FinOps X 2026 | Karl Kraft, Walmart

Clips
More from FinOps X 2026

Karl Kraft

Sr Manager Software Engineering

Walmart

Why FinOps teams are turning to FOCUS for AI cost accountability

As artificial intelligence drives a new wave of token-based spend across cloud and on-premises environments, enterprises are discovering that understanding what they actually owe — and to whom — is harder than ever. The FinOps FOCUS specification is emerging as the open standard that provides practitioners with a shared data language to cut through multi-provider billing chaos.Fragmented cost and usage data across cloud providers have been the quiet tax on FinOps teams for years, forcing practitioners to build bespoke normalization pipelines before any analysis can begin. That pain is familiar ground for Karl Kraft (pictured), senior manager of software engineering at Walmart Inc

play_circle_outline Walmart's adoption of FOCUS to harmonize multi-cloud and on-prem billing data

play_circle_outline Effective cost column for prorating monthly subscription costs to daily/team usage

play_circle_outline FinOps Chargebacks and Incentives: Pod-Level Billing, Provider Split Allocation, FOCUS for All Orgs, and SQL Use Cases

play_circle_outline Harmonizing Cloud Billing: FOCUS 1.4–1.5 and SKU Pricing Dataset Standardization

Info
Transcript

Karl Kraft, Walmart

Karl Kraft

Sr Manager Software Engineering Walmart

In this interview from FinOps X 2026, Karl Kraft, senior manager of software engineering at Walmart, joins theCUBE's John Furrier and theCUBE Research's Paul Nashawaty to discuss how the FOCUS open billing standard is evolving into a foundation for AI spend accountability and enterprise-wide cost governance. Kraft traces his journey from manually reconciling incompatible Excel billing feeds across two cloud providers to championing FOCUS as the common language that eliminates that friction. He walks through a concrete example of how FOCUS's effective cost col... Read more

explore Keep Exploring

What led to involvement with FOCUS, and how does FOCUS help address the challenges of disparate cloud and on‑premises cost and usage data? add

How does FOCUS convert monthly billed (all-you-can-eat) costs into an effective daily cost and attribute that cost to the actual consumers? add

How does FOCUS convert billed or build costs into effective daily costs allocated to the actual consumers, and is FOCUS suitable for small organizations as well as large enterprises? add

How has the transition from FOCUS 1.4 toward 1.5 progressed from your perspective, how quickly has it moved, what needs do you have for that transition, and what might be included in the 1.5 release? add

bolt Powered by CUBE AI

Karl Kraft, Walmart

search

John Furrier

>> Welcome back to the Cube's live coverage here in San Diego for FinOps X2026. I'm John Furrier, host of The Cube with my cohost, Paul Nashawaty with the Cube Research covering this area like a blanket. We've been talking about the cost, understanding the cost of the FinOps, the token economics, but you got to know the data and FOCUS has been a real topic. 1.4 is out, focuses the FinOps, open cost, usage specification. Karl Kraft is here, senior manager of software engineering at Walmart. Very successful DevOps shop, very successful cloud native and I'm sure they got a lot of AI. Carl, thank you for coming on the Cube.

Karl Kraft

>> Yeah, thank you for having me.

John Furrier

>> So obviously cloud native has been a great run watching Kubernetes and all the developers using the cloud. But now with AI and tokens, token economics and AI factories, the data centers is in play. People are using AI, using tokens. Budgets are like popping. Values being created, but on premises and in the cloud. It's highly accelerated. So what's going on with FOCUS? How are you leaning into the FOCUS as a practitioner and a user? I guess customer, it's open source, but you're using it.

Karl Kraft

>> I think to really understand where I am with FOCUS, it helps to understand where I came from. So when I joined the FinOps practice, back before it was called FinOps, I was handed some Excel files of our cost and usage from our cloud service providers, two different cloud service providers. I looked at it and thought, "What am I going to do with this?" So we created a system. We started ingesting the files. We started using it, but we still were dealing with two fundamentally different pictures of our cost and usage data, as well as what we do on prem. So I heard about FOCUS at FinOps X four years ago when it was first kind of announced by UdeM and I was like, "That I want to be involved with. That is exciting." So all of the work that I've done to try and make sense out of the multiple different billing standards and put it into on cohesive picture for our internal teams, I don't want everybody to have to go through that pain again, and again, and again. With the number of new companies adopting FOCUS for their cost and usage data, it's becoming a lot, lot easier to onboard a new platform, to take something you thought you understood already and break it down into something a lot more granular so you can get the cost back to the teams who are actually consuming it.

John Furrier

>> Give an example. Give an example of where that new platform came in or new provider. People don't know, what's it like? Give an example.

Karl Kraft

>> So let's actually just take an existing billing feed that I had before and moving to FOCUS, right? So certain costs are billed once a month. You get an all you can eat for a month and guess what? At a company of our size, there are a lot of different people who are consuming and using it. How do you take this ... You've bought a bus pass for your family or for you and your roommates. How are you taking that and getting it back to the cost on a daily average, getting it back to who is actually consuming it? So with FOCUS, we have the effective cost. It's a new column. There's still the build cost. I can still see what I paid at each point in time, but now there is the effective cost. So I can see over the month my $30,000 is $1,000 a day, and who consumed it each day rather than just a big upfront cost that, great, I have this cost and I don't know what to do with it.

Paul Nashawaty

>> Yeah. So Carl, I love talking to Walmart as a company because you have lots of different use cases. You have lots of different things that are going on with regards to the scale of the number of employees you have, 75,000 plus configurations. Before we got on the air, we were talking about the Kubernetes, and the clusters, and such. You manage all of that. So organizations that are looking at this going, "Well, it's Walmart. Of course they have big systems like FOCUS to use." Is this something that small companies can use and they can get started? Or is it because you have multiple different kind of areas where you need to harmonize that data, is it easy to have that entry point to get started with FOCUS or is it like, do you have to be a very large organization?

Karl Kraft

>> I would say it makes sense for anybody. There are certainly a lot of, there's a use case library out there. I can go and I can take generic FOCUS data and grab SQL queries from the use case library to accomplish some of the easier FinOps tasks. So we talked about Kubernetes a little bit ago, right? So the 1.3 specification introduced the provider defined split cost allocation. It's a mouthful, but it's a way for providers to natively tell you how much the individual pods are costing within your virtual machines. Rather than just seeing the cost at a virtual machine level, you need to build this instrumentation, you need to go take an open source project and start using it. No, you can get it natively in your billing feed now and that's supported as part of FOCUS.

Paul Nashawaty

>> Yeah. That's big part of the chargeback model. Now you can do allocation. So instead of having these siloed systems and siloed processes, now you can have that harmonized with FOCUS. Is that what I'm hearing?

Karl Kraft

>> Yeah, absolutely. To me, chargebacks are all about aligning the right incentives and driving the accountability. So I want to make sure that I'm showing the cost back to the right teams. I'm not doing allocations that don't make sense for the business, because when an individual team optimizes, the entire business should see the benefit of it.

John Furrier

>> We talked to the FinOps Foundation about 1.4, kind of teased out 1.5, they're already kind of jazzed up for it. Take us through your view of the transition. How fast has it moved with FOCUS from your perspective and what do you need? What are your needs? They're watching.

Karl Kraft

>> It has come incredibly quickly. I've been involved in the FOCUS specification and helping write it since 0.5. You say 1.4, I'm like, wow, we're already on 1.4. Actually, as maintainers and contributors to the FOCUS specification, we've begun our dev cycle on 1.5. The introduction of additional data sets to really help paint a much broader picture has been incredibly helpful and we've got some of those that are scoped for 1.5 release already.

Paul Nashawaty

>> What does next year's 1.5 release look like?

Karl Kraft

>> I'm not committing to anything.

Paul Nashawaty

>> You can go in. Where would you think it would go?

Karl Kraft

>> So we are looking at a SKU pricing dataset because that's a common thing that we have from the different cloud service providers, but there isn't a standard approach to it. Even within one cloud service provider, you might have three different versions of the same data set that all communicate slightly different things. There's a lot to learn, and a lot to know, and a lot to keep in your head. So having this all well documented and standardized is incredibly useful for somebody new starting up.

John Furrier

>> I love your story. Your first example of the pain and suffering, I'll throw suffering in there. You just said pain. I'll put suffering in there. Necessity is the mother of all invention. What are the big first principles? How would you talk about the first principles of FOCUS? Because you start to see everyone seeing this. Common language makes sense. Open source is the perfect vehicle for that. People want to reign in the cost. They got to know things. Data sets are coming online faster. Is there a first principles or a philosophy you knocked down the biggest problems first? Is there an intentional roadmap? Talk about the first principles and then like that roadmap thinking.

Karl Kraft

>> It is an open community and anybody can get involved in it. That's what's so great is that we don't need to have these one person setting some foundational first principles that we all try to operate off of. Everybody's bringing their own perspectives. The more people you get bringing their perspectives, the better understanding we have of the problem. We can wrap your arms around it a lot better. So we're getting a lot of feedback from the community. FOCUS.finops.org. You can go there. You can submit feature requests. Talk about your pain points. We listen to the community and we take all of that feedback in. Many of us are practitioners. We have people from the cloud service provider side. We have vendor tools. Everybody's all coming together to make this specification and make it work, make it solve the problems that we're all facing.

Paul Nashawaty

>> You talk about the ecosystem, which is incredibly important, right? So let's talk about AI because we can't have a conversation without having AI in the conversation. So how has AI changed what you're doing at Walmart with regards to FOCUS, with regards to like you had your previous way of delivering, now you have you using FOCUS, now you're looking at like next generation of where you want to go with this. Is AI the pivot point of it? Is it making it faster for you? Are you using agents? How is this all happening?

Karl Kraft

>> So actually let me talk about the development focus and really how fundamentally that has shifted over the last six-month dev cycle that we've had. We have introduced guidelines for agents. Everybody's kind of been using AI a little bit, and you run it through your process, and you still submit it as yourself. Now with the guidelines, we're able to accelerate how quickly we're delivering the new features and it is a great second set, third set, or fourth set of eyes. We've all been staring at this for so long. It's like when you write a white paper and you read it and you read it through your own lens, having AI read all of it's like, "Actually you said this thing over here, you said this thing down here, you said this thing over here," and these don't all jive. So it's really helped us identify some of the subtle flaws that we've had. It's like, "Oh, yeah, we did say that thing, but we also said this other thing." We are able to improve the spec quite a bit by having something that can ingest all of it into its head. Somebody described to me that a problem that is too large to fit in one person's head needs to go on a whiteboard, right? We're onto multiple whiteboards at this point and it's, again, wonderful to have something that can fit it all into its head.

Paul Nashawaty

>> But I'll say that 2025 was the year of experimentation and innovation. AI comes out, there's a lot of ways of doing things, a lot of things that are happening and kind of coming together. According to our research, 25% of IT budgets were allocated to AI projects in 2025. Okay, now we're in '26. '26 is the year of implementation. People want ROI. They want to know when they're spending their dollars on AI, these AI initiatives, and what they're doing with regards to cost optimization. What does that mean to Walmart? What does that mean to the foundation? What does that mean to the ecosystem when it's no longer just like a sandbox kind of playing, it's actually expecting ROI for the spend that's being allocated?

Karl Kraft

>> I think it's important to understand what you're spending your money on first. So you need to be able to tie the entire story end to end to really get the value statement out of it. In 1.5, we have specific features we're working on around helping understand the AI spend a lot better. Breaking down what type of tokens, the what and why of it.

Paul Nashawaty

>> The Tokenomics, right?

Karl Kraft

>> Much more detail. This is such a large thing that ... There is the Tokenomics foundation. I have heard about this recently. I was like, wow, it absolutely makes sense. It's a different set of practices, but in many ways it reminds me of the journey that we had from on prem to cloud. You need to relearn how to do things because things are moving faster. You get more granular, and of course, the scale of spend is so large suddenly we need a new way of thinking about the problem.

John Furrier

>> Karl, final question for you. I'm curious. So you can answer it or take an attempt. It's more of an opinion. AI is all the rage and I hear terms like injecting intelligence into your organization. AI factories is intelligent edge. So when someone says intelligent FinOps, what does that mean? Are you injecting AI into FinOps? Is it being smarter with algorithms, data analysis? How should people think about bringing ... Because I'm sure people are going to say, "Throw some agents at the problem." I mean, you can strain agents down the road, but intelligence is coming. You can do more whiteboard space with Gen AI. What does intelligent FinOps mean?

Karl Kraft

>> So there is FinOps for AI and there's AI for FinOps. In order to really understand all of your cost effectively, FOCUS is a common nomenclature that's going to really help the agents excel and process the data.

John Furrier

>> Speak the same language, basically.

Karl Kraft

>> Exactly. Yeah.

John Furrier

>> Which is a source of truth.

Karl Kraft

>> Yeah.

John Furrier

>> Karl, thank you so much.

Paul Nashawaty

>> Thank you, Karl.

John Furrier

>> Great to have you on the Cube. Really appreciate it.

Karl Kraft

>> Yeah. Thanks for having me on.

John Furrier

>> All right. FinOps, speak the same language. Getting those costs understood is about speaking the same language. Harmonizing the data, getting access to the data, of course tracking all the operations. I'm John Furrier, Paul Nashawaty. Thanks for watching.