service and a platform as a service provider, you know, what we deliver to the customer goes to that level. And I think really where a lot of the vulnerabilities come in is more than on, you know, the end user level that they're responsible for, you know, keeping their applications and what they install on our infrastructure and platforms safe, but we do everything that we can to make sure that what's delivered to them is patched. And then the services that we deliver that we operate the backplane for,

like our VMware solution or Nutanix or, of course, our OpenStack and platform as a service offerings, of course, you know, we're managing the patching and things for those for the customer. But if they're buying, you know, bare metal or some other, you know, dedicated resources, then of course they're, they have to manage, you know, the application level. Yeah. Well, yeah, it's good. So I'm glad that that came through and you're able to share some of that history. And I guess, too, I mean, this goes back to, again, I'm a hardware guy, so I get stuck on this.

But when you bend your own stuff, you select your own parts, you standardize on some things, it does make it easier in some ways to be able to ensure the root of trust there to build that out to make sure that you're testing and validating firmware and then pushing it to hopefully fewer different components across your entire stack. So we started this project with you guys on the GPU servers. I know that's a relatively new offering for you in OVHcloud US anyway. You've got the Nvidia V100s in there. Jordan on our team did some of his AI work through that instance, those cards.

Also Vince from our video team was in there messing around a little bit too. And both of them came away with like, this is pretty neat, easy to use and really cost effective. I think all of those themes came through in our review, which we'll link to in the notes for anyone that wants to check out the GPU instances that OVHcloud offers. But talk a little bit about the GPU instances and why that's important going forward for your customers. Yeah, I think it really ties into, or ties together everything we've been talking about.

So that's certainly where the interest is in the market right now. And what we're offering at the moment is just, you know, the bare GPUs. So they're part of our public cloud stacks. You get them in a public cloud VM with one, two, or four GPUs attached to the instance. And to tie in the story, so they're an older model, right, at this point. So they're an older model, right, at this point. So they're V100S, but they're still really appropriate and set up to do the key things like, you know,

learning or, you know, training the large language models and, you know, a lot of the things that customers are interested in right now with the craze of AI and machine learning. And we are going to be coming out with, you know, more recent H100s and A100s. But everybody knows, I think, that those are A, very expensive, and B, very difficult to get. So we're making a great offering out to the market that it's not the latest and greatest. But if you can't find anything else and or you can't afford the latest and greatest, these will definitely get the job done. And the way we have them priced again, I think it's a very compelling offer. So they're a little slower, but, you know, the price offset, you know, I think gives the performance, the price ratio, you know, well, well above one compared to, you know, something that's newer or faster. And then looking forward...

Well, yeah, before you go forward, on that point, just to make it abundantly clear, is having access to basically the graphics DDR is really what you're after. And so these V100s still have a reasonable footprint there. And for anyone, any small org, any org at all, that's developing models on a workstation of some kind now that has some GPU or limited GPU in it, even an A4, 5, 6,000, and you want to take that to the next level, that was our use case with what we were doing here, saying, look, all of this action is happening on workstations. Much of it's happening on workstations. So if we segment out HPC, if we segment out the top Fortune 100, 150, whatever, and then look at everything else, a lot of that's happening on workstations.

And so the cloud, and especially the instances that you guys offer, yours are a super cost-effective way to take that model to the next level. And I don't think you always need the latest NVIDIA or Instinct or Gaudi or whatever to be able to bring those to life. It's really about AI as a progression. And there are instances like this out there that a lot of people don't know about. So that's why we were excited about it. And I know you said that already, but I really wanted to hammer home the point that this isn't just an alternate GPU instance. This is a perfect instance for a lot of use cases to go forward with. Yeah, thanks for that.

I appreciate it. And again, yeah, there's no lock-in. There's no long-term contract. You can get it. We've got this great pricing for you just pay for it by the hour. If you want to commit to a whole month, it's cheaper. And if you want to buy high volume and you've got some large training sets that you want to run for a couple of weeks or a couple of months, you'll call us and I'm sure we can make a deal. Absolutely. But you're evaluating, I mean, you must be all the time looking at new GPUs, new accelerators, new, I mentioned DPUs before. I mean, there's so much stuff out there. It feels

more dynamic than ever in terms of networking interfaces and protocols, GPUs, FPGAs, smart this, smart that. Your team always evaluating these things, but they must be either having the time of their life or the living nightmare to sort through not only the testing and validation, but as you said, the access to even be able to, a couple hundred units of whatever it is that they're that they want to adopt. Yeah, we've got some great and in a in a positive in a positive way. Yeah, like some really geeky, nerdy, you know, great engineers to Yeah, they love this. They love building and testing and, you know, you know, evaluating what the next products will be. And what we're looking at in addition to just offering,

like I said, the instances with the GPUs for people to use directly is we know a lot of people are very interested in large language models and they want to run it on their own data, but they don't want to run it on, don't want to expose their data to these services in the cloud. So as part of further platform offering

that we're going to come out with, we're looking at building out that offering where you can have a large language model that you can have operate on your own data. And again, that's a huge thing for OVHcloud. Like you're always in control of your data. We don't move it and put it somewhere that you don't know where it is or make backups and put them offshore or something. So we respect our customers' data and they always know and are in complete control of their data. And so that's why, you know, some solutions that we're looking at coming out to

the market in the future are in that sort of vein. Well, it's funny you bring that up because the podcast that will air right before this one is Jordan from our team and I talking AI. And we talked about this specific part of the progression too, is that, you know, how do you get started in AI? And most of us know AI through chat GBT or one of the image generators or something along those lines. But you're right that organizations are apprehensive to put their data into those models in the public version, but there are APIs, there are other ways to consume those things. And what you're talking about, it sounds like a productized kind of way to do that. And that will be very interesting. So certainly keep us in the loop as you progress

on that. I think a lot of smaller organizations especially would be very interested in that. Yeah, it's a cloud AI machine learning model, a know catchphrase over in Europe and it's starting to get some traction here Although it's probably not the right term is you know data sovereignty They really want to be in control of their data And so you know we'll we'll guarantee that you know if you run it on you know these resources with OVH cloud You know you're in complete control And you know we don't look at our customers' data or even do, we don't even do any AI

or machine learning ourselves, like looking at their metadata, trying to understand anything about how they're using it. That's just not our mindset or how we're designed. We really respect our customers' data and their privacy. Well, that is another key point on the privacy side. I mean, I think it was Zoom that got in trouble for that a couple of weeks ago, making some sort of intimation, or maybe it was a full-fledged statement that they were doing that kind of

thing. And the customers are all like, well, hold up now. Are we sure? Yeah. Okay. So it's not the conversation itself, but the metadata. What exactly does that include? And what sort of inferences are you making based on that

data? And that's very uncomfortable when you talk about, and there's a perpetual fear of trade secrets being stolen, IP being harvested, especially if you're in the physical manufacturing of anything or even software. I mean, we see this every week, there's another breach or leak or competitive espionage kind of thing going on. It's relatively scary from that front. Yeah. So I'm glad to hear you're not doing that. Yeah. Well, and yeah, we also always say, yeah, like I said, we're a cloud provider and that's all we aspire to be.

So we're not trying to compete with our customers. So we want to deliver them the cloud that they can use and then build their service and deliver their value add to their customers. And yeah, we're perceived that, you know, there are stories out there about other, you know, large cloud providers who might be doing some of that metadata mining or whatever, and then say, oh, hey, yeah, this customer of ours, they seem to be pretty successful or getting a lot of traffic on this thing. And then, you know, in a couple weeks, you know, they've got their own version of whatever that, you know, tool or software is. That's not how we operate at OVHcloud. That's good to know. And I would tell you that I'm

getting some feedback from the Discord team and other messages that a lot of these guys up in Toronto want to come see your data center to the extent that I don't know where you are in the process there, you guys probably don't want to show too much, but if you do, let me know. We'd be happy to send a crew of cross between AI nerds and home labbers that want to come check out your operation. Okay, great. Well, yeah, we can certainly explore that in the future, but I think we have, and we might need to, you know, do a little more editing on it. But I think for an anniversary that we did, it was the 10-year anniversary of the data center in Montreal last year, or about 18 months ago. I think we did a virtual reality, you know,

run-through of the data center. Let me see if we've got that and maybe we can post it or share it with you and you can share it with your Discord followers. Absolutely. That'd be great. I've got another question about GovCloud or FedCloud, depending on where you are. Do you have that support for that or other

super high security options for that type of critical user data? Is that available or something that you guys are evaluating? We don't yet here in the US, only because it just has not come time for that in the company's history. We've only been operating for a little over five and a half years. So yeah, we don't have FedRAMP and we're not currently pursuing that. But I'm sure in the future we will have it. We do have a lot of customers who are, you know, SLED, state and local government and higher education that don't need the full FedRAMP certification.

And like I said, we've got HIPAA and PCI DSS and some other certifications. Now over in Europe, that is a big play for them. But again, they've been around for 20 years and they're really well known in France. They do a lot of government and public sector work in France, and they do have specific offerings that meet those types of need over there. One of them is called Secnum Cloud that they have for, again, French government and public sector and military and things like that. So in the future, I'm sure, but we're still evaluating the market demand for the sort of customers that we're servicing or going to service in the future.

Okay. So anyone can go check out the OVHcloud website or the OVHcloud US, depending on where you are. We do have a lot of US listeners, but also global as well. But I do have one last question for you. You talked about the guys spending a dollar a month. How can young people or people that are reskilling that want to learn how the cloud works, how some of these instances work, or how can a small business kind of get engaged with you to start small? Do you offer any sort of starting credit or free tier or anything like that to help them kind of edge into a relationship with OVH? Yeah, great question.

So we've got a couple options. So, yeah, if you want to jump right into a bare metal server, like often, you know, we have some of the, you know, the older ranges on sale. The, you know, private cloud, you know, the VMware and the Nutanix stacks, those are larger clusters and more expensive. So that's probably not appropriate for who you're talking about. But on the public cloud side, and again, we do it slightly differently than the hyperscalers and some other competitors. But yes, you can sign up right now today, go to our website, us.ovhcloud.com. You can create an account, create a public cloud project,

which is the first step. That's where all your VMs and resources are going to live. And we'll give you $200 credit to get started. So you can have $200 worth of credits to operate until they run out. So that's different than you'll find some other providers where you get only like the small micro mini resources. Now again, we have some limitations so you can't come in and just get necessarily anything until we've vetted you a little bit.

But yeah, you can certainly get $200 credit to try out our public cloud. And then if you're a small or medium business and you want to move to the cloud, like I said, we're a cloud provider through and through, and we don't want to be, and we're not going to be a managed service provider, but we do have something called professional services. We've got experts in cloud architects and system engineers who, for a fee, provide you and show you how to get set up or how to move to the cloud. And then if you're a partner, if you are a managed service provider

or a system integrator, we're always looking for partners who can help us stand between our cloud and end users. So if you're an MSP or a system integrator and you want to work with OVH cloud, please contact us. We're happy to meet and want to work with you so that, again, you can help all the end customers who might need your services on very high performance for price cloud

that they might not be able to afford with some other providers. Well, that's all very good info. The credit to let these guys get in and get started, I think is a pretty compelling feature. So we'll make sure, or option, we'll link to that in the description so people can find you. This has been a great conversation. I appreciate it, Gregor. It sounds like everything that we had sort of anecdotally thought about OVH, because I'm

going to be honest, like when we did the review of the GPU servers, your team pretty much said, hey, Storage Review, what do you guys want to do? And we said, we're not exactly sure, but give us a couple weeks to get in there and figure it out. And they said, OK, and set up, to be clear, set up a credit for us. And Jordan and Vince and Kevin and the guys got in there and got to work and played around with it. And the fact that you guys just let us come in and do our thing without being prescriptive, without asking, honestly, that many questions was kind of brave for a first engagement on a review. So, you know, appreciate the transparency and more of it here in this podcast and your willingness to engage with a couple of these questions on Discord. So, you know, like I said, we appreciate it.

It's been a great relationship so far. We can't wait to see, you know, where you go with this and what's next. And, you know, like, you know, we'll link to the resources you described on the conversation today for those that want to learn more. But yeah, thanks again for doing this.

We greatly appreciate it. Oh yeah, thanks so much for having me. Great conversation. And yeah, really enjoyed talking to you. And hopefully all of the listeners out there got something out of it. Absolutely. All right. Thanks again, Gregor. Appreciate it. Yeah. Have a great rest of your day. Take care. You too.