You heard a lot of things today about product and product innovation, about the network and what is in flight. I will try to get all this together and reflect it in the numbers that we've been presenting to you in the past and how we think it's going to impact us moving forward. There is no exciting CFO presentation that wouldn't start with the safe harbor slide. I'm not going to read this to you. I just assure that you took notice before we get started. Let's move to the highlights. I think if there's anything to take away from today's event is what Matthew calls the heart of Cloudflare and the brain of Cloudflare and the center of Cloudflare is innovation.
This is what drives the company and enables all the opportunities that we have in front of us. If you look at what this has allowed us to do over the last four years alone since we've become a public company, it allowed us to literally quadruple the TAM, the market that we address and that we disrupt. I'll show you more insight of how this has built and what products are contributing in the later part of the presentation. This is really the driving force behind us. It's what enables so many independent growth vectors that are driving the top line, and this is, you know, what got us to 54% revenue growth last year. This is the sixth year and counting year of delivering north of 50% of growth.
This is what allows us to be on a customer journey that gets us to larger and more prominent customers over time, and we'll share some new insights into what our customer footprint has done and where it seems to be going. More importantly, especially for CFOs, that we're able to deliver this revenue at a gross margin of 78%. If you need one measure to see the efficiency of the network architecture that Matthew talked about, you know, standard off-the-shelf hardware, a unified, completely integrated software stack on top of that allows us to run every product and every service we have on every server. That is the number that measures that.
We'll dive a little bit into, you know, why that is the way it is and why we are able to use our gross margin literally, as we said before, as a strategic weapon. It's also the dial we have that impacts productivity. The start of 78% gross margin is literally what allowed us to break even a year ahead of our plan. This is the dial we have to control operating margin moving forward. With this being said, let's talk about the TAM. You know, we, when we started our journey to become a public company in 2019, you know, we sized the TAM around $30 billion, and it was literally products or core offerings around Cloudflare application services.
We coined a term at the time that says, if you are new to us and you want to understand what we do, think about us as a Cisco-as-a-Service. Every time you go to a Cisco or Cisco-like company, you buy a hardware box that you put in your data center. This is what we are turning into a service that is delivered off our network. I met this Cisco CFO a year ago and said, "If you continue to use that, I will charge you license fees for saying you are Cisco-as-a-Service." You know, it must have resonated with them too. In 2020, we announced a slate of products that you get a good insight into today where they are going.
We started with what we call Cloudflare Zero Trust services, the VPN replacement products, the gateway, the Data Loss Prevention product, the Browser Isolation. This is where email is going to go in. We announced a slate of products that is protecting network services or is offering network services. This allowed us over the last two years after the IPO to add another $50 billion of TAM closely, getting us to $80 billion of TAM that we are disrupting. If we want to stay in the world of Cisco-as-a-Service, you know, in these two short years, we added a Zscaler business model, and we added a little AT&T on top of what we started out to do. We are just in what we call Platform Week. You heard it today.
The R2 object storage announcement takes us to another level in terms of TAM that we address and TAM that we disrupt. We think this takes us now for, well over $100 billion of TAM that we address, a TAM that we think is going to grow to up to $135 billion, over the next two years. It still doesn't include a lot of opportunities that we think is still in front of us. You know, we talked about D1 and D2 today, that this takes us in the database business model. We heard a lot about the progress we've been making on Workers and edge computing. We are still not sizing that in our TAM. We said we want to be conservative. Numbers, they get awfully big awfully fast, and we don't want to lose credibility.
We are getting close to that point where we can start to think about this also from a TAM disruption perspective. You get a lot of insight today in terms of developer adoption, use cases, and all the features around Workers that are still in flight. This is what drives revenue growth? This is the core engine behind our revenue growth. 51%, as I said before, over the last six years and counting. 54% in the last year. It is driving an engine that has allowed us to be rather durable and predictable from a revenue generation perspective. Matthew coined that saying, you know, this is not about rip and replace. We see more and more of this happening with new customers.
This is what we call winning the war of attrition, where customers put us in front over an installed on-premise infrastructure, and then upgrade and expansion budget dollars move to us over that time. With the explosion of the product portfolio with this, you know, relentless innovation engine, the levers we have to increase the vectors, the independent vectors of generating revenue have been rather increasing. I will, in the next couple of slides, give you a little bit more insight into the main drivers. There's for sure the large customer momentum that has been building.
I also want to show you more insight in what has been one of our really quite impressive improvements, how we have been building the sophistication around our expansion engine and how DNR has been dollar net retention has been a driving force behind continued revenue growth. We'll open some new KPIs and give you insight into product adoption for some of the new products that we have been talking about today. Let's get to large customers first. You know, this is a journey, and I remember, you know, some of you sitting in our teach-in sessions in front of the IPO, where the question was, where is your product portfolio? You're not known to be an enterprise company. You are serving small and medium-sized businesses, and that is true.
That is where it started. I think still a very important part of our business because this is where the efficiency from a customer acquisition perspective is coming through. Our journey on the enterprise sales side, our investment into building an enterprise sales organization is really showing off in terms of ROI. Our large customer cohort, those are customers that pay us more than $100,000 a year, has shown in terms of revenue a CAGR of close to 80%. The revenue over the last five years and the revenue contribution from those large customers has now risen from 32% of revenue in 2018 to 58% in the first quarter.
What's really important. Well, first of all, I think this is a sign not only that we are shipping out many products and features, but the products are enterprise grade ready. I think that's a very important note to make. We are not buying this revenue. Our. Despite this journey, despite the fact that our large customer cohorts are growing in terms of size, numbers, and revenue contribution, our customer acquisition costs have actually come down since the IPO to now about $1.21. Significant efficiency on the customer acquisition side, despite the fact that we have been moving up the enterprise stack quite successfully. I wanted to also update you on what the cohorts are doing. You know, the $100,000 cohort measured in, you know, $100,000, $500,000, and $1 million increments.
The larger the customer cohort in terms of size, in terms of revenue contribution, the faster the growth rate. This has continued to be true. You know, our $1 million cohort has been growing north of 80% for the last three years, 72% in the last quarter. We are now at 62 customers beyond $1 million of revenue contribution. We have now a handful of customers that are getting close to $20 million of ARR. You will see despite that success that there is zero revenue concentration from an exposure perspective, but we'll cover that in a minute. Finding larger customers and new revenue with new customers is one measure of and one driving force behind our revenue growth. The second big contributor is the improvements we have made on the DNR.
You know, the comment I had in the teach-in from you guys slightly ahead of our IPO was, you know, 115%, that is the number that we showed at that time. That doesn't sound good. There's nothing that motivates us more than, you know, people showing us room for improvement. Michelle continues to say, "Once we get started and focused, there's no hold back." The improvement of 12 percentage points to 127% now in the last quarter is really a result of, first of all, throwing a lot of data science at the problem in terms of where do customers come from? What is the vertical they come from? What is the first product they buy?
You know, what data do we have that allows driving expansion into new products and attach? We'll show in a minute, you know, what that is driving in terms of result. It allowed us to have DNR now to become a significant tailwind in DNR contribution. You have to keep in mind what we said two years ago is still true. For us, DNR is a very true, honest, and conservative measure because it truly covers 100% of our revenue, all the cohorts, and with that, also our pay-as-you-go business, where companies give us $20 a month to $200 a month. It's an all-you-can-eat plan, and by that nature, really hard to expand. We've seen significant improvement or really good performance from a customer retention perspective.
When you break down, go one step below on DNR, when you're one click down, this is literally where go-to-market and innovation comes together. You know, we talked two years ago about, you know, how many of our customers have four or more products. It's a KPI we dropped because it's almost maxed out. Here you see our journey of being able to attach new products to customers. The number of customers that is now beyond five products, usage of five products, is north of 80%. The speed at which the attach of plus products happens, the n+1 step is getting faster and is accelerating. The good news is we have plenty of opportunities though to get to 100% or close to 100% across the products.
Every product we add, every feature we add, you know, extends that opportunity to the right-hand side. It's one of those really important growth vectors, independent growth vectors, if you're on a call. In our journey to continue to grow and go after an opportunity, a market opportunity that gets larger and bigger and more substantial with every product we ship. You heard a lot today about new products. We wanted to share a new stat with you and update that. What is the attach rate? Are we not only shipping products, are we successful attaching new products? Application services, you saw that on an earlier slide, is what we was where we started. 75% of our customers, or 75%+ of our customers are using our core services.
The Zero Trust product portfolio is now adopted by more than 10% of our customers. The network service offering that we talked about, both together, our Cloudflare One, is north of 10% of contracted customers. You've seen a lot of customer testimonies today around Workers adoption. It's now 15% of contracted customers have Workers somehow attached to their contract. That is huge progress. It's also a huge opportunity in order to drive those numbers up. You know, another one of these growth vectors that will help us be durable from a revenue growth perspective and hopefully predictable too.
Despite all the growth we have seen, despite all the success we have seen on the large customer cohort side, and not only with the number of customers we have, but also the revenue on average they contribute, we've seen, we stay true to our attempt to stay as broadly diversified as possible. Regardless what vector you pick, you don't see any concentration. Neither around customer type. While 54% of our revenue comes from large customers, I talked about mid-market and SMB, that is really important to us, not because they are important customers only and contribute to revenue, but this is truly where the efficiency of the products come from. In that the customers are able to get behind our network in five minutes if it's your website.
It doesn't take a lot longer if you're one of the largest banks in the world. That is where this efficiency comes from and this is why this is, in addition to all the goodness from revenue contribution perspective, why this is important to us. We are making progress on our geographic diversification. Nothing, no exposure and concentration there either. I think what is really important to point out that despite all the success in the large customer cohorts, growing revenue, growing that cohort in numbers, we are not exposed to any one of these customers. The top 100 customers still are below 30% of our overall revenue.
We moved from if innovation is the engine and really the heart of what drives this gigantic opportunity that is in front of us, then gross margin is the measure of the efficiency of the architecture. I cannot get tired of repeating what Matthew said. If you want to understand one of the most important competitive moats we have as a company, then it's the off-the-shelf hardware stack and the completely integrated homogeneous software stack that runs on it. That we are able to offer every product we have and every service we have on every server increases the degrees of freedom we have to manage cost and capacity and supply and demand across the whole network surface. Every server we add, every location we add, increases that flexibility and our degrees of freedom to manage that.
This is the main reason why we are now, you know, five, six quarters in a row above our target margin, despite the fact that we are supporting more customers than ever, that we are shipping and offering more products and services than ever, that our traffic on the network has outgrown revenue growth by far, and we've been able to digest this within the model we have. There was one question on the earnings call that said, "When are you going to make use of that gross margin?" You know, "When are we going to see that? You talk about it in as a competitive threat." The fact is we've been using it all along. You know, we unmetered DDoS. We offered Zero Trust for free at the beginning of the COVID epidemic.
We started the critical infrastructure project at the beginning of this year for free. R2, that move to open data will be highly disruptive. You heard that. We've been overcompensating for this in terms of productivity and network efficiency. I think that is, you know, the true sign that speaks and is the testament for this incredible efficiency in the network architecture. We'll come to guidance in a minute. You have to keep in mind, we're able to build this network because of the hardware being off-the-shelf. The software stack almost runs agnostic to the hardware that is below it. Doesn't matter where the CPU is coming from, doesn't matter whether it's a x86 or whether it's ARM, allows us to drive that.
CapEx in percentage of revenue has actually come down from the mid-teens to 9% in the first quarter. We expect it to be in the range of 12%-14% over the course of this year. It already reflects the investment that we think we still have to make when it comes to the R2 offerings. This is the measure pricing for the efficiency of the network, but it's also the dial we have that allows us, you know, to be hopefully predictable when we come to our promise to stay above break even. That's the 78% gross margin is a strong asset to work off. We have not been holding back on achieving operating leverage either on our expense side.
R&D, despite all the relentless innovation we are seeing, is already in the zone of where we want to be. It came down to about 19%. A lot of measures in there in terms of where we put people. Ramping up Lisbon as our European development center. We've seen significant progress and scale on our G&A side. The investment really goes into supporting the building out the go-to-market infrastructure. We've been saying that as long as we see superior returns of this investment, as long as we see high growth potential and the market opportunity staying where it is, we will continue to invest. This doesn't take away that we have dials to work on that.
You'll see when we come to our long-term operating model, that doesn't take away from our ability to deliver substantial operating profits in the future and moving forward. Since this is the finance section, I also want to reiterate the guidance we delivered and maybe give a little bit more color on some of the additional parameters. We guided for 49% revenue growth for the current quarter. Would take us between $226.5 million and $227.5 million for the quarter. The guidance both for the quarter and for the year includes EPS, revenue growth of 43%-46% for the year.
I want to reiterate the network CapEx guidance, 12%-14%, not because it's an important number, only because I think it shows that the efficiency of the network and the gross margin is not coming at the expense of a high liquidity need or high CapEx. As I said, CapEx as a ratio of revenue actually has come down despite the growth we have seen over the last two years. We anticipate to be free cash flow positive in the second half. That's a really important statement. We made that at the end of last year, not knowing where the macroeconomic environment would take us. In today's environment, that is an even more important statement, and we'll be relentless in achieving that commitment. Let's reiterate the long-term model.
We've made significant progress in achieving operating leverage. We also said that we want to be around slightly above breakeven as long as we see there is return. The levers that allow us to dial that in are really, really strong, I want to repeat that. Based on the gross margin that is now in the 79% range, slightly above, where we want to be, not because of a lack of opportunity, but because the infrastructure team that runs our network and builds our network has done a really good job finding productivity and levers to increase productivity, and compensate, even overcompensate for the places where we use gross margin as a strategic leverage.
I think, you know, where product innovation and this relentless innovation comes together is it drives our marginal cost to put products on the network and serve our customers actually down. You know, every new product we ship literally comes at almost zero additional margin dollars from a cost perspective and allows us to aim for a significant long-term operating margin opportunity that we think is well above 20% in the long term. With all that being said, you know, summarizing a business model that I think is unique in many ways. It's unique in the markets we disrupt. It's unique in this, with the speed and innovative capacity we have.
I think if anything has become clear today and at the start of the week with platform week, then it's the innovation engine. There will be many more innovation weeks to come over the course of this year. We have been able to use that and turn that into a record of delivering strong and consistent revenue growth. I think durability, predictability and consistency is really key here. We have seen that we can manage this revenue growth while we continue to significantly invest in our enterprise go-to market and continue to see returns, not only in terms of superior growth, but also in terms of driving our customer acquisition costs down. It's a highly efficient business model. Now I shared that with you before I joined Cloudflare.
That was what really fascinated me coming from one of those more established software companies. What is the efficiency of that end of that model that was visible. Boy, did I fall short in terms of seeing the opportunity that model presents. I think you get a glimpse now how the flywheels that drive that efficiency are thriving and accelerating growth. Highly impressive. With that, I think we have multiple levers to achieve and get to what is really important in this environment. Not only free cash flow upside, but being positive at the end of the year and staying positive moving forward. With that, I would close it here. I would invite Michelle and Matthew to come on the stage here with me and answer any questions you might have. Thank you.
Good job.
Thank you.
Good job.
My favorite investors bring me cool mugs. The team at Whale Rock brought me, "Matthew, you are too awesome." Which is definitely there in the lead. Next investor day.
Good. That's a high bar.
That is a pretty high bar. D1 you can do something fun with, maybe.
Yeah.
Yeah. Anyway. Jason, are you moderating?
Yeah.
Oh. All right. How about we'll start. I will moderate. Yeah.
If you were to explain to a five-year-old the whole increased freedom thing and how that works, like, can you do that? Can you like explain Cloudflare and how it works? Like, give me an example. Like even runs like DDoS attack to Morgan Stanley out of Sri Lanka, like this is what it is like. How does that not add cost to you? How does that not add latency if it takes up all the way the other side of the planet?
Yeah. The way that I think about it is that every packet we process has a different value to us. Morgan Stanley pays us more than Bob's blog and has a different latency sensitivity that's associated with it. A way of understanding this is we have this giant network. We have a whole bunch of free customers. Don't pay us anything, right? One of the assumptions is that to service those free customers, it costs us a lot. Actually, it doesn't cost us very much because there's always excess capacity somewhere on the network. Oftentimes, it's really close. Oftentimes, we can service that free customer here in New York from a data center that's probably, you know, within a few milliseconds of where we are.
If Morgan Stanley is having an attack or if ProPublica publishes an article that goes viral and we need to make capacity, then it's gonna be those free customers that get shoveled off. For the most part, they're still gonna have a better experience than they would have without Cloudflare, because again, that capacity is everywhere. The real key is figuring out the ability to, first of all, be able to move traffic flexibly around the network to wherever you need it. Having the, you know, very incredibly simple architecture at some level, which is every single service can do whatever we need anywhere in the world.
Being able to say, "This customer is paying us a lot," or has an application which is particularly latency sensitive, so serve that at sort of a high level of service, whereas this other customer isn't, and we're gonna. They're still gonna get great service, but it might be there. I think, you know, in explaining it to a five-year-old, it's like, listen, there are certain things that, you know. I don't know. I just. My wife and I just have. We have a three-month old. I'm trying to figure. I haven't quite figured out the five-year-old thing. For three-month-olds, like sometimes she really needs to eat, and she needs to eat right now, otherwise she turns into a purple screaming pterodactyl.
Like, that's a high priority, and we're gonna make sure the latency on that is very, very low. Whereas, you know, does she get a bath tonight? Like, yeah, it's probably good to give her a bath on a periodic basis, but she probably doesn't need one every single night. We might deprioritize that. But we have architected our lives to be able to allow me to come out here, and the house hasn't burned down yet, it seems. We've architected Cloudflare's network in a similar way to be able to set that priority. But it starts with that radically simple architecture and just. Like, that's hard to actually get, and it's hard to write the software. But now that we have it, that's how we're able to innovate as quickly as we are.
That's how we're able to launch products as quickly as we are. That's how we're able to scale and get the efficiencies that Thomas was able to show. Does that make sense?
What I would add is for this example, there's another example that Matthew gives from time to time where you don't shift customers around. There's always a part of the world that is asleep.
Where capacity is idle. You know, that idleness moves around the globe over 24 hours. There's always capacity that we can use to offload compute tasks that are less latency sensitive because we have over capacity or idle capacity, not over capacity. Idle capacity in parts where the world is at night, for example.
One other thing I would just add, just because we've been building this for 11 years and we've done a lot of conversations, this comes up a lot. It's when we started and we told people how we were gonna build Cloudflare and our approach, a lot of people told us that was impossible.
This is not how others before us solved the problem, and we really kinda innovated and found a path forward. I think that that's why it's kinda hard to explain. No one else is really doing it, and a lot of people told us that this is not gonna work, and there's lots of reasons, technical reasons why, but those were false assumptions. We made it work.
Thank you guys for I wanna ask you about Workers. I think this is a solution that you've been very good at showing us the potential of sort of where this could be, but somewhat measured in terms of what the impact gonna be to Cloudflare in the near term. Last week on the earnings call, you were talking about getting to that critical mass for Workers. You have a lot of solution out there today. You have a lot of applications that are deployed. Can you help us more granularly understand what does critical mass mean? Like where do you have to get to to get to that critical mass for Workers, number one?
Number two, what can you do to help sort of jump the shark a little bit and sort of accelerate the that time to getting to that critical mass?
Yeah. A couple of different ways to measure that. One, I think that if you look at developer platforms, 1 million developers using the platform tends to be a milestone. We believe we are on path to get to that milestone this year. It takes time for people to adopt things, but if you just you know, if you're all sitting here with us, the most interesting presentation is going and sitting with the developers next door 'cause the enthusiasm and excitement that they have for it. I mean, if you just search Cloudflare on Twitter right now, and you say, you just watch people go, "Wow, I am considering moving my application from AWS to Cloud flare." That's because we've got the tooling that allows them to do that.
I think there were some components that we were missing. One of them was an object store. That's gone open beta this week. It will go GA this fall, probably in September. Second is a database. You need a database to be able to build sophisticated applications. We launched that this week. We expect it'll go open beta actually much more quickly as a result, and we'll build that. We actually are gonna need multiple different styles of databases, so we call this one D1. There will be a D2. Obviously, if we have an R2, we have to have a D2. So we're heading down that path. There will probably be a D3 and a D4 as well 'cause you need different styles of databases, and we won't necessarily build them all.
We also might interconnect, and we already do that with other vendors that are out there. I think those are the three components that you really need to do that. The challenge with developer platforms, and we've studied lots of them, is that it always takes between eight and 12 years for successful developer platforms to really catalyze. The reason for that is you need a community to develop. You need sort of the first killer app to be out there. You need the people who built the first killer app, some of them to leave that company and then go migrate to other companies where it starts to be like, "Oh, if I invest in this platform, I'm going to be successful in my own career." We took the time to look at how did Microsoft build their platform?
How did Salesforce build their platform? How did Apple build their platform? And were there any shortcuts somewhere along the way? Basically, I wouldn't describe it as if you said jumping the shark, but it's that 'cause that means you're past your prime. I think it's like how do you leapfrog to the next opportunity? I think the only thing that we see that's a shortcut is how do you build your platform into existing successful platforms. Earlier this week on Tuesday, we announced what is sort of Workers as a service, where you as a platform can build Workers directly into your platform. The partnership that I'm the most excited about with that right now, although I think you're gonna see more in this space, is with Shopify.
If you think about all the developers that are building storefronts on Shopify today, tomorrow, right? If they are building storefronts on Shopify, they are learning the Workers platform because Workers has gotten built directly into that. You can imagine all of the other platforms that are out there, where if we can say, "This is a way to make your platform more programmable. It's a way to make it more scalable. It's a way to integrate it more." I think that's how you go from instead of taking twelve years, taking it eight. But again, I can't believe that Walmart is building on Workers. I can't believe that Atlassian's building on Workers. I can't believe.
In NCR, if you're swiping your credit card at the grocery store, that's being powered by Workers behind the scenes at some level in their anti-fraud systems. It is catalyzing more. I think that we're measuring number of developers as the KPI. I tend to think of Cloudflare's business as having acts, sort of our app act one where the application services business is. You know, again, we've got tons of penetration there. That's what we're known for. Act two is really the Zero Trust services. That'll be the big driver of revenue for us over the next period of time. That is a significantly larger market than the application services market. Act three is Workers. That's a significantly larger market than Zero Trust. Act four is we're already working on as well.
I think that's how I would think of it. Developers is what we're watching, and then we think revenue will follow.
Fatima Boolani from Citi. Thank you so much for hosting us. Michelle, this question is for you. When I think about the slide that Thomas put up, with respect to your $100,000 paying customer cohort, they derive, I think, roughly 50%-58% of your revenue. Outside of 10x-ing your sales and marketing spend, how do you get that sort of 1% of your base to grow much faster? Maybe if you can put it in the context of some of the executive hiring you've been doing to really build out a more traditionally defined enterprise go-to-market motion.
Great. Yeah, thanks so much for the question. You know, it's a little bit like these different acts where you started. We started by servicing everyone, and then we kept getting pulled up market, and we continue to get pulled up market. Yesterday we had our customer advisory council where it's all these large customers, all the logos you've heard of. It's that we have over 1,500 of these large customers that are using our service. It's not gonna stop there. There's so many more ahead of us, and so it's been this journey which we're really proud of our go-to-market.
One of the things we said how we think about it internally is instead of investing in marketing to go get large customers that then go drive revenue, how do we use our large awareness, our large install base to drive demand and then go invest marketing behind the curve? We kinda look at sales and marketing to help fuel our business in that way, which is part of the reason why you've seen our cost to acquire a customer come down because our audience has gotten bigger install base. Let me just give you like a really tangible example, 'cause we were just meeting with customers yesterday and they're here, so you should talk to them about it. One of them was just saying, "Oh my God, like I love when you blog.
All my engineers circulate these blog posts internally at this. This was Epic Games. They're obviously a really technical company, but they need all of our services. He is like, "And then we find new ways to engage with what you're doing at Cloudflare through your blog." Right? It's just these really. You think the blog is for developers around the world, which is true, but they're employed by a lot of big companies, and then they pull us in to find problems that they have internally to help us to do that. Those are examples of how having a large audience, a large awareness, sets us apart from maybe other discretionary B2B companies that don't have that, right? Where they're really doing just a very straightforward demand generation activity.
We do that too, but we have these other assets on top of all of it and we're really proud. That's a little bit of like how we think about it. You know, in terms of scaling up that go-to-market motion, it's going well. We hire account executives like you would expect. We have pre-sales engineers and post-sales engineers and customer success team servicing these customers. We know the practitioners, we know the economic buyers, we know the C-level executives, and those are all different journeys. We're doing that and it's going well globally. I can say it's going well globally.
Can you talk a little bit about?
You guys asked a question already. We'll come back, I promise.
That's fair.
Just to be fair.
Hey, Gray Powell with BTIG. Thanks for hosting the event. So this one's probably best for Thomas, but I just thought it was really good to see all the different attach rates in some of the product categories. I just wanna make sure I have it correct. Was that across all customers or enterprise customers? And then on the ZTNA side, is that mainly Cloudflare Access is getting the most uplift? And sorry for the multi-part question here, but just how much does ZTNA add to the typical customer's bill?
The attach rates were beyond large customers. They were all contracted customers. If you're also below $100,000, I think, which is the even more impressive stat, that it's 10% across all contracted customers, in the study you mentioned, not only large customers. The last question you have to repeat for me.
Yeah, sure. On the ZTNA side, how much of an uplift are you getting there to the typical customer's bill when they adopt it?
I think.
It'll depend. It's very much seats-based.
Yeah.
We have some customers that will have very few employees, but will push a lot of you know our application services. There are some that have, you know, spend very little with application services, but would spend quite a bit with us. I think we gave some examples during the last earnings call where, you know, we're seeing the ZTNA, the greater ZTNA story, and that's both the Access and Gateway products.
Yeah.
Coming together. You know, I think that is typically for a contracted customer, hundreds of thousands, you know, $100,000+ additions to what the spend is. But we like that it can start with an individual team. So if you look at someone like L'Oréal, they had an individual team they needed access. They started with that individual team. That account has then significantly grown over time. It might start with a, you know, $5,000 or $6,000 self-service contract.
Yeah.
self-service, you know, someone paying with a credit card that then quickly turns into something where they see that it works, expands across the entire organization. That could turn into a $100,000 contract. Then as they expand from Access to Gateway, you know, that might turn into a half a million dollar or more contract. I think that we're seeing very good attach rates and that this can meaningfully drive additional revenue from those existing customers as we can take that 10% adoption, you know, up over time, which we fully expect we'll be able to do.
The only other thing I'd add, just it might be obvious, but in case it's not. We have all these four axis maps you talked about. You know, we have this wheel, it's the four.
There's four parts to a wheel. Customers start in any one of those four parts of the wheel. Some people start with zero trust. Like the Twitter example Matthew gave today, like they started using Cloudflare because of a zero trust need, for example. Other customers start using developer services. Other customers start with an application performance and security, which is what we've been doing the longest. Customers start in these different areas. They expand within that area, but then it's also they can expand within other areas. It's all these expansion levers. I know that's obvious, but it's not really that obvious. That's just something that I wanted to point out where it's they can start in both places.
We're constantly saying, "Hey, how do we get them to expand within the use cases that they're already using, and then find more use cases within the organization too?
What that also leads to, it doesn't make our job easier sometimes, is there's really not a typical path for a customer to come to us. You cannot say there might be a story around a correlation around the vertical a customer comes from. The financial customer probably starts with security. An e-tail platform probably starts with a performance-based product. There is not a typical starting point, and there is not really. We are getting better at predicting what the second product is going to be than predicting what the first product is so. We are improving rapidly.
Hi. Phil Winslow at Credit Suisse. I just wanna focus on zero trust again. Obviously, you talked about in your slides, Thomas, you know, 10% of contracted customers use the zero trust service. When you talk to these customers, particularly the larger ones, what is it about Cloudflare that's making Cloudflare stand out versus some of these, you know, long existing vendors like Zscaler you mentioned or Netskope, another startup? Sort of what is it that you have that differentiates you?
I would say there's a handful of things that we hear over and over again. The first is our product is truly integrated across any situation you want. You are worried about an attacker which is coming at you using multiple different vectors. They're gonna DDoS your site. They're gonna try and attack your web application and try and bribe employees. They're gonna send phishing emails. This, we see this. These are really sophisticated attackers that are targeting across it. We can give you a picture across that entire platform. Zscaler can't. Zscaler actually pays Cloudflare to be their DDoS mitigation service because they just aren't good. They don't have the capacity to. They have two orders of magnitude less capacity across their network than we do to be able to even deal with that.
That unified platform I think is a killer feature. Second thing that I think that we're differentiated with is ease of use. We just have ease of use in our DNA. We want things to be self-service. We want it to be easy for you to sign up. We have to be able to service the millions of customers that we have in an incredibly easy way across that entire platform and make it work. The third is performance. If any of you are using Zscaler, it sucks in terms of its performance. It slows you down. It is slightly better than your old corporate VPN, maybe slightly better than Citrix, but not much. Users hate it. IT tickets are higher. It slows customers down. It breaks things. People don't like it.
We have to make our products work at literally a consumer level. Like, again, in Russia, the app that is our VPN is being tested in what is the most hostile network environment anywhere on the planet right now. It works even in those circumstances. That means that when somebody deploys Cloudflare's product, it makes it much, much more available. Then I think the fourth thing is just we can deliver a much higher ROI. Because we have this broad diversity of products, we can often bundle things together in such a way that we can say, "Listen, you want... You know, you're willing to pay X for Zscaler.
We'll give you exactly the same functionality that they have for exactly the same price, but we'll throw in DDoS mitigation or WAF or something else for free. I think great companies that are in the SaaS space that go from the tens of billions of dollars of market cap to the hundreds of billions of dollars of market cap do it all in one way, and it's bundling products together. Every single platform that does it figures that out. Most people have to do it through a series of acquisitions. I think what's unique about Cloudflare is we've been able to organically build that platform that has the sufficient breadth that allows us to do that bundling strategy. That's a really tough thing if you're Zscaler to compete with. Like if we just say, "Listen, we'll give Zscaler away for free if you buy DDoS," right?
That's a move that we can make in the same way that you know that Microsoft uses their E5 license in order to you know take on some of the folks that are there. I think that's gonna be a huge part of our story. We're already seeing it. Let me tell you the flip side of your question. We'll tell you what's good. I'll tell you what we're bad at. We have not yet really mastered what Zscaler is amazing at, which is partnerships. Like, that's I think a place which is a big opportunity for us to continue to invest. We love the fact that Accenture selected us.
Mm-hmm.
In order to be who they would go to market for their big federal contracts. I think being able to work with those sorts of systems integrators prove that we can do that, and also going out to even the small VARs that are out there and saying, "We're gonna take care of you. We know how to run a program. We know how to run a partnerships program." That is a huge focus for us right now, and it's actually one of the sort of secret benefits really of us buying Area 1 Security. They had almost all of their sales were channel-based sales. They really knew how to do that well. We're absorbing that DNA, and I think we're gonna. We know that that's the case.
That is different than what our traditional products are because these Zero Trust products do often take a little bit more handholding, and so there's much more of a role for a systems integrator or a VAR, and I think that that's an opportunity for us to improve.
Thank you. Shaul Eyal with Cowen. Thank you for hosting. Great to see everybody. Probably one of the key takeaways or at least a new data point, Thomas, that's I think back to Gray's comment, that's on the attach rates. You know that the question's already coming, and we start already getting, I believe you mean, like, you know, it's through the emails. So if we're to sit here in like two, three years, and the buckets that you've showed us and provided us with, which one do you think probably is growing the most? I think that's on the one hand. Maybe, Matthew, you know, just great that you've mentioned that partnership.
Matthew or Michelle, on VARs and, you know, on Accenture, should we be expecting more announcement along these lines? I know Accenture, it's mostly on the federal side, but what's happening on the enterprise front in that respect? You know, typically it'll go the other way around, first enterprise then federal, but with you actually it's a little different. Just curious about your thinking about it.
I'll take the first bit and Michelle can take it. The way we think about our business when we think about products is we're stacking a series of S-curves one behind another, and different products are at different development points on those S-curves. A product like DNS, like we are the biggest DNS provider in the world. DNS, super mature. We're at the very end of that S-curve. There's not gonna be a ton of growth from DNS, but it's strategically important. We'll continue to do it. We invest in it. We make it better. We don't want anyone to be able to sneak in behind us, so that's that. On the other hand, something like Cloudflare for Offices, which is a true MPLS replacement and taking on that, I think that's an act four product.
Like, that's at the very early stages of how that's gonna develop. Exciting. We've got. You know, we're learning a lot. We're trying to figure out how that's going. We wanna have products that are all of those different stages. Depending on what your timeframe is, I think how much the buckets grow really, really depends. If I think your timeframe that you laid out was sort of in the next two-three years, Zero Trust is gonna be a big part of our story over that time, and that's the place where I think you're gonna see. That's right hitting that hockey stick growth. I think it's gonna be great for us. I think it's gonna be great for Zscaler. I think it's gonna be great for Palo Alto Networks.
I think you're gonna be actually bullish on all three of us, in that space. When we go head-to-head with them, I think we love our win rates, and I think that we can make that business bigger than what all of our traditional platform services businesses are. If you had said, "What do you think over the next 10 years?" Now you're in Workers land, right? The opportunity that's there I think will dwarf everything that we've done with platform services and Zero Trust. If you said, "What do you think over the next 20 years?" Now we're doing Cloudflare for Offices. If you look at IT budgets, what's the biggest thing that every Fortune 500 company spends IT budgets on? Telecommunications and connectivity.
We've built a network where we're not that far from every physical office in the world. If we can build an intelligent network that says, "Well, we'll take over some of that telecommunications and connectivity spend," I think that's a big opportunity over the long term, but it's a much longer term. I think that it's gonna shift over time. We're gonna make our way through that S-curve. What I have faith is that our product team will continue to launch new things, find new areas, and find something else where you're kind of at that slow initial part of the S-curve that will continue to stack over it. That's why our revenue, I think, is so durable over the long term.
On the channel front, we do more business with the channel today than a year ago and more than two years ago. It's actually a fast-growing distribution channel for us. It's going well, but it's still early. Where we see a big opportunity is with Zero Trust, as for all the reasons Matthew said, because with Zero Trust, you gotta manage the endpoints. You gotta manage like the fleet. A lot of organizations are looking for like a managed service provider to do that for them, and we're not in that business. We don't wanna be in that business.
That's opportunities there where managed service providers, the SIs, the VARs, depending on which piece, where it's like, "Hey, we wanna be able to help design the Zero Trust solution, deploy it, and then manage the policies and the devices over time." That's why the channel is so important in that segment. We're doing a lot there, and you'll see us do more and with more announcements and more. It's an important piece for Zero Trust, and we're actively investing in growing there. The channel for us actually is a very fast-growing channel. It's growing faster than our overall business, so it's becoming more important to us.
One thing to add, and this maybe is a Thomas point. There is so much margin for us in these Zero Trust products.
Mm-hmm.
that it is an easy thing for us to be a good partner in.
Hello? Hi. It might be a segue here, Matthew, to something that Thomas said here, KeyBanc, Tom Blakey. About share gains, I think that's super exciting. You mentioned there's a large opportunity that you're sitting in front of, Tom. I'd love to if you can give some color around what that opportunity is that you're sitting in front of about large customers that haven't yet adopted the full Cloudflare platform. I know you can't give us dollars, but some color there. As a follow-up, logical follow-up to that, what's holding these customers back? I'd love to hear about those, what's in the sausage-making factory, Matthew. You gave great examples on the call about ripping out legacy and
Using Cloudflare at these large customers, you know, what's involved in rewriting policies and what's holding these customers back and before we hit the inflection point there? Thank you.
Maybe I get started on the first part of your question. You know, I mentioned that we have now getting to a handful of customers that are far north of $10 or $15 million, getting close to $20 million of ARR. Not even the largest one is using all of Cloudflare products. I think that gives you an indication of how much expansion room there is. 127% is a DNR measure across, as I said, all of our cohorts, including our pay-as-you-go business that is difficult to expand. There is a lot of room you have to assume in the much higher performance already on the enterprise side. The bigger the customer is, the faster we expand and the higher the DNR is.
You know, I think these really are new products. Like, we really had a competitive Zero Trust suite for under two years. There's time that takes to educate the market. Like, the more amazing thing is how many people are actually adopting them as quickly as they are. I think that I like that a lot. What I think is also impressive is the buyer isn't always the exact same. Our platform services buyer and Zero Trust buyer, they know each other. They probably sit on the same floor. They probably report to the same person, but they don't necessarily have the same budget. You have to navigate that internally, and I think it gets back to the question that Michelle was answering.
Like, that means that we continue to invest in hiring what look like a traditional field sales, relationship-based sale. I think that it is a mistake 'cause it's inefficient to just say that the way that you're gonna grow is just you're gonna hire as many salespeople as possible. I think it's also a mistake to sort of say, you know, the old Twilio model, we don't need salespeople at all, right? The right answer is you want a more efficient sales process where you're investing behind demand, you're building trust over time, you're capturing more and more share of wallet, you're intelligently using bundling, and you have a great AE customer success, solutions engineering team that can support that as well.
When Michelle talks about how we want to invest behind demand, not ahead of demand, I think that's it. You fast-forward 10 years from now, our sales team looks exactly the same as everyone else. The difference is, I think, we're less likely to make a mistake because we're using the signals in the market in order to make intelligent choices. I'll give you a really, really, really specific example. As we were thinking about expanding our presence in Germany, we had a handful of early adopter German customers. One of the questions that we asked them was, "Who are the sales leaders that you have really enjoyed working with?" They gave us a name, a guy named Christian.
Stefan.
Stefan, sorry. Christian's a different place. Sorry, Stefan, if you're watching. Stefan, they give Stefan's name. We talked to Stefan. He was awesome. We hired Stefan. He immediately had a good rapport with those existing customers. He expanded those customers and was able to go get more customers. Again, that is different than hiring a head, saying, "Oh, we're gonna expand into the Japanese market. We're gonna hire a headhunter. Go find someone." Like, if you can actually get the feedback from your customers, you can listen to them, you can say, "Who do you want to be supported by?" It is much more likely to have a culture of success. The person you hire is much more likely to make their quota number.
They're much more likely to hire people who will make their quota number. Again, I think that's just a better way of building a sales team. You end up in the same place, you just make fewer mistakes along the way. At least that's been our experience so far.
You know, it's interesting. Just the only thing I would add is, when you said, "Hey, what holds people back?" Maybe it's less of that, but I would say that like a lot of your research actually is helpful because we're a newer company, and so it's awareness. Are you gonna be around? The credibility. One thing that's actually been surprising to me is going public has been good for that. Going public, being more in the market. People can look us up. They're like, "Oh, wow, okay. Other people are writing about you." It's actually been good from a credibility standpoint 'cause before, when you're private, small companies are worried that you're gonna go out of business because it just, it's a lot of startups have gone out of business. We're not that anymore.
The credibility, the awareness has been very helpful. Thank you to all for all your research notes. It's actually very helpful. People read them all the time. Our customers are always referring to them, saying, "Oh," and employees who wanna join too. It's you play an important part in the ecosystem.
It doesn't mean you get commission. Yes.
Is this mic on? Yeah. Steve Cavan, Kingdon Capital. Two questions for Thomas or the second for everybody. I hear a lot of concern that R2 is gonna drive a lot of revenue, but also drag down gross margins and drive up CapEx. I think people are looking at comparables out there and drawing some conclusions. Can you just talk about the ramp of that and what it means for both the CapEx side and or how much that's offset by, say, Zero Trust, and also then the, the kind of the margin side? The second question is the ARM server shift, and I'm guessing that comes next year. AWS is seeing 30%-40% price performance improvement off the Graviton processor, which is an ARM chip. Do you guys think you're gonna be able to get a similar step function down in cost?
How are you gonna flow that through the business? How do we see that flowing through the business, both from a cost perspective and from a functionality perspective? Kinda when does that hit?
Yeah, let me get started, and Matthew might and Michelle might jump in. You know, we are comfortable that the efficiency of the architecture is going to hold up. I think, you know, I cannot repeat myself often enough. The architecture of the network and how the software stack sits on top of it is an elasticity that we think is comfortable to scale across the new business models we enter. We've launched not only low margin products, we have launched zero margin products before, and it actually improved our gross margin in terms of the traffic patterns it drives, the additional degrees of flexibilities it gives us to shift traffic around.
We've been rather careful in modeling the impact of the new products and gotten very comfortable around the fact that the elasticity and the efficiency of the architecture is going to hold up. We will not turn into a multi-billion-dollar-a-year-to-build data center companies. That is not part of that architecture. That is not something you should be concerned about.
The fact that we have been scaling the network to the size that you saw today in more than 270 cities, more than 100 countries, digesting up to 90% traffic growth in spike quarters, especially when COVID started ramping up the number of customers, carrying the products we carry, and the margin didn't go down but went up, and the CapEx ratio went down over that period of time, I think is an indication like that. That doesn't mean we have to have all high margin products, but we can put the margin at those products to use to be disruptive in areas, in other areas. I really like the examples that I gave you. You know, some of you asked me how does pricing and discounting work at Cloudflare?
It's not like we look at, you know, what is the competitor's price and do we need 5% or 10% or 15% discounting. We do that from time to time on important deals. The fundamental question we always ask is what is the most disruptive move we can make in this market, putting that margin to work? When you go back to DDoS, it was not, "Well, let's sell our DDoS product at a 20% discount to the main competitor." The most disruptive move was to unmeter it and offer it for free, so to speak, if you are a large corporate customer of ours, unmetered DDoS is part of that bundle.
The reason for that is because how the network is architected, that the defense of that DDoS attack happens at the edge of the traffic network, that we don't have to pay high bandwidth costs to channel traffic back and forth to scrubbing centers like other companies. I think you have to look at us from that perspective when we launch products like this and when we talk about trying to be disruptive. We model this a lot. We think the efficiency is going to hold up.
I'd add starting at the you know, 10 inches level and then going up to 10,000 feet. At 10 inches, I would say that this. We've looked at what the capacity that we have in excess storage across our existing network, 'cause if you took all of the servers that are running all around the world and you put them in one data center, it'd be one of the biggest data centers in the world, right? We've got all these servers all around the world. They all have storage. We ask how much of that storage is being productively used right now. The amount that's excess would more than satisfy the demand that we've seen for R2 at this point. Which is another way of saying that to satisfy the demand that we have requires no additional CapEx spend today.
Now, we anticipate that demand will go up, and then we will add more servers, and then we'll service that. What will we have excess of? CPU, RAM, network capacity. What will that allow us to do? We can add more Workers functions. We add more security functions. If you looked internally at Cloudflare, we have our infrastructure team, which is one of our secret weapons. We have our product team. You met Jen earlier. We have our ETI team, which is a team that thinks about what are the sort of disruptive next generation products, and they are constantly asking themselves every single quarter, "Where do we have excess capacity on the network?
With that excess capacity, what are additional products that we can build? Again, I think this is a perfect example of exactly what Michelle talked about, how we invest behind demand, not ahead of demand. We decided to launch R2 because we had the capacity to do it. We didn't say, "We're gonna announce R2," and then go build the capacity, right? We did it the other way around. We can do that because we have the flexibility to be able to deliver that service over time. This is at 10,000, maybe this is at 100,000 feet. I was a student of Clayton Christensen, right? Like studied The Innovator's Dilemma. My essentially chief of staff is his former research assistant. Someone here approached me about it a little bit earlier.
I really do believe that one of the ways that companies go wrong is that they overly focus on making sure we're gonna hit a margin target with every product that we launch. That's how you get disrupted, and that is how you lose. We don't do that. Instead, what we say is, we're going to, as Thomas said, launch products that will completely disrupt the marketplace, and then we will engineer the margin back in, and that's what we're constantly doing. That's a good segue to your second question, which is ARM.
I have no idea exactly how much more efficient ARM will be or otherwise, but what I do know is the fact that we've done the hard engineering work to make all of our platform run on ARM, and we have ARM servers in the field running right now, and they work great, and they are cheaper for us to buy than that. That then helps us put pressure on Intel and AMD to drive their costs down. As they drive their costs down, that lets us put. Having the ability to have multiple vendors not being locked into any one platform, making sure that every machine is effectively just a drone in a giant army, where you can't tell one from another, that is the engine that allows us to continue to. We're really tough on our vendors, right?
We push costs down, and we can do that 'cause we have the flexibility that if one of them says, "Well, you know, we won't sell you chips," we're like, "Great, we'll go buy a bunch of ARM chips." That is, I think, the more important thing. It's less about ARM is so much more efficient. It's actually more about it gives you more leverage over the x86 guys. By the way, if, you know, Power or, you know, RISC-V or whatever it is the next thing that allows us to drive that down, we've built the efficiency in our platform where we can continue to drive more and more efficiency and continue to drive our underlying cost down because we have this radically simple architecture. It all comes back to that, so.
Hey, this is Anusha from RBC. You touched on the telco opportunity. Could you expand on that and talk about how you're helping customers reduce legacy telco spend and how big of an opportunity this could become?
I missed which, the hardware?
The telco opportunity.
Oh, the telco opportunity. I think it's way too early to say. I do think that everyone agrees technologies like MPLS are way too expensive, way too slow to provision and actually not that secure. If you can find more efficient ways to give people the ability to have a software-defined network that has either physical presence in the buildings that you have to access or uses radio to get you that same performance, that there's gonna be an opportunity. We are to something like 90%, you know, well over 90% of office workers in the world were within a 5G radio hop of where they work. We are usually within a dark fiber light up of where they work. That gives us an opportunity to use our network.
When we do that, we've already got customers who are doing that. You talk to someone like Garmin. Like Garmin, if you're sitting in a Garmin office, and you go to one of their internal applications, or you go to another cloud SaaS customer, you never touch the public internet. You go to Google, you go from Garmin's office through a piece of fiber optic cable to the edge of our network, cross our backbone to another edge of our network, and then we hand off directly to Google. You've never touched the public internet. That opportunity to be on both sides, I think, is there. There are ways that we can do that more efficiently than others, and I think the advantage. There's never been a truly global telecom. Even Global Crossing or Level 3 were not truly global telecoms.
If we can figure that out, and again, that's act four, that's. We'll be really talking about that in 10-20 years. If you can really figure that out, I think that is, you know, more than 50% of the Fortune 500 IT spend.
Got it. That's helpful.
Okay. Operator, we are going to take one last really good question from an analyst right here.
2. I have to hear.
Maybe two.
Gentleman in the red shirt who's asking his second question. We'll do two. Yeah.
Yep.
Just on R2 really quick. 9,000 customers in the private beta. You just opened up the regular beta, and you're gonna release at the end of the year or sometime in the second half, you said. What's the best milestones for us? Like, you've given us your Workers goal is to get to 1 million, you know, Workers this year. What's like are the best-
You have to take all that.
you know, milestones for us to think about when we're thinking about R2 in terms of how we should be thinking about that in revenue contribution going forward? Even though it's in act three, I know.
Yeah. I mean, I think that it's. At some level, what I care the most about are customers building really innovative, disruptive applications. If you can be supporting that and building that, then the rest of everything follows from that. Like, that's how I would look at the overall Workers space in general. I think R2 could go one of two directions. One direction is the rest of the market doesn't respond at all, and we just capture a huge amount of object store. In that case, then success is how many objects are on the platform, you know, how much are we charging for that super sticky business if you have it. And then, you know, can we maintain our margins, which we again have confidence. That is not.
Like, Zero Trust products are super high margin. R2 is in line with what the margins is. It's not as super high margin as that, but how can we, how can we do that? We think it makes sense to do that. That's one measure of success. A completely different measure of success is, Andy Jassy says, "I am sick and tired of these guys taking our objects away. We're dropping our egress fees to zero." I would be so excited because all of a sudden now we have unlocked the ability to be that CPU and the ecosystem. We can be the network that interconnects the cloud together. I think that's if we force the market to drop egress fees, that's in some strange way an even bigger victory for us.
Where R2 then probably, I think there's still some things that we can do with it that are gonna be really interesting. For instance, as data locality and data residency requirements become a bigger deal all around the world, being able to say, "Here's an object. I want you to store it in Germany, and I never want it to leave Germany, so it can move from Berlin to Frankfurt to wherever, but it has to stay physically there." We think that's a premium product that we can actually charge a. It actually can be a very high margin product for us. I think that will be a long-term opportunity for R2, even if everybody else matches on egress. It's kinda hard to say what I'm rooting for.
Like, I love the revenue of us, you know, winning the object store space, but I actually think the long-term play of us driving egress fees in the cloud to zero is the bigger strategic win for us over time. We have two more. One there.
Great. Thank you. Alex Henderson over at Needham. I wanted to ask you, since this is a development environment here that we're talking about, and there's lots of developers here, something along those lines, which is you have done a really good job with Workers of basically taking away the infrastructure element that the coder has to deal with. There's another company out there that spends a lot of time working on that issue for coders. You are one of the largest users of Consul, for instance. I guess my question is, over time, should we be thinking about the multi-cloud architecture that people are talking about becomes transparent on your network that I don't need necessarily to build all of those elements in?
Is this a direction that changes the mechanics of code as infrastructure over time?
Yeah, you know, how I've thought about this problem has changed and I think evolved. When I thought multi-cloud in the past, what I imagined was you write an application and you run the exact same application on AWS as you do on Azure, as you do on Google Cloud, and you play them off each other and drive the cost down and that would be and you have higher resiliency and those things. That's not how any serious customers work. With maybe a few exceptions, like Zoom does that a little bit, a handful of others, but it's really hard to make that work. When big enterprises talk about multi-cloud, that's not really what they mean.
Talking with customers, and we spend tons and tons and tons of time talking to customers, and they drive our roadmap, and they help us develop what it is. When they talk about multi-cloud, what they mean is one of two things. Either, one, because of M&A or just different teams or whatever, they've got a bunch of stuff across a bunch of different clouds, and they need one consistent control plane across it to make sure it's secure. It's not, you know, it's not gonna expose secrets. It's not gonna do those things. We play an incredibly good role in doing that.
It's probably in that sense, the fact that we are that neutral network that can be in front of everyone and be consistent across whoever you wanna use is a really powerful differentiator, even if one of the big public clouds replicates one particular feature or another. I think that that is a valuable thing for us to do. Having that consistent control plane, a place to enforce security policies, a place to be able to see what's going on, you need a network. It's really hard to solve it in some other way, even if you abstract the APIs and do whatever it is.
I think that's actually good for us to abstract those APIs and make it, but I think it's almost playing for what my misassumption of what the public, like, what multi-cloud meant. It's less, "I'm gonna run the same thing in three places." It's more, "I'm gonna be running a bunch of things in three different places. How do I have a consistent control plane?" The second bit about that is that I think that what customers really want and the thing that AWS's egress fees is stopping them from doing is they wanna be able to pick and choose from all of the different vendors and use best of cloud. Google's great machine learning tools, right? But you may not think that some of their other cloud services are as robust.
Maybe you wanna use AWS, but you wanna use Google Machine Learning to do one particular thing. Or you heard about something really cool that IBM is doing on AI. You're not gonna move your entire application over to them, but you wanna be able to use that one thing. What keeps people from doing that today is the egress fees. It's not the APIs. You can figure out the APIs. It's you can't get the data back and forth efficiently. Where, again, I think our long-term value is how do we push that interconnectivity down as much as possible? How can we be that fabric that connects the various clouds together technically, but also then makes it financially attractive?
If you do that, then the multi-cloud universe isn't. You know, I'm running the same thing on three different applications. It's I'm using a little bit of Google, a little bit of Microsoft, a little bit of AWS, a little bit of IBM, a little bit of this startup over here, a little bit of who knows what in the different regions, and I get to build the best applications with the individual tools. That's painful for some of the cloud providers that really think that it's, you know, you got to go all in on my application. I think it's inevitable over time that these clouds get picked apart into what they are best at and that people use. People are gonna spend money on all of them.
I think over time, it's actually better for the ecosystem 'cause it's gonna encourage more competition and more innovation and other things. I think we are going to be the catalyst that leads to that inevitable long-term future.
Okay. We'll wrap it up. It seems to me like one of the least understood advantages that you have is your network utilization levels. I'd love for you to just touch on the comparison between your network utilization and let's say like an AWS and why AWS could or could not bring it up there. Then lastly, the relationship between increased network utilization and COGS, gross margins or however you wanna see it. Like, if you have a huge spike, if you're 60% now, you go to 90%, what happens to your margins? Like that kinda sort of like thought process.
Yeah, what we sell is different. With AWS, what you are typically selling is you're saying they are essentially renting a server to you. Now, they virtualize that in various ways, and it's not literally a server, but it is just conceptually a way of thinking it is, I'm renting a server to you. If you're spinning up an AWS in a region, it's because they are taking, and again, it's not literally this, but it's effectively literally this. They're taking a box, they have spent a certain amount for the box, and they are saying, "I am reserving this for you." The problem is that customers aren't actually that good at getting high utilization rates. There's a ton of instances out there which are just sitting completely and totally idle.
Now, AWS has to have that up and running. Their model works that way. I don't know what their utilization rates are, but the best estimates I've heard are that they're in the teens, right? In terms of what their actual utilization is. That's. They don't really care 'cause they're paying for that server, and they're getting paid for the server, and it makes sense that they've built in the models and it all is effectively. It all effectively works. What we sell is different. We're not selling you a server, we're selling you the promise that your application is going to run when you need it to run in a way which is efficient, as fast as possible. There's a trade-off for that. We don't give. You can't run certain.
You can't take SAP HANA and run it on Cloudflare Workers. Never is gonna happen, right? Now, I think the next competitor to SAP HANA builds on Workers, but there are certain applications that you just can't pick up, move and drop back down. You have to purpose-build it for. If it is purpose-built for us, now we can slice out every single function and we can say, "Where does that function make the most sense to run?" Either from a cost perspective, from our perspective, or from a performance perspective. It might not always be run the function as close as possible to the user. It might be run the function as close as possible to the data, right? We have the flexibility to be able to move that around. At some level, Cloudflare is just a giant scheduling engine.
Fundamentally, what we sell is different. It's almost one layer abstracted up. There are some things that we then can't do. But if you buy into that platform, then we can have a much higher performance, much higher availability, much higher reliability, much higher scalability, and a much lower cost. Because instead of being in the teens in terms of utilization, we're in the seventy, you know, five percent range in terms of utilization. That's just a much more efficient spend of CapEx on whatever it is we do. If that seventy-five percent went to ninety percent, that would make us nervous 'cause we know to have stability and reliability and everything else, that we can do that.
We're diversified enough in terms of our traffic and everything else, that it would mean something went wrong in terms of our procurement. You know, one of the things that our infrastructure team likes to think about in the world is sort of just-in-time infrastructure. The internal KPI they have is, can we double the size of any of our major locations within 30 days? Which is pretty amazing, right? If they can do that. Now, that's the supply chain and everything else has made that tricky, but we have done a good job of managing to that. That is why the utilization rate line stays super close.
That when it drops down, it's when we deploy a whole bunch of additional hardware or spike up, it's maybe when we have a harder time getting that. But that's what drives it. But again, fundamentally, architecturally, the reason why we have just a way we can offer the same end result, which is your function gets run somewhere, for significantly less, is because architecturally, we're selling something different. We're selling the promise that we're gonna run your application versus a reserved, you know, server that you then have to figure out how to manage yourself.
Please join me in thanking Matthew, Michelle, and Thomas for their time today. Thank you.