The Infra Pod - Incubating Infrastructure at Scale (Chat with Dane at Cloudflare)

Episode Date: August 19, 2024

Ian and Tim sat down with Dane (SVP of Emerging Tech and Incubation at Cloudflare) to chat about Cloudflare's journey of incubating emerging infra areas like security and AI, and how it becomes products Cloudflare customers can use at scale.

Welcome to yet another Infront Deep Dive again. This is Tim from Essence VC and Ian, let's go. Hey, this is Ian Livingston trying to sneak into a platform. And I couldn't be more excited today to be joined by Dane Connect, who is early employee at Cloudflare, currently runs Emerging Tech Incubation. Ian and I have had several awesome conversations. I couldn't be more excited to actually record one this time for everybody else. So, Dane, could you introduce yourself, tell us a little bit about yourself,

and what in the world have you been up to at Cloudflare recently? Yeah, first of all, thank you, Tim and Ian, for having me. I'm excited to be here and talk with you guys today. At Cloudflare, I've been there for about 12 years. I run a group called Emerging Tech Incubation, which I think we'll touch on a little later. I hope everyone's heard about Cloudflare, but if they haven't, just at a high level, we are a giant edge network. About 20% of the internet flows through us. We have data centers in 300, over 300 cities, and we're 500 pops. We're a few milliseconds from almost everyone.

And our goal is to build a better internet. And by that, we mean, you know, protect people from bad guys from making it hard to get to your content and also make your site faster and safer in the progress and have now kind of turned that network into being able to do it. Security and development platform for CrossFit Enterprise. I think Cloudflare is such an interesting story.

It's been around as long as I've been building infrastructure. Get started is like basically like a firewall, like a nouveau waif,, right? How did Cloudflare get started and where is it today? Because those are not the same thing, right? So how did the company get started? Where did it initially build? What was the use case? And now what's the delta between that and today? Matthew, Michelle, and Lee, the co-founders, kind of came up with the idea. People should be able to protect their websites without having to go buy hardware. And they could do it by leveraging data that they had from a project called Project Honeypot, where they had data about largely who's good and bad on the internet.

They kind of figured out that the best way to do that was to have it as a proxy close to the users. One of the things that usually comes with security is sometimes there's these trade-offs you have to make. And so they quickly realized that you don't want users to have to make those trade-offs. And so performance and security are intrinsically tied together at Cloudflare. When we say make the web better, it's not just safer, but it's faster as well. That was kind of the

origin. There were a lot of network effects that kind of played into that, and that's why it kind of started as a free service where we could start to protect people and understand the internet at a large scale with a large amount of diverse population and slowly build up the product to serve the largest enterprises in the world with that information and learnings that we learned from the free customers. Awesome. And today, Cloudflare is just like large infrastructure providers, not just this secure proxy for your website to protect you against CDOS and other types of attacks.

It's this, you can build a full stack on Cloudflare. How did we get here? What was the journey? What was the thought process and the journey to come from? We're going to protect your website. I have compute. I have network. I have data.

I have everything I need to build a website. A lot of it came from our own internal need. We're kind of our own best customer. But the way I think about it is kind of in different acts. The first act was we built this giant network, and that's really our core asset. And the first thing that we did really well was we protected your infrastructure. We protected your APIs, your websites, your servers. We protected them and made them faster.

And then so it kind of naturally kind of progresses to, okay, well, if we already have this network that's close to the users and we're great at processing traffic in one direction, you know, we're kind of the front door. Why can't we also be the back door as well? And how can we protect your users and devices and offices from the rest of the internet? It started as, you know, some of our engineers love to watch the graphs on Grafana and didn't want to have to log into a VPN. Logging into a VPN sucks. And so the first thing we built over a weekend was with a couple of engineers, actually, kind of like the SRE team, where we built it so that you could log into our Grafana, our internal dashboards, with something we called EdgeAuth at the time. And it worked with our IDP, so you could authenticate it. And it went through our Edge. That was kind of the early days of

access where we could just kind of built it for ourselves. You know, obviously, then that kind of grew into what the full zero trust security suite is. As Convert got larger, we are one of the top targets on the Internet. And so, you know, we have to build quite a lot of tooling to protect ourselves. And so, you know, once we know it's good enough to protect ourselves, we want to allow our customers to use it as well. And then kind of the third act was as Con conference got larger and more and more engineers, we kind of wanted to continue to innovate. We have a reputation for shipping a lot. Things got kind of slowed down in the 2016 timeframe and kind of looked at why.

And it was because we were building a lot of great technology, but we weren't building a platform that made it easy to onboard new developers. And so we kind of had to step back and said, okay, well, how do we make it so that it's easy to build applications on us? That was, I mean, for ourselves, easy to build applications. And they said, okay, well, if we're going to do this, not all great ideas are going to come from us in the long term. So let's make sure we do it in a way that's a platform. And so over time, we can expose the great primitives that we have for building distributed global applications, not just to our own employees, but every enterprise that wants to be able to use it.

Amazing. And at what point did this go from, hey, we're just solving some internal use cases, we're going to actually turn this into a function? What you do today at Cloudflare is you actually run incubation, right? And my understanding is it's a completely separate organization and there's a lot of thought behind that. So can you help us also overlay, when did this go from R&D to now we're going to have a full, we're going to have, like, a full, like, we're going to focus on incubation as an actual business function versus, like, just a thing that happens off to the side and as a result of culture? Yeah, so, I mean, just taking a step back, explain what

emerging tech incubation means at Cloudflare. Cloudflare has kind of R&D split into three different horizons. Our CEO is a student of Clay Christensen and, you know, innovators to Limba and, you know, wants to make sure that we are, you know, not getting disrupted from above or below. And the way you do that is by continuing to innovate. You continue to ship new things, solve new customer problems. And, you know, it's best done when you're not having to kind of make tradeoffs with what you need to do next quarter. We have a core product engineering group like most companies do. That is a large part of our revenue and what we sell to our customers.

Their focus is largely, you know, what do we need to deliver to customers that we've committed to them? What does sales need to close deals? And how do we, you know, continue to build those products and kind of incremental innovation there? And then we have a working tech incubation, which is kind of like the next horizon. It's about, you know, 10 to 20 percent of our day spend, where my team is focused on how do you grow TAM? How do we find new use cases for our network that we can bring more value to our customers for every bit that already flows through us? When you take away the pressure of revenue, the teams can really swing through the fences. And my team's job is to not make a plan and get 90% or a delivery on time.

It's to take a lot of shots on goal. And some of those will be successful. Some of those won't. We like to build things where we get them out quick and early to customers to start getting feedback. And then we invest behind that demand. Sometimes the demand is crazy. We have to double down real quickly and quickly mature those products. And sometimes the customers are like, that's an interesting technical solution

or it's an interesting problem, but just not much demand. And then we kind of just kind of hold it off to the side. What's interesting though, is over the longer time horizon, a lot of those ideas kind of come back. About four years ago, we launched an AI primitive

for our workers developer platform. It's our serverless developer platform and got no traction. We made a lot of press, but did not get much customer demand. Obviously, over the past couple of years, that's, you know, people's appetite for AI and building AI-enabled applications has changed quite a bit. And, you know, the demand was crazy. And so we said, OK, let's pull this off the shelf.

And we've already de-risked it. We've already have a good understanding of what we need to build and how to build it. And we're able to move real fast to build that. And, you know, we launched Workers AI last September and growth has been amazing. And then we have a third horizon, which is our research team. They're less about kind of building what the next new product is, but they're trying to think about what are some of the hard problems that we have to solve either for us or for the internet? What are the new protocols that need to be built in order to enable things on our network? We did a lot of work early on with cryptography and trying to drive down the cost of SSL to make it so that we can launch a universal SSL for everyone. They also did a lot

of work with privacy, push privacy forward with like masks. And so then they kind of feed the ideas and that core technology into my group. And then we, you know, hopefully kind of figure out what the business application is and how to build a business around it. Eventually it kind of graduates over into the core organization when that flywheel is going and, you know, great product market fit,

great roadmap, you know, early customers. You kind of think of it like a series B, you know, series C company. And then you get to put the entire Cloudflare, go to market org behind it, and then you can just let it fly. So I'm curious, like, obviously Cloudflare, go to market org behind it, and then you can just let it fly. So I'm curious, like, obviously Cloudflare, you know,

like I said, started with sort of like a one surface area, but it grew and grew and grew. At this point, we're looking at the products, actually touches not just the network and security from different angles, but applications and workers and a lot of stuff.

So there's so many things you can actually try to find research opportunities and innovation opportunities. And AI now is kind of going crazy. How do you decide what to actually focus on at a point in time? Like there are probably layers, different layers are having different sort of state-of-the-art expansion or adopting different things.

Is there a sort of like a framework or a way you tell folks that this is how we should think about what is worth pursuing as a group? Do you try to pick only a few to go for or do you try to have like a little bit more breadth for search? I'm just curious like how to decide what to actually try to like innovate here. Yeah, so, you know, in that kind of that first horizon, the corporate organization, a lot of those

ideas come from talking to customers, from Gartner, from things that we need to innovate on the core products. And that makes sense because their job is to grow revenue. The emerging tech incubation team, our job is to find new opportunities. And Tim, you invest in a lot of companies. It's hard to pick the winners. You do that by having a portfolio and taking a lot of shots on goal. Don't try to pick the winners too hard up front. We do try to focus them around different areas. In Cloudflare, we want to grow the amount of customers that we have, obviously, bring more value to the customers. So those are just kind of like large North Stars there. How we get there is ideas that people are passionate about. The best ideas for me is when an engineer comes and you know that large North stars there, how we get there is, you know, ideas that people are passionate about. The best ideas for me is when an engineer comes in and like, you know, you know

that they're not going to sleep until they can prototype it, prove it out, get feedback. So, you know, all the ideas is if someone comes and pitches me something and you can just tell that they believe in it, you know, we can pull off one or two people to work with them for a month, a quarter to try to prove it out. You know, sometimes they become huge successes. Sometimes it's things like the early AI work, where like it was cool technology, but the market wasn't ready for it. And sometimes, you know, just not a fit. And, you know, I think the key there is we celebrate failure. The successes are just as good as the failures because they usually lead to us learning something about the technology, about the platform, about customers, about different pain points and

go from there. No kind of like strict, stringent framework there. The key is finding great people that want to work on those problems. I'm curious, you kind of mentioned that you have a portfolio approach similar to a VC. I'm very curious in your experience, I mean, you've been doing incubation at Cloudflare for many years. You've had some successes, obviously, like workers, R2, et cetera. In your experience, does it have the same sort of power law distribution in that a few of the batches actually end up being like 10, 20, 1,000x in terms of return, and the same with the VC does? Do you think that holds true both in incubation at Cloudflare's size

in comparison to Tim's VC portfolio? I mean, absolutely. The winners far outpace the small wins or the losers as far as their impact on the business. And how do you manage sort of the life cycle of a bet? Like, I'm sure you have some type of internal process. Hey, we're going to spend six months. Do we have some like KPIs or something

that you measure over a period of time? I'm also curious to understand, like, how does a business think about financing innovation? At some point, you're going to have some headcount pool that I assume that you're working with. How do you think about the function of an innovation lab, broadly speaking? When we budget for the year, emergency tech incubation gets a fixed part of the budget. It's not the idea that each project you try to go fund it. It's that, you know, kind of broad liberty to be able to spend that money and go experiment broadly and quickly. But, you know, we have to stay within

that envelope. Once those ideas, you know, kind of the flywheel gets going and they do need to grow and accelerate, I would have the choice of either, you know, try to put all my head against that, but then I'm no longer making a lot of bets, or that's when the right time is for me to move it into the core organization where they can then go invest in it in a way that's more correlated with revenue and traditional opportunity from that perspective. Different bets take different time horizons, but in general, we're able to get something off the ground, especially because we have the workers platform that we build on top of ourselves. Within a month, most things you can get an in demo. Within three months, you can probably

ship something to the customers that they can start to test and get feedback on. We have a great pool of people that are always willing to test things and give us feedback on them. And the process of what wins and what doesn't is a little more organic in the sense that, you know, if the product manager and the engineers don't hear feedback that they want, they kind of shovel themselves and start moving on to the next thing. They want to be working on the things that are going to bring the most value to the company. So we haven't had a case really where you have to go say, okay, let's shut down that product kind of from top down. It usually kind of happens more from bottoms up.

In your experience, are there rules that you've developed around doing incubation? It's like, hey, I know that this type of stuff just doesn't work. These are things I refuse to do. How do you figure out whether something's a good thing to incubate? In addition to that, at what stage can you incubate something? Is there some prerequisites you have to have as a company to know, oh, this incubation has a chance of being successful because we have this platform built? What's the mental model for thinking through the prerequisites for a good incubation?

In general, I prefer guidelines, not rules. Rules can kind of be a filter for things that are kind of outside the box, not ever kind of reaching their potential. My goal is to obviously hear everything. Early on, I kind of thought about it in the traditional sense of the TAM is this or within 18 months we could get to a million or 10 million or 100 million in revenue. I quickly realized that that was actually

a bad way of thinking about it because that's actually how disruption happens when companies get too focused on that. Really, it's really around adoption. I want people to build things and if they can get people using it, in general, we'll figure out how to generate and share value that we're creating there.

So I think, you know, we're going to go many, many places when we kind of deeper dive into infrastructure, fun stuff here. But, you know, AI is everything. You know, you have a dedicated page for AI. So I see that Cloudflare has already started to have quite a lot of, you know, message and features and

products talking about AI and using AI here. I'm curious, how do you see where AI fits into the Cloudflare strategy here? Because when I'm looking at the product offering right now, there is inference part of it. There's application building parts of it. As a VC and, you know, Ian, we're all looking at all kinds of AI frameworks and architectures and models and everybody has like their own like take on the future. Just curious, like how do you see Cloudflare's advantage being able to like capitalize on this AI wave here? Yeah, so there's two sides of that. So one side of it is, how do we use AI in our products today to deliver better products and more vital to our customers? From that perspective, I mean, we've been using AI ML since almost the beginning.

Every time a request goes through us, it gets scored to help determine whether it's a good or bad request, or if it's a human or not a human, and how we should process it beyond that. And that's actually part of why the great network-effective Cloudflare, the fact that we have, you know, the free traffic as well as the paid traffic. We see everyone on the internet multiple times a day. And so we can use that data to train our own models to build better security products. We can use that to answer questions customers might have on their data, their traffic, using

AI inside of our products. And then the other side of it is, you know, in our developer platform, it's how do we kind of offer the infrastructure for companies that want to build the next great AI company on top of us? You know, Workers AI sits inside of our serverless platform where it makes it really easy for someone,

you know, you're building your full-stack application on us. You want to add AI and AI enable a feature. You know, we have all the models that are sitting on our edge that are powered by GPUs that are, you know, one hop away from the users and make it just super easy for them to add it to their application.

We think it makes sense for inference to run at the edge. You know, I think from a privacy perspective, it makes the most sense for inference actually to run on device. But, you know, you even see with Apple with their recent announcements, that even with people with the best chips, that's not always going to be possible.

And so the next best place we think for running inference is on the edge, where it can stay in region, local, both from a performance and from a data privacy sovereignty perspective as well. When you kind of think about what we do and all the AI companies that you're funding and everyone else is funding, today I think about 70% of all AI companies' traffic is behind Cloudflare. And that's largely using everything from our developer platform, but also using the application security products. And that's because one of the largest costs for AI companies is obviously the GPUs and there's a lot of bots and malicious traffic that I can kind of not just DDoSing your server, it's now DDoSing your GPUs and the bills that you're paying from behind it. So being able

to have covers in front of it and be able to protect and ensure it's only those good users that are getting to your services. Our customers have found a lot of value in that. I'm curious, like as you think about AI, like today you're talking about like network protection effectively. It's like, hey, we have this amazing suite of protective software that protects you from malicious attacks and malicious users. How do you think about the future of AI on the edge and the way that we build our apps? Has Cloudflare developed a mental model for how you think about what the future architecture of an app will look like in AI? How much of that lives at the edge? And I'm curious, have you thought about specific drivers?

There's a reason for developers to move from centralized cloud to distributed cloud, like edge cloud or distributed cloud, however you want to frame it. The fact that we're moving compute away from a centralized core to someplace else. I'm curious to understand how you think about that trend in AI and how Cloudflare fits into that picture and what levers are going to drive that transformation.

Yeah, I mean, kind of going back to the phone and privacy, I mean, the questions that you ask AI and the inference can be sensitive. And I think someone who cares about privacy in general, I think you want to keep that information as close to you as possible. In the best case, it's actually not even on the edge. It's running right there on your phone

or a device in your house. But from a performance and a cost perspective, that's not always going to be possible. So then the next best place is at the edge. And then if you can't do it at the edge, then you want to do it at a more regional compute hyperscaler. But someone on the other side of the world, the last thing that they want is their data to go all the way around the world

to have it processed in AWS East. I think those workloads in general are just going to have much more privacy and data sovereignty. A lot of countries are getting ahead of the data problem here and actually saying that they want their citizens' data to stay in-country. Early on with OpenAI, I believe Italy for a while, they had to turn it off because Italian Data Privacy Org wanted all the infrastructure to stay there. I think they were able to work it out and be able to open it up tiddly.

But I think that's going to be the trend where developers used to not have to worry about data localization. They worried about speed. But I think more and more people are going to have to worry about where am I storing my users' data? And it's not going to be good enough to say, you know, even I have an EU and a US data center. People are going to want it actually in their countries, whether it's India, whether it's Singapore, Australia. They want their data sitting there with them and with their users. So I think naturally running the AI on the edge

where you can have an application that kind of natively sits there and already today natively scales horizontally, you press once and you deploy to region earth. It makes sense that you'd want your inference there as well too. Absolutely. I'm curious, do you think we're at a point where like developers actually have all the tools they need to truly build edge apps? Or is there like big gaps that are missing that actually enable developers

to come in and build a full-fledged app that entirely runs at the edge? Largely, I think that we have the primitives, but I think a lot of the tooling is not there yet, whether it's the frameworks that the developers rely on to be able to take advantage of those primitives. Ideally, the developer doesn't have to think about sharding their data across databases in 100 different locations. And that's possible on Cloudflare,

but you have to build those pieces to get there. I think some of the hard work is done on the infrastructure side, and now it's the DX side that needs to catch up to make it to the point where it's easy for a developer to build in this kind of distributed manner as it is for them to build an application that they deploy to a single server. And how do you think, do you think this is one of those transitions that is like a whole new evolution where everyone's going to be building basically a new stack in the same like in the same way that we kind of went from, you know, there's some people that like lift and shift to the cloud, but broadly speaking, it was like lift, shift to the cloud, then rebuild. Like, do you think this is like, what do you think the motion in the ocean is going to be as this trend broadens out?

And as we think about some of those DX primitives that enable people to have apps to the edge, like, do you think is this this another way, sort of like on-prem to the cloud, cloud to the edge is a way? Is that how you think about it? Or how have you all thought about that, broadly speaking, both for Greenfield but also Brownfield apps? Yeah, I mean, early on, a lot of the early adopters of our platform are Greenfield apps just because it is such a different architecture and way of building, way of thinking. But obviously, after you get to those Greenfield apps,

you know, a majority of the world's workloads are still, you know, in these legacy apps. Some people haven't even finished the on-premise cloud migrations. And so, you know, a lot of things we've released in the past few years are to be able to help companies that wanted to start to peel off workloads. So, you know, we have a product called HyperDrive,

which allows you to connect the edge to your databases sitting at the hyperscalers. You know, you're not going to move your Postgres database from one of the hyperscalers to the edge right now. And most of your application logic is going to be running right next to it already. But, you know, you can start to peel off things like peel off, you know, authentication, peel off the user experience, the web presence, and then start to peel off

data aggregation as you're collecting data from users. And you can pre-process it at the edge before sending it all the way back to the hyperscalers. I think right now we're at that phase where people are starting to take those brownfield applications and peel off the workloads that make sense. And so I don't think there'll be rewrites from the ground up. I think they'll, you know, just be a more of a, you know, as they need to add new functionality and refactor, they will kind of choose the best tool for the job.

So I'd like to move to a section, our favorite section called a spicy future. Spicy futures. As you know, you might know, Dane, what is your spicy hot take of the infrastructure landscape giving your role and your super biased opinions here? I just mentioned that people are just finishing

their migration to the cloud from on-prem. I think we might be hitting a cycle where it's kind of going to bounce back. People are going to start to realize that they're spending a ton of money with these hyperscalers and they're going to want to start putting data back in their own data centers or on the bare metal where they can have greater control once they're at scale and start building back on-prem for a lot of things. Excited about companies like Oxide Computer where they're kind of rethinking what on-prem means and being able to provide some of the things that

people thought were annoying about on-prem earlier and just kind of offer that all in a box. And with that, you then do lose some of the things that you get at scale with a hyperscaler, like the network protection, things like that. But kind of self-servingly, I think that works out very well for Cloudflare in the sense that we can then go sit

in front of your infrastructure, whether it's at the hyperscalers or whether it's on-prem, and you can decide for yourself which workloads you want to move to the edge, which workloads you want to serve on-prem, but have one control plane that kind of sits in front of it and kind of routes everything and protects everything. So you believe in like a hybrid approach,

that we're going to move some workloads to go to on-prem, some will go to the edge, and kind of like the hyperscalers who don't have edge facilities will kind of suffer? Yeah, rarely do I think, you know, one size fits all works for anyone. And so I think the smartest companies will look at, you know, what they're trying to build and the problem they're trying to solve

and go pick the right solutions for each of those components. And I think that they will find, you know, over time that, you know, going back on-prem will be much more economical and easier to manage and make a lot more sense for their business over time. And when I say hyperscores, I also think the same goes for a lot of SaaS services. I see some applications today that they built their business logic, but then there's 20 different SaaS solutions that manage everything from, you know, their login to their logging to their payments. And, you know, while some things, you know, definitely belong at their party, you don't want to do things like payments yourself. I think a lot more people are going to want to

self-host, you know, some of those building blocks of the application. I think we'll see a pullback to pulling things back into your application and kind of owning your full stack. And largely, that's just because things have gotten easier. It's easier to build those things yourself. You don't need the extra complexity of all those vendors. I guess the question is, do you believe in the unbundling of the cloud? Like I think there's a famous Andreessen post written by a mutual friend of all of ours, I'm sure, where they talk, Andreessen talks a lot about like, you know, the hyperscalers,

two pizza box teams, they can only build so much, so they get constrained in terms of DX. So I guess what I'm hearing from you is you don't necessarily believe in the unbundling. You kind of believe, in fact, more of that spend is going to move out of the clouds and we'll end up spending some amount in cloud, but broadly speaking, we'll continue to buy more from the hyperscalers instead of this burgeoning, broad set of SaaS services for auth, identity, and all these other different aspects. I think all these things are cycles, right? You're always in an unbundling or bundling cycle.

In this case, we're starting to move back to a bundling cycle. You see that a lot, I mean, whether it's on the security side with consolidation of vendors on the security side. I talked to a lot of CISOs. They talk about managing 20, 30 different vendors for that and trying to get, you know, goals of getting that down to under five. There's a time and place, you know, in an innovation cycle where you have a lot of niche vendors that solve things that the larger bundled vendors weren't solving. And then those other vendors kind of catch up or acquire those companies and they start to bundle things back together. And then, you know, the cycle just repeats itself again.

So in the spirit of Spicy Takes, you know, if you were advising an entrepreneur who's starting out, great team, money's there, and they're trying to figure out where they should, you know, place a bet, where do you think you would send an entrepreneur? Like, where do you think they should spend their time? Where would you suggest they should go look for opportunities to innovate on their own? I still think there's a lot of opportunity and security. I think that problem is still getting worse. So I think there's a lot of

places to innovate there. Most people don't set out to build a platform. They're usually solving a problem. And so I would still tell those early stage entrepreneurs, just keep going. If they think they have a better mousetrap and they're solving something uniquely, go and solve it. And over time, they'll either turn into a platform or they'll hopefully get acquired by a platform. I would still tell the entrepreneurs to take whatever ideas they have and if they're passionate about it to continue to explore it. And the great ones will find a way to make it a large, big business. So you mentioned this bundling and bundling side. We're seeing the cycle always happening. But I feel every cycle,

you're also seeing the environment, the paradigm shift. Plus there's also probably more enabling infra that's required, right? Because Oxide, you just mentioned, that's down to hardware level. If we're going to rebuild up the convenience of cloud, we'd certainly need more than just servers. That's more easier to be purchased and plugged in you know and cloudflare sort of sits in the complete almost like opposite end of the world right do you see what's the immediate need of abstraction or infra that's

required to get more and more people or companies much very simpler to build so just like on-prem data sort of infra applications like i know we're seeing more and more frameworks already but in your eyes what is like yes i mean infrastructure is code uh companies that are kind of building those paradigms to make it easier to manage not just on the cloud but on-prem or really a hybrid environment there are a lot of opportunities there to help companies manage it and seeing that across both the how do you manage your infrastructure how do you you know do, how do you do security, and across those hybrid environments,

I mean, that's probably where I would go invest. Players like Terraform and those type of tools usually have a lifetime just because the way that they were designed and built was amazing for migrating to the cloud. But if people want diverse options, there's probably some new paradigm that needs to be found

that will better serve those things. And what's the next Terraform that's going to be built that makes it easy to manage what infrastructure looks like today? And do you also have an opinion of how this new Terraform that requires to push

and help orchestrate and manage your, I guess, your hybrid architecture, how does Cloudflare also fits? Like the worker side, the edge side also fits here? Do you see the opportunity where it's also extended? In a cloud world, there's like multi-network companies helping to bridge the gap between, you know, on-prem and your cloud. If you go for on-prem, is there like opportunity to kind of connect to the edge in some more

direct way? Absolutely. We acquired a company earlier last year around multi-cloud networking. And more and more, we are getting asked to sit in front of much more diverse infrastructure. And so we needed a way to be able to communicate back to those infrastructures, whether it's in one of the cloud providers or whether it's on-prem. More and more people are saying, how can I expose these internal private resources to workers, as an example? And like HyperDrive, I mentioned before, how do you expose an internal database to the edge?

Or Cloudflare Tunnels, how do you expose a container that's running in your data center to the Cloudflare workers but can still run everything that Cloudflare does great and should be distributed and should be close to the users, still out at the edge and not have to worry about those things as you kind of build things that make sense to be built on-prem there. It's never going to make sense to do training at the edge, right? Cloudflare, we're much more power constrained at each of our locations

and training obviously requires a huge amount of power and data store and it makes sense to do training and it's not, you know, real time and, you know, there's not those latency requirements. And so it makes sense to do it where there's cheap power and we're not trying to compete for every single workload.

But we want to be that connectivity cloud that can connect everything together. Be the mesh, the glue that holds it together. That makes a lot of sense. I'm curious how you foresee data management of the future, right?

Where you have some data at the edge, you have some data centralized, that data's going to be centralized for training. Have you thought broadly about what the future of sort of the data stack looks like and what data lives where, what tiers? A lot of things that we're trying to build and add to the work of the platform still

from a product perspective is around data management. And, you know, and added queues, announced kind of pipelines and workflows. And that's all about making sure that you can ingest data and do the computation that makes sense out there. And then you can move it cheaply and quickly back to the places where you need it for those type of applications. Cloudflare's R2, which is our S3-compatible blob storage,

kind of launched with free egress. We're not going to hold your data captive. We're going to let you remove your data anywhere you want for free so that you can do what you need to do with the data at the different places. Awesome. So, Dane, we touched on so many topics today,

and I'm not sure there's so much more we could, but just given the constraint of time, we'll definitely wrap up here. What is the best way for folks to learn about innovations happening in Cloudflare? Do you guys have, do you, or do you guys have like a certain, like a publication or a community

or some sort of places to go find you or learn more about this? I mean, definitely the Cloudflare blog. Our blog is focused around technical blog posts. We try to make sure that we explain how we build things and the process

and the technologies behind them. We have a Discord community and then, you know, I'm always available on Twitter. Thanks, Dane. This is super awesome.

Folks, if you're interested in Cloudflare, check out Twitter and all this Discord and all the communities here. But thanks a lot, Dane, for having you on the pod. Yeah, thanks a lot. Thanks for having for having you on the pod. Yeah, thanks a lot.

Thanks for having me. Amazing. Thank you so much.