Screaming in the Cloud - The New Cloud War with Martin Casado

Episode Date: August 30, 2022

About MartinMartin Casado is a general partner at the venture capital firm Andreessen Horowitz where he focuses on enterprise investing. He was previously the cofounder and chief technology officer at Nicira, which was acquired by VMware for $1.26 billion in 2012. While at VMware, Martin was a fellow, and served as senior vice president and general manager of the Networking and Security Business Unit, which he scaled to a $600 million run-rate business by the time he left VMware in 2016.Martin started his career at Lawrence Livermore National Laboratory where he worked on large-scale simulations for the Department of Defense before moving over to work with the intelligence community on networking and cybersecurity. These experiences inspired his work at Stanford where he created the software-defined networking (SDN) movement, leading to a new paradigm of network virtualization. While at Stanford he also cofounded Illuminics Systems, an IP analytics company, which was acquired by Quova Inc. in 2006.For his work, Martin was awarded both the ACM Grace Murray Hopper award and the NEC C&C award, and he’s an inductee of the Lawrence Livermore Lab’s Entrepreneur’s Hall of Fame. He holds both a PhD and Masters degree in Computer Science from Stanford University.Martin serves on the board of ActionIQ, Ambient.ai, Astranis, dbt Labs, Fivetran, Imply, Isovalent, Kong, Material Security, Netlify, Orbit, Pindrop Security, Preset, RapidAPI, Rasa, Tackle, Tecton, and Yubico.Links:Yet Another Infra Group Discord Server: https://discord.gg/f3xnJzwbeQ“The Cost of Cloud, a Trillion Dollar Paradox” - https://a16z.com/2021/05/27/cost-of-cloud-paradox-market-cap-cloud-lifecycle-scale-growth-repatriation-optimization/

Hello, and welcome to Screaming in the Cloud, with your host, Chief Cloud Economist at the Duckbill Group, Corey Quinn. This weekly show features conversations with people doing interesting work in the world of cloud, thoughtful commentary on the state of the technical world, and ridiculous titles for which Corey refuses to apologize. This is Screaming in the Cloud. This episode is sponsored in part by Honeycomb. When production is running slow, it's hard to know where problems originate.

Is it your application code, users, or the underlying systems? I've got five bucks on DNS, personally. Why scroll through endless dashboards while dealing with alert floods, going from tool to tool to tool that you employ, guessing at which puzzle pieces matter? Context switching and tool sprawl are slowly killing both your team and your business.

You should care more about one of those than the other. Which one is up to you? Drop the separate pillars and enter a world of getting one unified understanding of the one thing driving your business, production. With Honeycomb, you guess less and know more. Try it for free at honeycomb.io slash screaming in the cloud. Observability, it's more than just hipster monitoring. This episode is sponsored in part by our friends at Sysdig. Sysdig secures your cloud from source to run. They believe, as do I, that

DevOps and security are inextricably linked. If you want to learn more about how they view this, check out their blog. It's definitely worth the read. To learn more about how they view this, check out their blog. It's definitely worth the read. To learn more about how they are absolutely getting it right from where I sit, visit sysdig.com and tell them that I sent you. That's S-Y-S-D-I-G dot com. And my thanks to them for their continued support of this ridiculous nonsense. Welcome to Screaming in the Cloud. I'm Corey Quinn.

I'm joined today by someone who has taken a slightly different approach to being, well, we'll call cloud skepticism here. Martin Casado is a general partner at Andreessen Horowitz and has been on my radar starting a while back based upon a piece that he wrote focusing on the costs of cloud and how repatriation is going to grow. You wrote that in conjunction with your colleague, Sarah Wang. Martin, thank you so much for joining me. What got you onto that path? So I want to be very clear just to start with is I think cloud is the biggest innovation that we've seen in infrastructure probably ever. It's a core part of the industry. I think it's

very important. I think every company is going to use the cloud. So I'm very pro-cloud. I just think the nature of how we use clouds is shifting. And that was the focus. When you first put out your article in conjunction with your colleague as well, I saw it and I have to say that this was the first time I'd really come across any of your work previously. And I have my own biases that I started from. So my opening position on reading it was, this is just some jerk who's trying to say something controversial and edgy to get attention. That's my freaking job. Excuse me, sir. And who is this clown? So I started digging. And what I found really changed my perspective. Because as mentioned

at the start of the show, you are a general partner in Andreessen Horowitz, which means you are a VC. You are definitionally almost the archetype of a VC in that sense. And to me, being a venture capitalist means the most interesting thing about you is that you write a large check consisting of someone else's money. And that's never been particularly interesting. You kind of cut against that grain and that narrative. I mean, you have a master's and a PhD in computer science from Stanford. You started your career at one of the national labs, Lawrence Livermore, if memory serves. You wound up starting a business, Nicira, if I'm pronouncing that correctly, that you then sold to VMware in 2012,

back at a time when that was a noble outcome and rather than a state of failure, because VMware is not exactly what it once was, you ran a $600 million a year business while you were there. Basically, the list of boards that you're on is lengthy enough and notable enough that it sounds almost like you're professionally bored.

So I don't, so looking at this, it's okay. This is someone who actually knows what he is talking about. Not just, well, I talked to three people in pitch meetings and I now think I know what is going on in this broader industry. You pay attention and you're connected disturbingly well to what's going on to the point where if you see something, it is almost certainly rooted in something that is happening. And it's a big enough market that I don't think any one person can keep their finger on the pulse of everything. So that's when I started really digging into it, paying attention, and more or less took a lot of what you wrote as there are some theses in here that I want to prove

or disprove. And I spent a fair bit of time basically threatening, swindling, and bribing people with infinite cups of coffee in order to start figuring out what is going on. And I am begrudgingly left with no better conclusion than you have a series of points in here that are very challenging to disprove. So where do you stand today now that, I guess, the whole rise and fall of the hype around your article on cloud repatriation, which yes, yes, we'll put a link to it in the show notes if people want to go there. But you've talked about this in a lot of different contexts. Having had the conversations that you've had, and I'm sure some very salty arguments with

people who have a certain vested interest in you being wrong, do you wind up continuing to stand by the baseline positions that you've laid out, or have they evolved into something more nuanced? So yeah, I definitely want to point out, so this was work done with Sarah Wang, who's also at Andreessen Horowitz. She's also a GP. She actually did the majority of the analysis. She's way smarter than I am. So I'm just very, feel very lucky to have worked with her on this. I'll make sure she gets due credit on this. So let's talk about the fear. So like, I actually thought that this was kind of interesting and it started a good discussion, but instead like the amount of like response pieces and like angry emails I got. And I kind of thought to myself, like, why, why are people so upset? I think there's three reasons. I want

to go through them very quickly because they're interesting. So the first one is like, you're right. Like I'm a VC. I think people see a VC and they're like, oh, lack of credibility, lack of accountability. It doesn't know what they're doing. Broad pattern matcher. And like, I will say I did not necessarily write this as a VC. I wrote this as somebody that like, listen, my PhD is in infrastructure. My company

was in infrastructure. It's all data center stuff. I had a $600 million a year data center business that sold infrastructure into data centers. I've worked with all of the above. I've worked with Amazon. So you sold three Cisco switches. That's right. I remember those days.

Those were awesome, but not inexpensive. That's right. Yeah. So I had 15. It was kind of a culmination of that experience. So that was one. I just think that people see VC and they have a reaction. The second one is, I think people still have the first cloud wars fresh in their memories. And so they just don't know how to think outside of that. So a lot of the rebuttals were first cloud war

rebuttals, like, well, but internal IT is slow and you can't have the expertise. But they just don't apply to the new world, right? Listen, if you're a Cloudflare, to say that you can't run a large operation is just silly. If you went to Cloudflare and you're like, listen, you can't run your own infrastructure, like they take out your sucker and pat you on the head. And not for nothing, if you try to run what they're doing on other cloud providers from a pure bandwidth perspective, you don't have a company anymore, regardless of how well-funded you are. It's a never-full-money pit that just sucks all of the money. And I've talked to a number of very early idea stage companies

that aren't really founded yet about trying to do things like CDN-style work or streaming video. And a lot of those questions start off with, well, we did some back-of-the-envelope math

around AWS data transfer pricing, and if our numbers are right, when we scale, we'll be spending $65,000 on data transfer every minute. What did we get wrong? And it's like, oh yeah, you realize that that one thing is up per hour, not per minute. So slight difference there, but no, you're basically correct. Don't do it. And yeah, no one pays retail price at that volume, but they're not going to give you a 99.999% discount on these things.

So come up with a better plan. Cloudflare's business will not work on AWS, full stop. Yep, yep. So I legitimately know basically household name public companies that are software companies that anybody listening to this knows the name of these companies who have product lines who have 0% margins because they're basically like for every dollar they make, they pay a dollar to Amazon. Like this is a very real thing, right? And if you go to these companies, these are software infrastructure companies. They've got very talented teams. They know how to build like infrastructure to tell them that like, well,

you know, you can't build your own infrastructure or something. I mean, it's like telling like an expert in the business, they can't do what they do. This is what they do. So I just think that part of the fear, part of the uproar was like, I just think people were stuck in this cloud or one-on-one mindset. I think the third thing is, listen, we've got an oligopoly and they employ a bunch of people and they've convinced a bunch of people they're right. And it's just always hard to change that. And I also think there's just a knee-jerk reaction to these big macro shifts. And it was the same thing we did to software-defined networking. My grad school work was trying to change networking to go from hardware to software. I remember giving a talk

at Cisco, and I was this naive grad student, and they literally yelled me out of the room. They're like, it'll never work. They tried to burn you as a witch, as I recall. And so to your specific question is like, how have our views evolved? But the first one is, I think that this macro downturn really kind of makes the problem more acute. And so I think the problem is very, very real. And so I think the question is, okay, so what happens? So let's say if you're building a new software company and you have a choice of using like one of the big three public clouds, but it impacts your margins so much that it depresses your share price. What do you do?

And I think that we've thought a lot more about what the answers there are. And the ones that I think that we're seeing is some actually are, companies are building their own infrastructure. Like very famously, Mosaic ML is building their own infrastructure. Fly.io, which is building their own infrastructure. Mighty, you know, so Hale's company, building its own infrastructure. Cloudflare has their own infrastructure. Fly.io, which is building their own infrastructure. Mighty, you know, Sohail's company, building its own infrastructure. Cloudflare has their own infrastructure. So I think if you're an infrastructure provider, a very reasonable thing to do is to build your own infrastructure. If you're not a core infrastructure provider, you're not, you can still use somebody's infrastructure that's built at a better cost point. So for example, if I'm

looking at a CDN tier, I'm going to use Fly.io, right? I mean, it's like, it's way cheaper. The multi-region is way better. And so like, I do think that we're seeing like almost verticalized clouds getting built out that address this price point and like these new use cases. And I think this is going to start happening

more and more now. And we're going to see basically almost the delamination of the cloud into these verticalized clouds. I think there's also a question of scale, where if you're starting out in the evening tonight to, I want to build, I don't know, Excel as a service or something. Right. You're pretty silly if you're not going to start off with a cloud provider

just because you can get instant access to resources. And if your product catches on, you scale out without having to ever go back and build it as quote-unquote enterprise-grade, as opposed to having building it on cheap servers or Raspberry Pis or something floating around. By the time that the costs hit a certain point, and what that point is is going to depend

on your stage of company and lifecycle, you're remiss if you don't at least do an analysis on is this the path we want to continue on for the service that we're offering. And to be clear, the answer to this is almost entirely going to be bounded by the context of your business. I don't believe that companies as a general rule make ill-reasoned decisions. I think that when we see a decision a company makes, by and large, there's context or constraint that we don't see that inform that. I know it's fun to dunk on some of the large companies' seemingly inscrutable decisions, but I will say, having had the privilege to talk

to an awful lot of execs in an awful lot of places, particularly on this show, I don't find myself encountering a whole lot of people in those roles who I come away with thinking that they're a few fries short of a happy meal. They generally are very well-reasoned in why they do what they do. It's just a question of where we think the future's going in some level. Yep, so I think that's absolutely right. So to be a little bit more clear on what I think is happening with the cloud,

which is I think every company that gets created in tech is going to use the cloud for something, right? They'll use it for development, the website, test, et cetera. And many will have everything in the cloud, right? So the cloud is here to stay. It's going to continue to grow. It's a very important piece of the ecosystem. It's a very important piece of IT.

I'm very, very pro-cloud. There's a lot of value. But the one area that's under pressure is if your product is SaaS, if your product is selling software as a service, so then your product is basically infrastructure. Now you've got a product cost model that includes the infrastructure itself, right? And if you reduce that, that's going to increase your margin. And so every company that's doing that should ask the question, like, A, is the big three the right one for me? Maybe a verticalized cloud, like, for example, whatever, Flyer, Mosaic, or whatever, is better because the cost is better.

And I know how to write software and run these things, so I'll use that. They'll make that decision, or maybe they'll build their own infrastructure. And I think we're going to see that decision happening more and more exactly because now software is being offered as a service, and they can do that. And I just want to make the point, just because I think it's so important, Exactly because now software is being offered as a service and they can do that. And I just want to make the point just because I think it's so important that the clouds did exactly this to the hardware providers. So I just want to tell a quick story just because for me, this is so interesting.

No, please. I was only really paying attention to this market from 2016 or so. There was a lot of the early days that I was using as a customer, but I wasn't paying attention to the overall industry trends. Please, story time. This is how I learn things. I hang out with smart people and I come away a little bit smarter than when I started. This is literally my favorite. This is one of my favorite topics is what I'm about to tell you, which is, so the clouds have always had this argument, right? The big clouds, three clouds are like, listen, why would you build your own cloud?

Because like you don't have the expertise and it's hard and you don't have economies of scale, right? And the answer is you wouldn't unless it impacts your share price, right? If it impacts your share price, then of course you would because it makes economic sense. So the clouds had that exact same dilemma in 2005, right? So in 2005, Google and Amazon and Microsoft, they looked at their cogs. They looked like, okay, I'm offering a cloud. If I look at the cogs, who am I paying? And it turns out there's a bunch of hardware providers

that had 30% margins or 70% margins. Like, why am I paying Cisco these big margins? Why am I paying Dell these big margins, right? So they had the exact same dilemma. And all of the arguments that they use now apply then, right? So the exact same arguments. For example, AWS, you know nothing about hardware. Why would you build hardware? You don't have the expertise. These guys sell to everybody in the world. You don't have the economies of scale. So all of the same arguments applied to them. And yet, because it was part of COGS and it impacted the share price, they could make the economic argument to actually build hardware teams and build chips.

And so they verticalized, right? And so it just turns out if the infrastructure becomes parts of COGS, it makes sense to optimize that infrastructure. And I would say the big threes foray into OEMs and hardware is a much, much, much bigger leap than an infrastructure company foraying into building their own infrastructure. There's a certain startup cost inherent to all these things. And the small version of that, we had in every company that we started in a pre-cloud era,

renting virtual computers from vendors was a thing, but it was still fraught and challenging and things that we used then, like GoGrid, no longer exist for good reason. But the alternative was great. I'm going to start building and seeing if this thing has any traction. Well, you need to go lease a rack somewhere

and buy servers from Dell, and they're going to do the fast expedited option, which means only six short weeks until they show up in the data center and then get sent away because they weren't expecting to receive them. And you wind up with this entire universe of hell between cross-connects and all the rest. And that's before you can ever get anything in front of customers or users to see what happens. Now it's a swipe of a credit card away and your evening's experiments round up to 25 cents. That was significant. Having to make a significant tens of thousands of dollars of investment just to launch is no longer true.

And I feel like that was a great equalizer in some respects. Yeah, I think that- That cost has been borne by the astonishing levels of investment that the cloud providers themselves have made. And that basically mean that we don't have to. But it does come at a cost. I think it's also worth pointing out that it's much easier to stand up your own infrastructure now than it has been in the past, too.

And so I think that there's a gradient here, right? So if you're building a SaaS app, you would be crazy not to use the cloud. You'd just be absolutely insane, right? So if you're building a SaaS app, you would be crazy not to use the cloud. You'd just be absolutely insane, right? Like, what do you know about core infrastructure? You know, what do you know about building a backend? Like, what do you know about operating these things? Go focus on your SaaS app. The calluses I used to have from crimping my own ethernet patch cables in data centers have faded by now. I don't want them to come back. Yeah. We used to know how to do these things. Now, most people in most companies do not have that baseline of experience for excellent reasons. And I wouldn't wish that on the current generation of engineers, except for the ones I dislike.

However, that is if you're building an application. Almost all of my investments are people that are building infrastructure. So like they're already doing these hardcore backend things. That's what they do. They sell infrastructure. Would you think like someone like at Databricks doesn't understand how to run infrastructure? Of course it does. I mean, like, or Snowflake or whatever, right? And so this is a gradient. On the extreme app end, you shouldn't be thinking about infrastructure, just use the cloud. Somewhere in the middle, maybe you start on the cloud, maybe you don't. As you get closer to being a cloud service, of course, you're going to build your own infrastructure. Like, for example, listen, I mean, I've been mentioning Fly. I just

think it's a great example. I mean, Fly is a next generation CDN that you can run compute on, where they build their own infrastructure. It's a great developer experience. And they would just be silly. Like, they couldn't even make the cost model work if they did it on the cloud. So clearly, there's a gradient here. And I just think that you would be remiss and probably negligent if you're selling software not to have this conversation, or at least do the analysis. This episode is sponsored in parts by our friend EnterpriseDB. EnterpriseDB has been powering enterprise applications with PostgreSQL for 15 years, and now EnterpriseDB has you covered wherever you deploy PostgreSQL, on-premises, private cloud,

and they just announced a fully managed service on AWS and Azure called Big Animal. All one word. Don't leave managing your database to your cloud vendor, because they're too busy launching another half dozen managed databases to focus on any one of them that they didn't build themselves. Instead, work with the experts over at EnterpriseDB. They can save you time and money. They can even help you

migrate legacy applications, including Oracle, to the cloud. To learn more, try Big Animal for free. Go to biganimal.com slash snark and tell them Corey sent you. I think there's also a philosophical shift where a lot of the customers that I talk to about their AWS bills want to believe something that is often not true. And what they want to believe is that their AWS bill is a function of how many customers they have. In practice, it is much more closely correlated with how many engineers they've hired. And it sounds like a joke, except that it's not. The challenge that you have when you choose to

build in a data center is that you have bounds around your growth because there are capacity concerns. You are going to run out of power, cooling, and space to wind up having additional servers installed. In cloud, you have an unbounded growth problem. S3 is infinite storage. And the reason I'm comfortable saying that is that they can add hard drives faster than you can fill them. For all effective purposes, it is infinite amounts of storage. There is no forcing function that forces you to get rid of things. You spin up an instance, the natural state of it in a data center as a virtual machine or a virtual instance is that it's going to stop working two to three years left on maintain when a raccoon calls it off into the woods to make a nest or whatever the hell raccoons do. In cloud, you will retire before that instance does as it gets

migrated to different underlying hosts, continuing to cost you however many cents per hour every hour until the earth crashes into the sun or Amazon goes bankrupt. That is the trade-off you're making. There is no forcing function and it's only money, which is a weird thing to say, but the failure mode of turning something off mistakenly that takes things down, well, that's disastrous to your brand and your company. Just leaving it up, well, it's only money.

It's never a top-of-mind priority, so it continues to build and continues to build and continues to build until you're really forced to reckon with a much larger problem. It is a form of technical debt where you have kicked the can down the road until you can no longer kick that can. Then your options are either go ahead and fix it or go back and talk to you folks and it's time for more money. Yeah, or talk to you. There is that. I think everybody should, honestly. I think this is a board level concern for every company.

I sit on a lot of boards. I see this. And this has organically become a board level concern. I think it should become a conscious board level concern of cloud costs, impact cogs, any software company has it. It always becomes an issue. And so it should be treated as a first class problem. And if you're not thinking through your options, and I think, by the way, your company is a great

option, but if you're not thinking through the options, then you're almost fiduciarily negligent. I think the vast, vast majority of people and vast majority of companies are going to stay on the cloud and just do some basic cost controls and some just basic hygiene and they're fine. And like this doesn't touch them, but there are a set of companies, particularly those that sell infrastructure, where they may have to get more aggressive. And that ecosystem is now very vibrant, and there's a lot of shifts in it. And I think it's the most exciting place in all of IT, like personally in the industry. One question I have for you is where you draw the line around infrastructure companies. I tend to have an evolving view of it myself, where things that are hard and difficult do not become harder with time. It used to require a deep-level engineer with a week to kill to wind up compiling and building a web server. Now it is evolved and evolved and evolved. It is check a box on a web page somewhere and you're serving a static website. Managed databases, I used to think, were something that were higher up the stack and not

infrastructure. Today, I'd call them pretty clearly infrastructure. Things seem to be continually, I guess, slipping beneath the waves, to borrow an iceberg analogy. And it's only the stuff that you can see that is interesting and differentiated on some level. I don't know where the industry's going at all, but I continue to think of infrastructure companies as being increasingly broad. Yeah, this is my favorite question. I'm so glad you asked. This was not planned, to be clear. No, no, no. Listen, I am such an infrastructure maximalist, and I've changed my opinion on this so much in the last three years. So it used to be the case, and infrastructure is a long history of calling the end of infrastructure. Like every decade has been the end of infrastructure.

It's like you build the primitives and then everything else becomes an app problem. You know, like you build the cloud and then we're done. You know, you build the PC and then we're done. And so there are even very famous talks where people talk about the end of systems and we built everything right then. And I've totally changed my view.

So here's my current view. My current view is infrastructure is the only really differentiation in systems, in all IT, in all software. It's just infrastructure. And the app layer is very important for the business. But the app layer always sits on infrastructure. And the differentiations in app is provided by the infrastructure.

And so the start of value is basically infrastructure. And the design space is so huge, so huge, right? I mean, we've moved from like PCs to cloud to data. Now the cloud is decoupling and moving to the CDN tier. I mean, like the front end developers

are building stuff in the browser. Like there's just so much stuff to do that I think the value is always going to accrue to infrastructure. So in my view, anybody that's improving the app accuracy or performance or correctness with technology is an infrastructure company. Right. And the more of that you do, you know, the more infrastructure you are. And I think, you know, in 30 years, you and I are going to be old and we're going to go back on this podcast. We're going to talk and there's going to be a whole bunch of infrastructure companies that are being created that have accrued a lot

of value. I'm going to say one more thing, which is so, okay, this is a, this is a sneak preview for the people listening to this that nobody else has heard before. So Sarah and I are back at it again and the brilliant Sarah who did the first piece and we're doing another study. And the study is, is if you look at public companies, and you look at ones that are app companies versus infrastructure companies, where does the value accrue? And there's way, way more app companies. There's a ton of app companies. But it turns out that infrastructure companies have higher multiples and accrue more value. And that's actually a counter narrative because people think that the business is the apps,

but it just turns out that that's where the differentiation is. So I'm just an infraaximalist. I think you could be an inframaximalist in your entire career, and it's the place to be. And this is the real value that I see of looking at AWS bills. I mean, our narrative is, oh, we come in and we fix the horrifying AWS bill. And the naive pass is, oh, you cut the bill and make it lower. Not always. Our primary focus has been on understanding it, because you get a phone number looking bill from AWS, great, you look at it, what's driving the cost? Storage. Okay, great, that doesn't mean anything to the company.

They want to know what teams are doing this. What's it going to cost for them to add another thousand monthly active users? What is the increase in cost? How do they wind up identifying their bottlenecks? How do they track and assign portions of their COGs to different aspects of their service? How do they track and assign portions of their cogs to different aspects of their service? How do they trace the flow of capital through their organization as they're

serving their customers? And understanding the bill and knowing what to optimize and what not to becomes an increasingly strategic business concern. That's the fun part. That's the stuff I don't see that software has a good way of answering, just because there's no way to use an API to gain that kind of business context. When I started this place, I thought I was going to be building software. Turns out that there's so many conversations that have to happen as a part of this that cannot be replicated by software. I mean, honestly, my biggest competitor for all this stuff is Microsoft Excel because people want to try and do it themselves internally. And sometimes they do a great job, sometimes they don't. But it's understanding their drivers behind their cost. And I think that is what is often getting

lost because the cloud obscures an awful lot of that. Yeah. I think even to summarize this whole thing pretty quickly, which is like, I do think that organically, like cloud cost has become a board level issue. And I think that the shift that founders and execs should make is to just like treat it like a first class problem upfront. So what does that mean? Minimally, it means understanding how these things break down, A, to your point. B, there's a number of tools that actually help with ongoing of this stuff. Like Vantage is one that I'm a fan of. It just provides some visibility. And then the third one is if you're selling software as a service, if that's your core product is software, in particular, it's an infrastructure. If you don't actually do the analysis on how this impacts your share price for different cloud costs, if you don't do that analysis, I would say you're fiduciarily negligent just because the

impact would be so, so high, especially in this market. And so I think, listen, these three things are pretty straightforward. And I think anybody listening to this should consider them if you're running a company or you're an executive company. Let's be clear. This is also the kind of problem that when you're sitting there trying to come up with an idea for a business that you can put on slide decks and then present to people like you. These sound like the paradise of problems to have. Like, wow, we're successful and our business is so complex and scaled out that we don't know where exactly a lot of these cost drivers are coming from. Yeah, that sounds amazing. Like I remember those early days back when all I was able to do and spend time on and energy on was just down to the idea of, oh, I'm getting business

cards. That's awesome. That means I've made it as a business person. Spoiler. It did not having an aggressive Twitter presence. That's what made me as a business person. But then there's this next step and this next step and this next step and this next step. And eventually you look around and realize just how overwrought everything that you've built is and how untangling it just becomes a bit of a challenge and a hell of a mess. Now, the good part is at that point of success, you can bring people in like a CFO and a finance team who can do some deep level analysis to help identify what COGS is, or in some cases of some founders, explain what COGS is to you and understand those structures and how you think about that. But it always feels like it's a trailing problem, not an early problem that people focus on.

I'll tell you the reason. The reason is because this is a very new phenomenon that is part of COGS. It's literally five years new. And so we're just catching up. Even now, this discussion isn't what it was when we first wrote the post. Now people are pretty educated. They're like, oh, yeah, this is really an issue. Oh, yeah, it contributes to cost. Oh, yeah, our stock price gets hit. It's so funny to watch the industry mature in real time. And I think going forward,

it's just going to be obvious that this is a board level issue. It's going to be obvious this is a first class consideration. But I agree with you. It's like, listen, the industry wasn't ready for it because we didn't have public companies a lot of public companies where this was a real issue i mean this really we're talking about the last five seven years it really is neat just in real time watching how you come out with something that sounds borderline heretical and in a relatively short period of time becomes accepted as a large-scale problem and now it's now it has fallen off of the hype train into, yeah, this is something to be aware of, and people's attention spans have already jumped to the next level and next generation of problem.

It feels like this used to take way longer for these cycles, and now everything is so rapid that I almost worry between the time that we're recording this and the time that it publishes in a few weeks, what is going to have happened that makes this conversation irrelevant? I didn't used to have to think like that. Now I do. Yeah, for sure. Well, a couple, I mean, just a couple of things. I want to talk about like one of the reasons that accelerated this and then what I think is going forward. So one of the reasons this was accelerated was just the macro downturn. Like when we wrote the post, you could make the argument that nobody cares about margins because it's all about growth, right? And so like, and even then it still saved saved a bunch of money. But a lot of people were like, listen,

the only thing that matters is growth. Now, that's absolutely not the case if you look at public market valuations. I mean, people really care about free cash flow. They really care about profitability. And they really care about margins. And so it's just really forced the issue. And it also made kind made what we were saying very, very clear. I would say, as far as shifts that are going, I think one of the biggest shifts is for every backend developer, there's like 100 frontend developers. It's just crazy. And those frontend developers... And a third of a DevOps engineer. That's true. And I think those frontend developers are getting better and better tools to build complete apps.

Totally complete apps. They've got great JavaScript frameworks. They're coming out all the time. And so you could argue that actually a secular technology change, which is that developers are now rebuilding apps as front-end applications, is going to pull compute away from the clouds anyways, right? Like if instead of like the app being some back-end thing running in AWS, but instead is a front-end thing, you know, running in a browser at the CDN tier,

while you're still using the big three clouds, it's being used in a very different way. And we may have to think about it again differently. Now this again is a five-year going forward problem, but I do feel like there are big shifts that are even changing the way that we currently think about it now. And we'll see. And if those providers don't keep up and start matching those paradigms, there's going to be an intermediary shim layer of companies that wind up converting their resources

and infrastructure into things that suit this new dynamic. And effectively, they're going to become the next version of, I don't know, level three, one of those big underlying infrastructure companies that most people have never heard of or have to think about because they're not doing anything that's perceived as interesting. Yeah, I agree. And I honestly think this is why Cloudflare and Cloudflare workers are very interesting. This is why Fly is very interesting. It's a set of companies that are like, hey, listen, workloads are moving to the front end and you need compute closer to the user and multi-region is really important, etc. And so even as we speak, we're seeing kind of shifts to the way the cloud is moving, which is exciting. This is why it's like, listen, infrastructure is everything, man. You and I, if we live to be 200,

we can do great infrastructure work every year. I'm terrified on some level that I'll still be doing the exact same type of thing in 20 years. I like solving different problems as we go. I really want to thank you for spending so much time talking to me today. If people want to learn more about what you're up to slash beg you for other people's money or whatnot, where's the best place for them to find you?

You know, we've got this amazing infrastructure Discord channel. Really? I did not know that. I love it. It's like the best. You know, my favorite thing to do is drink coffee

and talk about infrastructure. And like I posted this on Twitter and we've got like 600 people and it's this the best thing. So that's honestly the best way to have these discussions. Maybe can you put like the link in like the show notes? Oh, absolutely. It is already there in the show notes. Check the show notes. Feel free to join the infrastructure discord. I will be there waiting for you. Yeah, that'd be fantastic. Thank you so much for being so generous with your time. I appreciate it. This is great. Likewise, Corey. You're always a class act, man. I really appreciate that about you. I do my best. Martin Cassano, general partner at Andreessen Horowitz. I'm cloud economist, Corey Quinn, and this is Screaming in the Cloud. If you've enjoyed this podcast, please leave a

five-star review on your podcast platform of choice. Whereas if you enjoyed this podcast, please leave a five-star review on your podcast platform of choice. Whereas if you hated this podcast, please leave a five-star review on your podcast platform of choice, along with an angry comment telling me that I got it completely wrong and what check you wrote makes you the most interested. The content here is for informational purposes only and should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. For more details, please visit A16Z.com slash disclosures. If your AWS bill keeps rising and your blood pressure is doing the same,

then you need the Duck Bill Group. We help companies fix their AWS bill by making it smaller and less horrifying. The Duck Bill Group works for you, not AWS. We tailor recommendations to your business and we get to the point. Visit duckbillgroup.com to get started. This has been a humble pod production stay humble