FinOps, AI, and the Cost of Cloud Chaos with J.R. Storment

Transcript

Corey: Welcome to Screaming in the Cloud. I'm Corey Quinn, and today my guest is something of a member of the Pantheon of finops royalty. If I can abuse two metaphors and smash them together, JR Storm is the executive director of the finops Foundation, but also one of the OG folks in the space. You were a co-founder of Cloudability.

Uh, thank you for. Taking the time to speak with me and avoiding strangling me.

JR: Thank you, Corey. I gotta say it's like. It's super surreal to be here because, uh, you know, as some of the audience may not know, you and I were kind of on like opposite ends of this spectrum in terms of what we thought about the space in the early days, and it's like really great that you've invited me to be here because it's like getting to know you and, and your process has been an amazing part of the last few years.

Corey: We, we've. Passing like ships in the night for a very long time, even before the Enos Foundation existed. I had a bit of insecurity around it where, uh, I go and I talk to these companies as an independent consultant, pred Duck Bill, and like, oh, so you do what, what Cloudability does? It's like, yeah, but I'm.

Better looking and it, it, it, it doesn't work. But I do remember that you were very kind and sat down with me for half an hour or so at reinvent at a bar years ago. Mm. And it was so validating because we were talking shop and so many of the things that I had a strong suspicion on at one point you, like I, my comment was, oh yeah, it feels like a strong majority of the spend, despite all the stuff they talk about is just EC2.

Maybe that's just the small number of customers I talk to. And you've busted out your phone with data on it. Like, ha, there's a reason you think this, it, it was so validating and it was, I appreciate you're taking the time to talk to someone That, frankly, was more a little annoying in a bunch of different ways.

JR: Your, your, your brand of snark though is one that I've also actually been, uh, not annoyed by, but intimidated by. I'll say, because you, I'll say at the start of this, you have always been a lot funnier than I am, and you've been much better on the, on the camera and on the podcast, and so I've actually, for the long time, and I'll, I'm being very honest and direct here, like.

I, I was also intimidated and, and by you and what you were doing, because I saw you as sort of the, the, uh, the better version of me who was out there able to like, get the jokes going and actually dig deep at this stuff. And I think, you know, to the conversation that you and I had. Uh, way back then and sort of over the years.

Um, it's kinda the reason that I, when I left the platform space, I got into this, which was technology doesn't really solve the problems. There's actually like a whole bunch of cultural learning company organizational stuff that like guys like you have been doing with those organizations for a long time.

And so, yeah, when I. Left the platform more. It was like, I wanna get closer to guys like you on what you're doing.

Corey: I had no idea. I just assumed you're like, oh boy. I need to find something more depressing than cloud billing. I know nonprofit world. We're gonna dive into that and see what happens.

JR: I did not expect this to go where it, where it went, honestly.

Yeah. I thought it was gonna be like a a side thing while I figured out what was next and put a little time into it and the. Foundation just kind of grew up really

Corey: quickly. History's not yet written. It couldn't just be this thing you detoured in for a decade or two and as a side project while you figured out what was next.

Yeah. It turned out what was next was retirement and okay, we can, we can workshop the narrative. It worked out.

JR: I'm, I'm six years into it now and I see a path for the next three to four years, so I, it might just end up being a decade. How long have you been in this?

Corey: Oh, dear Lord. Uh, I got surprise fired from BlackRock.

The best way to get fired, uh, in the second half of 2016, I spent some time puttering around, but beginning of 2017 is when I really got serious about it. Figured it'd be an 18 month side project. Uh, wound up after 18 months, taking a few interviews at various places. Mm-hmm. Got an offer from Amazon. I've

JR: heard about this.

Corey: Yeah.

JR: The famous offer. Yes.

Corey: Yeah, which was, I was, it was weird because I interviewed there twice. One was six months beforehand. And they're like, he, no, because they do have hiring standards. Six months later I interviewed for a different role, got the offer, turned it down because of their crappy non-compete.

But the original hiring manager who is in the space, Keith Jar, for years thought that I was just telling lies about like, we didn't hire you. What's the deal with this? And I told him at a car ride at Reinvent one, he's like, wait. You looped a second time. He's like, oh wow. I had a very different opinion of this.

It's why would I lie about something so banal?

JR: Well, this is, I think, maybe vindication for both of us. 'cause I feel like this whole space that we're in, which, by the way, you did not want to say the name of this space. I remember you tweeted at some point I was very hurt. I think you said. I shall not say finops.

I shouldn't say finops. And for the longest time I was like, God damn that Cory, he's trying to like cut.

Corey: I hate the term. And it won. Yes, and I will be fair. A lot of this comes from my historical background hating the idea of DevOps as a job title.

JR: Okay. Yeah.

Corey: I called it cloud economics, which is very intentional because Yeah, it makes sense.

Like I'm a cloud economist. What does that mean? Well, it's two words. No one understands, uh, cloud is a bunch of other people's computers. And economist means I claimed to know a lot about money and dressed like a flood victim. Put them together and no one knows what the hell I'm talking about.

JR: Yeah, it makes sense.

Well, look, I like, I, I didn't create the term ops and I didn't, wasn't a big fan of it either when I saw it, but when I saw it, it was like, oh, wait a minute. Na Naming things as hard as, you know. Oh,

Corey: yeah.

JR: There's all this stuff that people have been doing. Like it's, they're, they're like operating in cloud differently.

They're trying to integrate that into the business. Engineers are like, cost, what is this? The finance people say cloud, what is this? And when it started to kind of like pop in there, I was like, okay, this may not be, this may be an annoying name. 'cause by the way, it's a. It's a Port man tow, right? Mm-hmm.

Just like DevOps. It's not actually financial operations. It was clear, like it's, it's the short version. And I mean, I had a lot of those conversations the early days as we were out there talking about this stuff. People were like, no, I'm, I'm never gonna say that, bro. No, no, I hate that. And then by the end of the lunch they'd be, they'd be using it 'cause it's just shorter.

Corey: I have pivoted on this even in our marketing now and the thing that changed is our customers started calling it that, and I'm not gonna gonna fight that particular tide, but I called myself a cloud economist, and I'm sure you don't remember this, but back when the finops Foundation originally launched.

At an event you threw in San Francisco and invited me to February

JR: 19th, 2020.

Corey: 2019. 2019. Yeah. Yeah. I'm like, this is great. Like, can I join? You're like, well, you're a vendor. No. It's like, and at the time you were at Cloudability, it's like, well, I, I have some opinions on this, but you sat me next to Owen Rogers at the time of 4 51 0 4 5 1.

JR: Yeah.

Corey: Yeah. Before he went to, I wanna say s and p acquired those? I think so,

JR: yeah.

Corey: Ancient history, but he apparently had a PhD in what he had RET on as cloud economics. He's like, oh good, someone else. I can talk about the deep academic structures of this. And it's like, okay, I have two paths before me. I can come clean and.

Talk. I was like, alright, I made the title up because nothing else seemed to fit.

JR: It's a good title.

Corey: Or I can try and white guy bluff my way through it, where at that point I wind up like with a book deal or something. And this was before LLM. So I came clean rather than over confidently sounding wrong.

And here we are.

JR: And here we are.

Corey: Yeah,

JR: yeah, yeah.

Corey: And here we are right now where we have just seen the 2026 State of finops report. There were. A lot of trends evidenced in it. Uh, few surprises, but one that I'm going to hit you on because why not? There is, there are two different approaches, AI for finops and finops for ai.

And some of those, uh, one of them, and I don't remember, which was the top of the list of concerns, and one was six items down. But in spot checking, talking to folks who are in the finops world, people always forget which one is which. So I have questions as to how accurate the relative placement of those two things are.

JR: Okay. Okay. Um, well, I should disclaimer that those rankings vary greatly. If you look at what people are doing today mm-hmm. What their absolute pain point is like today, Monday morning when I walk in, versus what they're being asked. To consider and what they're afraid of in the coming 12 months. And I would say the thing that everybody wants to talk about, 'cause it's super sexy and it's fun and you know, like automation, it's gonna take all of our jobs away, is AI for finops, which is, how do I use Theis to get better at the finops?

But the thing that people are actually struggling more with now is the finops for ai, which is like, I've got all this ai, I'm not sure. What I'm getting out of it. I don't know what the value is. I can't forecast it. I can't allocate it. It's very much like, I don't know, Amazon and you know, 2013 before we had billing data or something like that.

So we're at that same point. So the two sides of the coin are obviously connected and I think everybody really. Wants to push and from a vendor and practitioner side, like how do we use this new technology to, I don't know, even get rid of finops, but we're not really seeing that materialize yet. We're seeing the, like how do we use the finops principles to apply to a new type of spend that we're being asked to manage and the type of spend that is a.

Going up faster than cloud since, well, probably since 2013.

Corey: Yeah. The numbers are smaller, but the growth rate is meteoric,

JR: which is exactly what we were saying 10 to 15 years ago about cloud. Like it's a tiny portion of your it, but the numbers are very

Corey: fast. Five to 7% of infrastructure spend right now is on AI ish.

But it's hard to say 'cause what counts as an AI service. I saw an Amazon job ad years ago. They called S3 and AI service. So we're just, we're just making stuff up now. Totally. Yeah. It is a differentiator. Uh, if you, I do all kinds of nonsense with. AI because why not? I think that there's some whimsy to it that people are sleeping on, but I'll use the Anthropic API directly instead of using the same model, same cost through Bedrock.

And the only reason I do that is Anthropic will let you spin out an API key for a project and show you what the cost against that API key is in near real time as opposed to get a good night's sleep. And then Amazon will tell you the same thing if you set it up correctly. So when. This may shock you. I wrote some crappy code and it started eating itself and burned through $15 in credits.

I knew what app was doing it and when I thought I fixed it, but didn't, I found out real quick

JR: one, so that what you're doing right there is exactly what I think we're seeing people do more today than what's coming is mm-hmm. We've cut our teeth on how to do this for Amazon. We cut our teeth on how to do Google and Azure and OCI, wherever we are.

And now it's like, wait, this, this isn't like a whole new set of concepts we need to come up with. We just need to, and hopefully it's your Anthropic example. There's better billing data examples than we had in the early days. 'cause yeah, when we were starting all this, you know, for you and me, a lot of these people in this a long time, I mean.

Finops didn't really come to be until we had the proper billing data from the providers that we could do things like allocate down to the resource, down to the second, you know, figure out what things were applied where. And we're entering a world where you've got all these AI services that are consumption based, which makes 'em very cloud-like.

And as a side note, we're getting a bunch of SaaS services now that through their application of AI models are becoming much more cloud-like. Everything is now sort of built on what you use, which is breaking a lot of models, uh, of what has traditionally been seat based licensing and these things. But all has to say these are kind of solved problems in finops land, right?

It's like you get the data early into the hands of engineers. In your example, you try and, you know, forecast that based on some allocation, you figure out where there's waste inefficiencies. Like you tie that back to executive priorities. These are things that company's been doing for years on the big clouds, and now it's like, let's apply it

Corey: to you.

Make sure that the engineers. Are brought to care about in the proper strategic way. I, I, yes. I have this problem myself all the time. Yeah. Where left my own devices. I'll sit here and try and golf 10 bucks a month off my $50 development account with mm-hmm. And I'll forget the whole Oh, right. I cost money, don't I?

I embezzle at in office supplies every week. It's fine. It, although with a partial remote office, your company's a little more just, you can, there you go. Exactly. It's just one pocket of the other pocket. Don't tell the investors they won't mind. Like, that has been a, a significant shift is watching folks care about it.

The other side of the coin is, oh, I'm gonna build an AI bot that will optimize your cloud environment. And I have yet to see one of those that was not. Abjectly. Terrifying. It works for the common simple case. Yeah. But the edge cases cut you to ribbons. Mm-hmm. Mm-hmm. I have people, a number of people now have demonstrated them to me and shown demos.

It's like, cool. Even in your contrived demo, here are the edge cases where in a certain context it just. Caused more problems and it fixed

JR: well. And, and not to like continue to be even more the gray hair guy. But like I just flashing back to 20 15, 20 16, 20 17, where the hot sauce and what was to become finops was automating actions.

And people were starting to do that. And there were a set of tools that did that. And platforms I worked on did that for a while. And the actions like were very valuable in theory. And if they were operating in a perfect, pristine, clean environment where everything was known and you know, but in the end, like.

People struggled to automate a lot of the things because there were, I don't know, edge cases

Corey: or

JR: perceived edge

Corey: cases.

JR: There were perceived edge cases, there were fears, there were trust. There were security. And I, and I feel like we're kind of right back there again, where I will say my mind has been changed in the last 18 months.

A bit on, uh. LLMs and chat bots even for understanding data because two years ago people were like, well throw a chat Bott on it. I'm like, that's dumb. I don't wanna chat with a chat bot. But now I'm like, oh shoot. You can get some really interesting insights out of that. So I think there are some,

Corey: they need to be validated because they're not always correct.

But it gives you interesting, in interesting threads to pull on. Yeah. Think is the right way of doing

JR: it. Exactly. So I, I think there actually might be some juice that comes. It is coming now from. Even that chatbot approach with finops, but there's still a, I mean, are you gonna throw like agentic, like against managing your infrastructure?

Maybe down the road, but there might be some trust issues in the meantime.

Corey: I have always been surprised. Every time I talk to a customer it, it hits new, a new where. So much of this is about psychology more than it is about arithmetic. Yeah. Uh, my first big customer in 2017, um, they needed to buy back In those days, our eyes, uh, $18 million was the right buy.

And that's a scary number. No, no one wants to hit more than they're likely to make in the course of their career on the buy now button. So they ham and haw analyze it for four months. It's look. You're gonna spend the money one way or the other. Maybe it's not the right number. Great. Cut it in half. Make the buy.

See what happens. Also, I know this works with AWS and it works with other clouds as well. Okay? Once you figure out what your buy is, they have a concierge team you can have make the purchase for you. Well. Why would I do that? Because when someone screws it up and they will, you definitely want it to be on them to fix it and make it right, rather than you sweating, blood, wondering if, I don't know, you're about to lose the entire contract and get sued into oblivion.

'cause someone selected the wrong region. That wasn't me, that was the concierge group. They fixed it because it was their mess up. They're the cloud provider. They can make it work on the backend. I don't wanna have to call in favors every other month because I didn't read a spreadsheet properly.

JR: Your story there is super reminded me of one of the first finops talks I gave in like 2017 to reinvent.

And it was about a giant RI buy with seven figures that went wrong.

Corey: Mm-hmm.

JR: And it went wrong because somebody misread the data and they made a big purchase and they were bringing their own Red Hat licenses, but they bought like, you know, standard Linux and they bought the RIS and they sat back and they waited for some number of months and didn't pay attention to anything and suddenly the bill was twice because they had the RIS and they had these things still being used.

But that's kind of the point of this stuff is like. Finops as a discipline. I like to talk about, not as about technology. It's a, it's a people cultural discipline, and that's good and bad. The bad part is people make dumb mistakes, right? Yeah. And, and they're hurrying. And they're rushing. But the good part is, and I think a lot of the, the ethos and conversation, the space has shifted is.

It is less about cutting costs, which is like the RI lever. And it's more about like, well, and AI's a great example of this. How, how do I get the right value out of this stuff? And some of that has to, what all of it should tie back to the business strategy and things. But some of it has to be like conversations between teams about like, why are you doing this?

Yeah, well I'm doing this because we're getting pushed from the board on using in this way. And that's okay right now. 'cause we're focused on growth, right? We're not focused on cutting, cutting, cutting.

Corey: Yeah. Optimization. Lowering the bill is a point solution in many cases. Mm-hmm. It's something that people care about at fixed points and then they go back to the things that they care about.

I think it's a terrible business in many respects, where, oh wow. Your, your, your tool's knocked 40% off my AWS bill, which is a weird marketing in itself. I mean, I cut my AWS bill by 95% by not checking my keys into GitHub anymore. Who knew? But mean, which is not even a

JR: joke. I've heard so many actual stories of that over the years.

Corey: But what have you done for me lately? Oh, wow. You added a lot of value upfront, but now you're just sort of maintaining it and people aren't logging into it. The tool story doesn't work as well. Also, the, the cloud providers themselves are getting much better at this, and the baseline level of knowledge that engineering teams have about the thousand weird pieces of trivia that around cost optimization.

JR: Mm-hmm.

Corey: Yeah. Are better known these days. I used to wander into large companies and they. Hadn't bought our eyes in 18 months and they were passing data back and forth like a hot potato. I don't see those large misconfigurations anymore. I see small stuff. But once you tell them about it the first time, they internalize it and they start looking for that.

And what my job here has done instead, it's the ongoing, how do we predict it? How do we understand it? If we're making a commitment to our, to external or internal, we're gonna spend how many millions of dollars a year? What's the right number to commit to? Mm-hmm. There are consequences to this.

JR: Well, you're hitting so many things in there.

One of the interesting things that came outta the, the data from the state of ops this year was that while optimization is still the top like pain point priority I have today, it was very fur much further down on what people are. Skilling up and looking ahead on, partly because of, I think what you're talking about, which is that stuff has become a required table stake and we've gotten a lot more mature at it.

And, you know, a couple years of the conference we jokingly had, I think this for you and I, were, were buddies and you were there, but we jokingly had this, this phrase that was, um, there's no runners. Because we, we kept hearing from people and fops space, they'd be like, 'cause we have a crawl, walk, run, model maturity.

Oh, we're, we're runners, we're, you know, we're, we're really advanced. And it's like, yeah, no, you are really advanced. All the stuff that we were talking about like two or three years ago. Mm-hmm. You have gotten there, great. You're doing your job right. But now there's a whole bunch of new stuff that's come and the current hot sauce, of course example is like, well, how, how are you applying this stuff to your anthropic spend?

How are you applying it to your opening AI spend? How are you, are you using MCP in your finance practice? Like these are things that the bar is constantly being, I was gonna say moved, but I guess raised in this conversation.

Corey: People also are, are of the opinion that their organization is the only dysfunctional one out there.

Oh, they're all

JR: broken.

Corey: Yeah. Someday I hope to have a customer who will explain their tagging strategy and coverage without sounding embarrassed while they do it. Because like, well, this is nowhere near what most of your customers have. It's like, well, I, I have a surprise for you.

JR: Isn't that such a funny thing?

So we have been hearing you and I respectively, from different customers that same challenge for over a decade.

Corey: Yes.

JR: And everybody says, well, that's. That's the simple problem. We'll automate that away. The AI will take everything, will take care of that, and yet no one has been able to move to that like perfect allocation model.

Corey: I, I will, I will push back on that. I think that tagging is one of the few areas where there is strong AI promise, because my, my baseline premise is you can adjust the number however you'd like, but for. Conversation's sake, 80% of the time it's gonna get it right and 20% of the time it's going to get it wrong.

So, mm-hmm. What areas is this acceptable in? Like, well, it's gonna handle our security. It's like, well, that's not gonna work so well. Uh, but tagging it's, you are already at 0%, so you're starting from nothing. Go down that path. It's better than what you have now. And, oh, we misallocated this one workload. It was really not the kiss of death and it's better than what you had before.

JR: Ward Cunningham was the invent of Wiki. He's from Portland. It's where I started, previous company, and he had a law that was, it was how Wiki was founded, which was basically the best way to get the right answer is to propose the wrong answer.

Corey: Yes,

JR: it's colloquially known as the the blank page problem.

Right. And so that's what I make instead of

Corey: a second account and, and Aer and uh, confidently assert something that is incorrect. Right. And your PhD's dropping out of the woods correctly.

JR: Exactly. Yeah. Yeah. Right. I'm not gonna contribute unless you say something wrong, then I'll tell you how dumb you are.

Right. That's

Corey: exactly,

JR: and so in some ways, I mean, I think that's, yes, that's a great use case of AI is like, well, why don't you do the 80%, the good enough? Pass and get us allocated, and then the humans will come and like, wait a minute, we need to clean.

Corey: Well, we're just gonna lean on our engineers to make sure that they tag things right.

Yeah. Yelling at them hasn't worked for over a decade. Uh, why is that suddenly gonna change now?

JR: Okay, so you didn't ask this question, but talking about state of finops and one of the challenges that has always been the historic thing is like, can't get the engineers to take action. I, I think what we're realizing now in today's, like more integrated finops land where finops reports into the CTO or CIO, we found like 78% of the time now where it's like part of the function of not just you need to spend less, but like this is part of your job now.

It's part of being a good engineer and the way the security is, the reliability you need to have, like, I don't know, keep the company bottom line working so that we can pay your bills. And as we're getting a better integration of engineering practices with the finops side of it, that problem of like.

Engineers not taking action was never really a problem that engineers were being lazy or weren't paying attention. It was, they weren't being told by their boss to care about this. Right. And so we're seeing that I think happen now where finops is getting positioned in a spot that they're getting guidance down from the C-suite.

There's not like a seat at this executive C-Suite table, but they're getting guidance about, we need you to do more ai. We need you to, oh by the way, we used to have one cloud, now we have three. 'cause we've signed deals. I need you to not only be efficient in those clouds, but make sure we're hitting the commitments that we've made to these clouds.

Oh, and we've got a data center that we've kind of just led and been letting sit there. We should probably like manage that a little more actively. So we're seeing these things come together into a much more consolidated practice. Now

Corey: I am going to, uh, take a break here, and this episode is sponsored by.

Well, us over at Duck Bill HQ where we are solving problems for the large, convoluted enterprise in enterprise environment. Uh, I'm gonna say that it is strongly endorsed by the finops Foundation to watch him flops. Like it's not, how, how do I wind up? Not saying, not not contradicting him, but No, it's this, it started off as a consultancy.

We now have a product called Skyway. Uh, we recently announced a round of funding. We're actively hiring as well. If this is something that sounds, huh, I sure wish there were a solution for large enterprises and also a bit more whimsy in our lives. When having conversations with the folks behind it, please reach out duck bill hq.com.

Remember, you can't duck the duck bill. Bill is absolutely not our motto because my business partner won't allow it.

JR: Corey, we said no sales pitches. What happened, man?

Corey: Exactly. There we have to pay the bill somehow.

JR: What I can say is that uh, Corey and Duck Bill are members in good standing of the finops Foundation and they focus on an area of finops, which is contract commitments and negotiation.

That is actually quite unique as vendors go and I'm glad you brought up that. See how I managed to not say you're great at it or anything I just said you do

Corey: it. Exactly. Terrific. We have to upgrade our sponsorship tier for that.

JR: Just

Corey: find the foundation brought to you by Duck Bill. Yeah. Like one of the post acquisition, like, like Splunk, a Cisco company at the bottom.

Same approach.

JR: Is this like a hostile takeover that you're gonna be doing here?

Corey: Oh, we're all friends. How hostile could it possibly be? I'm gonna need to raise a little bit more money first, but yeah. Turns out we have to buy the entire Linux Foundation to pull it off.

JR: I, I can connect you to some people if you, if you need to talk to somebody.

Um. So, Jo, joking aside, uh, 'cause I'm not as good at that as you are, you know, like, so you, you, you've always focused on this contract stuff, and I, I see you talk about contract negotiations out there. I see you list three letter acronyms that I, I didn't think that the cloud providers wanted listed in public.

Of course we don't, we don list those three letter acronyms. Uh, but I think in today's finops space. What you're doing is especially relevant because where it used to be like, yeah, go optimize my AWS bill. Now to that point of, I've got two or three clouds already. I've just heard a story from a huge like fortune hundred tech company where like we've got really good at finops and three clouds and then, uh, this guy named Larry called our CEO and suddenly we have a fourth cloud provider and we're being told we need to shift this many workloads to it.

And oh, by the way, don't reduce your, don't miss your commitments on these other three cloud providers. And so suddenly people are like. I'm worried about optimization because if I miss these numbers, I'm gonna pay for nothing. And so they need to get really into how do I make sure I'm getting the contracts at the right time, rightfully, all those things.

Corey: Oh, absolutely. And contracts are always a moving target because these things don't exist in isolation as you allude to. It's not a great, we're gonna suddenly triple our use on one cloud provider. What does that imply for everything else, because you're not an island usually speaking. Mm mm-hmm. Uh, this also gets far more complicated in most of the environments I tend to work within because what, in all the demos we see of the various products from the cloud providers themselves, the contrived stories, it's always.

Basically startup or scale up environments where there's one big workload and that drives the bulk of what you do. Great. I have, the majority of my customers have very large cloud estates, but there is no, uh, majority even plurality of workloads that driving the spend. It's great. Our largest workload. 10% of our entire cloud bill.

We just have several thousand workloads that need to be addressed. It changes the story of how you do these things when you're running just one big workload. Well, cost or architecture and optimization are awesome. We're a giant conglomerate, and each environment in different clouds runs things very differently.

How do you start optimizing that? It's a different story. In many cases, the deep architectural juice isn't worth the squeeze.

JR: I think there's a time and a place for it. Right. Of course. That gets to like, we don't wanna just optimize or do anything in a, in a, in a bubble.

Corey: The consultants or a vacuum rather.

Oh yeah. The consultants. Uh, rallying cry, it depends So much is highly contextually dependent. It's weird because a lot of this stuff is not as hard as it is made out to be in some cases. Mm-hmm.

JR: Mm-hmm.

Corey: But there is complexity to it, and especially in large environments, all of it is complicated. Uh, there are increasingly, even in the stuff that we've built there, we don't have an optimization component publicly of Skyway.

But on the internal version where I was using it early on for consulting projects, everything it found that looked like a misconfiguration or something that could be optimized was listed as a curiosity. When spend goes up. It's not red. Mm-hmm. We're not presenting a frowny face. That's because we don't have the context to know whether it's good or bad.

Oh, it's really bad. Our cloud spent spiked last week when we ran the Super Bowl at like, maybe that's not terrible. Wow. A good news. We dropped our cloud bill to zero by turning off production. I'm not allowed to do that. Effective as it might be. Yep. So there's nuance to this, and when you start eroding your own expertise by, by not appreciating the context.

Complicated customers, which is most of them. On some level, it erodes trust you. You cannot, like, this is the consultant approach. You cannot be wrong, which sounds like I'm gearing up to be incredibly obnoxious. But in practice, what it means is, I don't know. Let's find out is my most common answer to a lot of these things.

Mm-hmm. Mm-hmm. Because if I wanna just be confident and wrong, well, AI can do that. I'm a white guy with a podcast. I mean, being confident and wrong, you'd think would be my shtick, but computers have supplanted me.

JR: Yep, yep. Um, you, I'm gonna go back to you said trust. Okay. So, so trust, I think is, is nailing it.

So the, the other perennial problem, uh, allocation aside, I think in this space is I just need to understand my spending. And that seems like a solved problem now for folks like you and your customers, I think, and a lot of people in our, in our space, but we're, we're back to that space with ai. There's two sides to this.

There's the, like, everything seems different. It's all new. It's it's tokens, it's input, it's output, it's like it's inference. What training, what is that like? There's all this new concepts, but then it's all still. Usage optimization, rate optimization and these basic things. And what we see happening though a lot of the time with the trust aspect, and, and this is one of the reasons that I think you guys are participating in this now.

We, we, we ended up launching that, that focus product, which is the, the standard for open, what, what was starting as cloud usage and billing data, but has expanded pretty dramatically in the last few months. Um. It's because people don't understand billing data coming outta providers. And the perennial problem back to the, like brass or the, the table stakes is everybody wants to talk about optimization.

Everybody wants to talk about automation. Can we auto automate away the optimization? But first the thing is like, well, what are we actually spending it? And more importantly, what are we gonna be spending at the end of the year or later so we can make this commitment? And that's where we see this important like underlying, um, you're an AWS guy as James Greenfield at AWS said like, focus creates a substrate.

Underneath of a consistent billing data so that all you smart people, whether you be vendors, consultants, practitioners, can build on top of that and get to the outcomes.

Corey: And we are, I I, one of the most impressive things you either personally or as a foundation have done is driving adoption of the focus standard across all of the hyperscalers.

Like my, my question always been how did you, how did you do that? Reminds me of the old saw of every car is an Uber if you have a gun, but it's the how, what arm twisting did you do? And it turns out that you have the, the ultimate arm twist available for providers, which is. Your members are all customers and they all care about this.

Yeah. And any provider that does not at least pay some lip service to caring about the customer experience is not destined for greatness

JR: in my experience. So yes, you could say it's an arm twist, but like, okay, so you're a, you're a a startup guy officially. Uh, sure. You build a lot of companies. I've been in that world for a lot of, a lot of worse years too.

And the thing that I've, it's, it's the reason I'm doing the foundation and the reason it caught me soft guard back to like, how do we start this whole thing was there's nothing better than product market fit. There's nothing better than that moment where things just start to go even when you're like not even pushing.

And I'm really excited, like literally at this point in time with focus, because this last. 45 days, we've started to see something that is new for focus, which is we're not twisting arms. Mm-hmm. So in the last 45 days, just randomly we find out about new implementations that neo clouds and, and like AI providing.

The new hyperscalers are doing. So like neas launch focus support without us, even we, we heard about it in their blog post,

Corey: which means they don't have to reach out for help. Like, I don't understand the spec. It, it's,

JR: it's documented. It's the flywheel source. It's so, they, they launched it on their own. Um, you know, we did our summit announcement last week of, of, uh, various things like the state of ops and our mission changed.

And a few hours after that we hear Versal, uh, provides, I don't fully understand the services. I should have my research, but it's another ai. Provider launch focus support, and they're on the version 1.3 and they're the first provider in the world on the version 1.3. We haven't talked to them, we didn't know they were doing it, any of these things.

So this is the beauty of when you find that thing that is the problem. Right. And I, and I think you guys might have this hopefully with what you're building as well, right? Is is you put it out there and people come to it and the be the fun part, and this is the challenge also to the nonprofit open source world.

We don't control who uses it. We don't control who builds it all we can do. Is hope the spec does the right things, put it out there, publish it in GitHub and hope people go, you know, that's good. We're gonna use it. Right? And now that's starting to happen. And that flywheel is really exciting.

Corey: And as a nonprofit, you generally don't drop $20 million on a saturation billboard campaign in airports of focus 1.3.

Now with extra ai, it'll, it'll be awesome. Uh, at, at finops X last June, but this is the last time I really went in depth with this, with a number of customers because you get them all in the room and start talking about things, and eventually you run outta stories about kids and pets and you start asking substantive things.

Uh, the big complaint they all had was that the, the major, the, the hyperscalers were doing this, but. Moving up the stack, the SaaS provider coverage was somewhat dismal. Yeah. Is that still the case? Are there trend shifts there? I, I have confessed, I have not looked into this since last June. We need to soon, for obvious reasons.

JR: Turns out you have other things going on. Yeah. Right. Um, so right after we launched Focus, I went, I went on, I went on a bike ride. I live in San Diego with, uh, I'm not gonna say the company name. We could probably figure it out there. There's a, there's a very advanced finops practice down there. Mm-hmm. At a big company.

With a guy who worked there has been in ops for like a decade, and I was like, Hey, you gotta, you gotta check out this cool new focus thing. You know, like two or three clouds are supporting it. He was like, I ingest data from 67 different types of spend mm-hmm. Into our reporting practice. I got a couple clouds, I got, you know, data clouds, I got SaaS providers, they're all across the board.

He was like, I have no use for what you're doing. And I kinda like, oh, bummer. You know, we're not there. He was kind of right as to where this stuff needed to go and what we're seeing happen. So right now there's um, I don't know, there's 18 people supporting focus maybe. But a year ago there were like three where we're seeing a lot of the uptick is not like we're twisting the arms of these cloud providers.

It's all these longer tails of companies, SaaS providers, data cloud providers, like they're being asked by their customers like, Hey, we've got this great finops practice is managing our cloud with this really great billing data. We'd like to do the same with yours so that we can keep giving you more money.

We need to show value from it, you know? Can you maybe use this spec? 'cause if you use this spec, then we can just start ingesting your data. We can prove the value and it flows through. And what we're seeing with these new companies that are popping in. Rather than what Amazon had to do. And we talk about this in the, the old finops book, which was like 10 years of iterations on getting through to the data before the data became the ker file.

And that iterated, they can drop right in with like a blueprint that's open source that's aligned that they can get on GitHub and use for free.

Corey: Oh, Kerr not exist in its current form if the focus back had existed back then.

JR: Well, the Kerr has a lot of stuff that focused doesn't do. Right. And it will probably for a lot of years and, and it's not

Corey: skipping all this.

Cynical things I could say in response to that that are accurate, but a little too honest. Uh, Amazon is embracing it was cut

JR: to two hours later. They're still talking about the curve file. Yeah.

Corey: Two

JR: hours later.

Corey: And finally, Amazon is thrilled to talk about this. They threw the, your, the Fennels Foundation, head of the focus spec on Matt Cowert.

They threw the two of us on stage together. That's Yeah. At reinvent last year. And by the way, surprising with a few guard rails.

JR: I I was a little hurt that I didn't get picked to speak with you. I was like, dang it. They wanted Matt,

Corey: I, I lobbied for it, but at the same time, it's, I don't know, like I, I don't know how they picked, that's why

JR: you've been very kind to let me come on your podcast

Corey: so I get stay with you.

Exactly. This is the consolation prize. I

JR: always, it, it totally is. It's like a dedicated hour with, with Corey. Now

Corey: what I found though is that Amazon has, they do legitimately care what customers say.

JR: They

Corey: listen and. The smaller providers, which is basically everyone else have to, so how legitimate that customer obsession focus is, and those other focus is almost irrelevant because they have to do it where the, where you have, you've reached a tipping point where it is absolutely driving what needs to happen for serious enterprises.

Mm-hmm. Uh, the, the coverage is increasing across the board. I think that we're now. At a point where what version of the focus spec are they supporting, is the good conversation to have instead of what's focus.

JR: That's the the problem right now too. Yes. Is what version, because the problem we've realized in the space, and you've probably heard it, and it was in the state of ops data, is now there's so many different priors and so many different versions that people are unclear how to put that data together and not everyone was conformant to the actual versions.

And so we are, we had just announced, and we're working on the, the larger process, but a conformance program for this because Amazon's on 1.2. And they did, they did good stuff with 1.2, which is they launched documentation for it. They, they kept 1.0 around, in case you're using it. There's now some one provider, 1.3 we here.

There's a bunch more coming as part of that. So one of the things we need to reconcile in the coming years is how to put that all together. Now I'm gonna do the twisting of the arm of the provider thing, which I think aligns with what you guys are doing. The coolest thing about 1.3 is actually most closely aligned to what I understand your company does, which is it gets into contract commitments.

Mm-hmm. Both allocating spend against them, helping you figure out burn downs, giving just a vehicle a format to align what is basically rack rate costs and uses data to some form of negotiated agreement. I'm not gonna say the three liter acronym, of

Corey: course,

JR: some form of negotiated agreement. So right now none of the major cloud providers support that 1.3.

Think I don't know the data well enough and not speaking to anyone. I don't know if any of the major cloud providers currently make available a programmatic way to get access to that often complex and negotiated three letter agreement, pricing agreement that they have.

Corey: In looking at large customer bills, I periodically have seen.

Errors in the discounted pricing that looks suspiciously like you would get if you type out something in Microsoft Excel, which tells me they don't have a whole lot of programmatic, uh, wisdom on the other side too. And this compounds, uh, one of our customers said, uh, that their team negotiates 300 SaaS contracts every quarter.

It is, there is a massive explosion of this. So it's a target rich environment. Yeah. And they all have different pricing models, some of which are weirder than others. I do see usage based and consumption based pricing. Mm-hmm. Uh, taking over at some point, providers like it and customers. We'll accept it on some level.

There is a lot to dislike about it, but I wanna be clear that all pricing models are terrible. You have to find the one that sucks the least.

JR: Yeah. Yeah.

Corey: Like it, it's every time I would talk to people about my consulting work over years, they would think about it for half a second and then bust out with, you should charge a percentage of savings to which my response was, oh my God, I hadn't considered this.

This is gonna change everything.

JR: Years of that conversation for me too,

Corey: the least painful part we found there. Flat rate. Everyone knows in advance what they're going to pay and they love it. It's like the exact opposite of an AWS bill.

JR: Well, um, companies like predictability, right? And that's where we go back to ops and this whole space is about predictably understanding your costs.

And sometimes they'll take predictability over, I might be able to get a little less spent,

Corey: and it's counterintuitive, but the larger an environment gets, the more predictable it becomes. You tell me you're spending $80,000 a month on AWS. That could be anything in the universe more or less. You tell me you're spending $400 million a year on AWS.

I can tell you what your top five services are and what your architecture generally looks like without a whole lot. 'cause yeah, maybe you have some skunkworks group that's doing something bizarre with document db. Yeah, that's not gonna be the number one driver of of spent. It's gonna be EC2 RDS, the usual suspects.

Yeah. The challenge there is that when you do have weird things, misconfigurations, it gets lost in the noise. Like if I steal your credentials and mine a hundred thousand dollars a resource worth of Bitcoin or whatnot. That gets lost there. Whereas on my $80,000 bill that stands out from Orbit.

JR: Well, and I think that's hitting the point.

I know you're are, are you just focused on a s now or are you doing other cloud providers?

Corey: We are doing other cloud providers.

JR: So you're gonna have like last week in Oracle Cloud newsletter, last week in Huawei Cloud newsletter.

Corey: Uh. I do have last week in azure.com. Okay. I have for years, I'm shocked that hasn't come

JR: out

Corey: yet.

We've waiting for years. We're a sign up sheet, and I have one for last week in Oracle Cloud. Okay. Which is written in the form of a cease and desist, because of course it's

JR: okay.

Corey: Hey Annie, Jack as can start a law firm. But it takes talent for that law firm to also make databases. Now, the Larry Ellison slander is necessary.

It's, it's, it's my brand.

JR: I, I, I dunno what you're talking about.

Corey: Exactly. I was profiled many years ago in the New York Times, and the closing comment, like, it was like, what? Who is this? Wackadoo was the premise of the entire article.

JR: You've had some amazing press over the years, by the way.

Corey: Honestly, it's amazing because I don't dunno when to shut up.

JR: It's, it's very impressive. Yeah,

Corey: yeah.

JR: Well, thank you. Yeah. Yeah.

Corey: And at the end, it's like, I, I try, I don't punch individuals. Because it feels like it's punching down. But I make a special exception for Larry Ellison because quote, nobody likes him. An Oracle spokeswoman did not respond to an email requesting comment on Mr.

Ellison's popularity is how they closed the article, and that is one of my favorite things I have ever read in the newspaper. So.

JR: As fun as it is though, what we have seen happen in the last couple years is that there are four legitimate cloud providers selling Oh, Oracle. Oracle Cloud itself

Corey: is an amazing product

JR: and But selling into big companies, right?

Yes. And this, this is, this is the challenge, I think where the contract stuff comes and where I was kind of asking about your. Scope and purview now because we're finding that yeah, you need the specialists in the area. Mm-hmm. But there's a rolling up that's happening not just of your cloud providers, but with your SaaS providers, with your data center costs, with your AI spend, your open AI anthropic that's occurring, which is kind of the reason that, I mean, the focus spec is covering all these things, but we're also seeing, I think an elevation, and this is actually one of the question I had view as well, about your practice.

Two part when, how are you looking at other areas, but also how are you starting to get. More integrated into the executive conversations when you're going in to talk about a $400 million bill negotiation.

Corey: Mm-hmm.

JR: So is it a, yeah, two part. Is it a cross cloud and, and what are the execs thinking?

Corey: Yes, it's cross cloud and the trick, it's, we're still coming at it from the old consulting approach where we have demonstrated expertise and value.

And for us, the measure of successful consulting engagement is the end of it. They say, hell yes, can we do it again? So we help them with an AWS negotiation or whatnot, and then like, Hey. We're doing one with GCP or Azure, or we're thinking about using Oracle, but I don't know if I can provide the physical security for my kids necessary for that negotiation, et cetera, et cetera.

And there's a, can you help us with this? And initially the answer was, well, that's not really what we focus on, turned into. Well, let's take a look and see what happens. Yeah. And that turned into, oh. There are patterns here too. Well, as a bootstrap company, we were always, uh, let's, let's stay on our, stay in our lane and focus on the thing that we are best at.

It is smart.

JR: It's smart.

Corey: As a product, we have the option to expand that.

JR: Congratulations on your fundraising, by the way.

Corey: I appreciate that. You, that was cool

JR: to see. Yeah,

Corey: I, I am, I'm a little, I am confident in the future of Oracle Cloud. To some extent and the, all the existing hyperscalers, I'm a little more reticent on the path of a lot of the neo clouds.

Mm-hmm. Because in every case where they're providing GPU to these companies, folks aren't reaching for a neo cloud as Option A. They're doing it when, well, AWS claims that there's cloud scales forever. It doesn't source tried it. Mm-hmm. And okay, we need to go wherever the GPUs can be had. That feels like a temporary inflection

JR: supply and demand.

Sort

Corey: of 10 years from now, you're telling me that all the majors are not gonna have access. Has to enough GPUs for stuff. Mm-hmm. I, I question that.

JR: Yeah. I mean, brass tacks on the additional vendors though, as well. Sure. I mean, adding additional vendors into that mix also often helps with negotiation with existing vendors.

Right. So I I have to imagine there's a lot of that happening with the companies as

Corey: well. Yes. And people get that wrong a lot. Mm-hmm. Where they think the right way to negotiate is to pound the table and threaten to move your all in on AWS 15 year estate to another cloud provider, to which Amazon's corporate response is the version of.

Bet because everyone threatens that. Yeah. Yeah. If you, they will fight for, it's

JR: like when you walk into the used car lot and you're like, I'm gonna give you, and they're like, nah, we're just,

Corey: yeah. Right. You, they, it has to be a credible argument if you're gonna play that path.

JR: I, I can feel both of us getting pulled though into the like, all in oh, one cloud thing or, and or the like, you know, is there a best cloud all?

Maybe you're not, but I, my mind is going to this Oh. Constantly in the sense that I think what the reality has happened to the, it depends thing you said earlier is that we're in this super multi-cloud world, not because like. Brighter A, B, or C has like the better offering or better deals, better rates.

But because business is complex and what we hear most of the time, where this complexity comes from is m and a activities, right? Somebody goes and buys some company, I had two clouds, now I have three. I buy a company they come with. And then with that brings the people and skill sets, but it also then brings the integration of the concepts.

And this is where. To the naming things is hard and the reason we're doing focus and the reason we kind of started the foundation was getting all the people managing the different types of technology together to talk and use the same language is what you need to get the trust. Because even in Amazon, like a dollar is not a dollar.

You got blended rate, you got unblended rate, you're gonna amortize unamortized or you all these things. And so I think you know the magic of what all this is and where I, I hope to see a lot of what you guys do and as it intersects with focus and we're seeing the community is people just agreeing. On the terms and terminologies and hopefully even pro against the providers between that I said against.

See that's Freudian between the providers because what I don't think anybody's differentiating things should be is like, are we best at finding the optimization opportunity, right? It should be like, are we the best at providing the service? So like we want to hopefully normalize a lot of that.

Corey: The bill is going to be what the bill is going to be, and by and large, the economic drivers, people are not picking a provider because it's 8% cheaper than another provider.

JR: Generally there. I hear some of that. Yeah. There

Corey: are exceptions.

JR: Everything. It's more commoditized gen, but Yes. And it's not about which. Yeah, exactly.

Corey: I mean, honestly, I went from being running my, uh, front end of my newsletter production system on AWS Amplify to ELL as a paying customer. And the single reason I know Ell.

I do, I over ell very well.

JR: I'm not nerdy enough to know. I

Corey: feel bit the, and the single reason I did it is simultaneously makes no sense in all the sense at once is that I got tired of fighting with my AI assisted developer in Claude Code to convince it that I was using Amplify. It kept assuming I was on Versel and I got tired of that sissy and push the Boulder back up the hill task every week.

JR: Okay. Okay.

Corey: Which is probably a silly reason, but I'm probably not the only one either. I, I just do what the robot tells me.

JR: Yeah. And Billing's always the last thing that people wanna pay attention to when they're building a product.

Corey: Oh. They pay a lot of attention to it. Right after they should have.

JR: And the last thing we looked at was how to build our own billing system, even though we were looking at other billing systems.

Right. That's, that's how it goes.

Corey: Ugh.

JR: So, uh, are you gonna come back to San Diego in June?

Corey: Oh, I'll be there. I'll be there at FedEx Ops X this year. I assume you will too. It's for you as sort of a command performance.

JR: I'm, I'm debating we'll see this year when I Yeah, sure. No, it's, it's, it's great. It's, it's our, uh, we, we kind of align our whole.

Calendar year around it. We don't, I, I informally call it our fiscal year 'cause it's like everything is leading up to X and everything's a before and after.

Corey: Unfortunately. That's how I feel about reinvent and I wish I didn't.

JR: Yeah.

Corey: Some year I'm going to skip reinvent. It'll be amazing.

JR: I've been promising that to myself and my family for like 12 years now.

And I, yeah, but it's, every time I go, we end up getting value. So it's, it's tough.

Corey: If, if people want to learn more or ideally and they should attend finops X, where should they go?

JR: There's a website. X ops work. Uh, but I think, you know, the thing that I always want to, for folks who've not been there before, well, ironically first, uh, the first reinvent when we were starting our conference, I went and watched the first keynote from the first Reinvent and Jassy sit on stage and he goes, this is not a sales and marketing conference.

That was in his opening keynote. And so I've actually adopted that phrase, but I mean it, I think a little bit more.

Corey: Well, they meant it at the time. They

JR: meant it. They meant it.

Corey: And then the economics sort of got away from it. And one on an earnings call during the pandemic, they mentioned it was a down quarter because they didn't have the boost of reinvent ticket sales.

It was Oh, my stars.

JR: But, but that's, that's what I like to say is, is, is a little bit special about our conference, which is, um. We're, we're not trying to get you to buy more cloud or anything else there. We're trying to get you together with your peers to like, not just get better at your job, but to be able to talk about the thing that, I mean, I mean, we just nerded out on some stuff Oh, yeah.

That most people could care less about.

Corey: This is what I love. Have this

JR: conversation. Right. Yeah.

Corey: During the raise, one of our hard problems talking to some investors was the fact that, their question was, why is this a hard problem? It just, the bill says what it says, and then you pay it. It's, yeah.

JR: Hasn't Amazon just solved this?

Corey: Exactly. It's, they're gonna solve it. It's an entire problem that so much the world doesn't know exists. And if you start talking about this at a cocktail party, you're asked to leave.

JR: Oh yeah. I hate it when people ask me what I do at cocktail parties.

Corey: Oh, I just lie. It's faster.

JR: Yeah, yeah, yeah. Okay, great.

Oh, that's, that's nice. But what do you do for a day job? No, no. I, I'm, nevermind.

Corey: We're, we're saving the cloud whales.

JR: You do? Yeah. Yeah. I really wanna say, like, and I, from bottom of my heart mean this. Like, I'm, I'm honored that, that you had me here and that you actually. Come and participate in our conference.

'cause I'll let you know, I, uh, at one point I thought of you as a, as a nemesis.

Corey: Mm-hmm.

JR: And you through our good friend Rick Oaks.

Corey: Oh, I love

JR: Rick came to me at reinvent and sat down and broke the ice. And, uh, if you haven't, have you heard the Infinite Game by Simon Sinek?

Corey: Not yet.

JR: So he talks in there about moving people from.

Nemesis to worthy rivals and a worthy rival is one who has a bunch of skills that you don't have who's probably in the same space, who makes you better because you watch them go and operate and you're like, gosh dang it. And I was like, Dak Corey's so good at getting the press and doing these things.

And you know, he, he's smart and he's funny, but I shifted my thinking around you and you've like, turned out to be like a lovely, sweet human being who, well, I like being, has a great public snark, you know, persona, uh, into that. And, and so I'm. I'm honestly like grateful that you've had me here and I'm, I'm looking forward to, you know, getting

Corey: to you better.

No, thank you again, where we stand in different spaces in the universe, like there is zero chance of me starting a nonprofit. I assure you that

JR: I would've said that seven, eight years ago too. Oh,

Corey: yeah. I, at some point when I, when my actions start casting aspersions on groups larger than me and causing problems for others, okay, this, I need to be a little more careful at how I say these things these days.

It's at some point of scale. I'm no longer a fit and I start transitioning over to the liability side. So it, it, it keeps things interesting, but I'm looking forward to seeing what's next. My, which is finops X in June.

JR: It is. And yeah, I'm excited to see you there and uh, hear what happens with Skyway and your everything else

Corey: more to come and watch this space.

JR: Alright.

Corey: Thank you so much for taking the time Jr. Appreciate it. Thanks. JR Storm executive director of the Ops Foundation. I'm Cloud economist Co Quinn, and this is Screaming in the Cloud. If you've enjoyed this podcast, please, we have a five star review on your podcast platform of choice, whereas if you've hated this podcast, please, we have a five star review on your podcast platform of choice along with an angry comment, but make sure it's focused, it's written in the right focus standard so that we can internalize it and maybe take action on it someday.