Loading summary
Transcript20 lines
- [00:20]
A
Lately, here at Directions, Azure has been figuring into our conversations in a variety of ways. With Microsoft forecasting that more and more of its pricing and licensing soon will revolve around consumption, the concept of metered Azure services and how they affect customers bills has been top of mind. With so much pressure on Microsoft and other hyperscalers to grow their data center footprints for AI and other workloads, the idea of cloud compute constraints has been big around here, too. Two themes are better than one, right? We're going to talk about both of these hot Azure topics in today's podcast. Welcome to the Directions on Microsoft Briefing podcast. I'm Mary Jo Foley, the editor in Chief here at Directions. I'm your host for the series of podcasts for those interested in the Microsoft Enterprise IT ecosystem. My guest today is Lane Shelton. Lane is the Director of Product Development and Advisory Services lead here at Directions. Hi Lane. Thank you so much for joining me today on the podcast.
- [01:23]
B
Thanks Mary Jo. Glad to be here.
- [01:25]
A
All right, let's start with the whole cloud constraint idea first. So as crazy as this might sound or have sounded a couple years ago, the cloud is not limitless. Microsoft, Amazon, Google and others actually can and have run out of room in the cloud. I actually wrote about this back in 2019 when Microsoft was hitting capacity limits with some of its Azure resources resources in the East U.S. 2 region. And I wrote about it again in 2022 when supply chain constraints were wreaking havoc with some customers Azure plants lately with AI demands resulting in astronomical capex spending by Microsoft and other cloud companies. We're talking again about capacity constraints. So Microsoft's admitted they're trying to juggle their own first party needs with those of their big customers, including companies like OpenAI on. And then when you add supply chain issues to the mix, the cloud vendors can't even source enough of the parts and the power they need for these data centers. So Lane, with all this going on, what's the customer to do here?
- [02:37]
B
Well, Microsoft's capex problem is sort of their problem until it becomes your Azure service delivery problem. And the challenge is that enterprise customers get deprioritized in ways that very rarely show up in your governing contract language. And that's the real challenge. If you think about how most enterprise big customers are consuming Azure today, it's via a Mac, right? The Big Mac, the Microsoft Azure commit to consume macc, and that's basically guaranteed spend in the Azure ecosystem. But now the yardstick that guaranteed spend is measured against infrastructure that Microsoft themselves are struggling to maintain and if you think about how many Macs were signed before this really sort of arrived on the scene, you know, in the, in the massive way that it has now, that's going to be a challenge for, you know, for a lot of customers. So first of all, if you're signing a Mac or you're negotiating one, or you have one coming up, first and foremost, you need to dust that thing off and start looking at what are the provisions in there that address that, because there probably aren't very many. Right. You need to look at this carefully and know what you're signing and be way more skeptical. When you're planning your capacity, when you're planning workloads, especially AI workloads that could become mission critical, you really need to start thinking about how to bake this risk into your Azure Mac or whatever. If you're still in an EA or an SCE or something, whatever your governing contract vehicle is, you need be thinking about how to bake that risk in. But this is also an argument for multi cloud or even hybrid cloud wherever possible. You've got to hedge your bets in some ways.
- [04:18]
A
Yeah, definitely, yeah. Because sometimes I've heard Microsoft is advising customers who cannot get capacity in the region or data center closest to them to just take their workloads to a less populated data center. Okay, that sounds good. But then there are all these new issues like latency, data residency, and other things. So, I mean, do customers have any recourse if Microsoft just says, well, we're out of room here in usc, so go, go, go somewhere else? Right.
- [04:46]
B
Again, you know, most, most of the people that are going to be impacted by this sign their contracts already. You know what I'm saying? So. Yeah, and you're right. You know, I think data residency is sort of consideration number one because that, that has real teeth to it. But latency, you know, that's like the site that's like high blood pressure, right? That's like the high blood pressure of Azure Compute. Because, you know, it's the silent killer. You build, you build something that's mission critical and the latency factor piles up and impacts things in subtle ways that add up to, you know, material cost problems and design problems. So, you know, again, you got to go look at what your language says right now. And first and foremost, if you're running into the issue, document it right in writing and make sure you're escalating that to your Microsoft account team every time you got to build that paper trailer. But also, you know, wherever you are in your Mac cycle, or your contract cycle. It's time to, again, I'm going to repeat this a bunch of times here. It's time to get that thing out and look at it very carefully with fresh eyes through the lens of this problem to see what language is in there. And if the language is inadequate, then you need to talk to Microsoft about it. Especially if you're getting ready to renew or you're planning for the renewal. Now's the time to. Now's the time to get the lawyers and get together and figure out how to build this into your agreement.
- [06:04]
A
I know you and I have been talking and you said you and some of the analysts here have been talking about what would happen if a customer buys a certain amount of Azure capacity, pays for it, and then they can't find it when they need it. Right. And it's just not there. So besides looking at the Mac, is there anything else customers should be doing to protect themselves?
- [06:26]
B
Yeah. Well, first of all, again, you have to build the muscle of factoring this new level of risk into your, all the decision making points, right? When you're building, when you're planning something, thinking about something, ideating something, prototyping, something like all this, like this needs to be a factor in all of that. And that takes organizational muscle, Right? Big organizations take time to build muscle. It takes a while, right? So it's the time to start, that is right now. But also document, you know, document everything. So you want to sort of look at your technical roadmap and figure out where the fault lines are there because of this problem and get ahead of those fault lines, both contractually and in how you design things. But careful documentation of every time this affects you. I think back to a time. There was a time back where I'm going to tell a quick story here. There's a time when Office365 first came out and there was one huge customer out there that at the time nobody wanted to sign up for O365, E3 for everybody because it was so expensive. So the goal was to have some E1 users and some E3 users. But there were some big feature differences between E1 and E3. And so there was this huge retail company out there that signed up for a whole bunch of E1 and they minimized E3. And then they spent three years sort of struggling with that. But they wrote it all down like every time they ran into a problem where, oh, you know, this department turns out they had some Excel process that could only work in online Excel. And if they didn't have that process there, the whole business would grind to a halt. So they had to convert them back to E3s. They kept this paper trail and that paper trail was so educational for so many other customers. You know, in terms of like, how do you, how do you deal with this? So they kind of by documenting it and by, you know, not only were they able to get recourse from Microsoft in time, but that also gave, you know, that also created some real peer to peer value for, you know, in the, in the industry that helped a lot of customers. So I think we're, we're all in this together, right? So the more you document this, the more you talk to Microsoft about it, talk to your lsp, whoever, you know, know, get the word out so that it spreads and people can sort of start to evolve best practices for dealing with this new constraint.
- [08:41]
A
Okay, that's good, good, good advice. Let's take a quick break so I can talk about a secret that we've been keeping here for a while. We are ready to start rolling out our Atlas AI Assistant, which will be available to a few select users on June 1 and roll out to all directions on Microsoft members only starting July 1st. We may be biased, but we think this is going to be the AI companion of your Microsoft licensing and technology dreams. We built it to make it simpler for you to find the information you need based on the insights of our team of experts which they've developed and curated over the past several years. If you'd like to give the Atlas AI Assistant a whirl, sign up to be a member today so that you can check it out. Go to directionsonmicrosoft.com Become a member to join. All right, back to the Azure fast lane with Lane Shelton. Now we're going to switch gears. We're going to talk about a blog post that you wrote for us recently about metered Azure services. I know you've done a lot of work to uncover some of the growing number of Microsoft 365 services that are actually being billed under Azure. So give us a quick summary of what you found out.
- [10:05]
B
I found out that there's like 10,876 Azure meters that spin against 1065 services. There's a lot of them. There's an awful lot of them. You know, a lot of them are esoteric and variations of a theme. But the point is, you know, communication services, AI data platform, like a lot of these things, there's Azure meters that once you hit a certain condition in your user based licensing it triggers an Azure meter that starts spinning. That is something that's rarely accounted for in either side. In most big organizations. Azure is over here and the EA is over here. Those worlds have evolved. They've evolved in parallel, but generally on separate tracks. What's happened is quietly, those tracks have merged in the background, but they haven't merged in the foreground. So what ends up happening is there's a growing amount of cost for your M365 services that is present in your Azure agreement and hidden there. And that's the new challenge that we're starting to think about.
- [11:19]
A
Okay, so say a customer realizes they've they're seeing more and more Microsoft 365 services built through Azure. Is there a way they can get better control of the situation? Does something like Azure Cost Management help in any way here? What are you suggesting they do?
- [11:38]
B
Well, okay, I consider myself an Azure Pro amateur. I know just enough to be dangerous and break everything. But I do know that I like things simple. I'm a business user. The other thing about Azure is companies tend to put their best and brightest on their Azure ecosystem, and it doesn't always make it easy for the business owners of Azure to really understand what's going on. When there's 18,000 meters all spinning at different rates, it can be technically very challenging. However, Azure Cost Manager actually has some basic functionality. Like if you go in and make sure you're scoping it at the top and create a report that summarizes by Azure service name or Azure Meter. I forget one of those two. Give you a monthly read of the last 12 months by Azure Service. You can get a very quick and dirty Excel spreadsheet that gives a business, that a business user can read. If you even have like, you know, a fifth grade level of Excel understanding, you can, you can parse this spreadsheet and look and see, oh, this looks like, you know, oh, what are these Purview meters. Oh, you know, what are these, you know what, what's going on with the relationship between what I'm doing with Copilot and AI Foundry or things like that. So there are some basic things in Azure Cost Manager, some basic reports that you can get. I haven't found any tool vendors out there and I talked to quite a few of these companies that have tools that go in and pull data and analyze data to optimize your Microsoft investment. I haven't seen one. I'm not saying they're not out there, but I haven't seen one that specifically addresses this. But I've been talking to those companies about this saying, hey, you know what would be really cool? If you did this, just hopefully. And I imagine that at some point, if we keep talking about this long enough, somebody's going to figure it out and build a dashboard that says, okay, you know, here's your M365 in Azure Dashboard. If I had the technical skills to do it, I would.
- [13:32]
A
Maybe you could vibe code it, you know, literally.
- [13:35]
B
Literally, as you're saying that, I'm like, okay, I'm going to go vibe. So remember what I said? I told other tool for I know that's now my intellectual property. Don't go build that because I'm going to go buy code it with Claude.
- [13:46]
A
All right. In your blog post, you talked about the idea of illuminating the Azure blind spot, which I loved. I love that phrase. What can customers do to actually make that happen before their next renewal?
- [14:00]
B
Run that report and subscribe to my new application that I'm going to build apparently very shortly. But that's it. You've got to be aware of it and look at it and understand it and think about like, you know, it's probably not a ton of spend yet, right? This isn't. And that's why it's important to do it now while it's small. Now, obviously with some customers, it could get very big very quickly. Especially like if they're heavy into Purview and starting to really expand their usage of Purview beyond the borders of M365. Those meters are going to start, you know, running up real quick. Same thing with Copilot. Get heavy into Copilot Copilot Studio meters are going to start, you know, are going to start rolling very quickly. Start ingesting a whole bunch of data into Sentinel and your log analytics bill is going to explode unexpectedly because you're ingesting a ton of data. So while it's small, it's good to be aware of this. And I would say start the conversation like, where do you start? Start with a simple report that you can pull out of Azure Cost Manager. Anybody can pull that report that has the, you know, has the right credentials. Start looking at the data and start having the conversation. Especially if you've got that sort of bifurcation of function where you've got your EA team over here and your Azure team over here. Start that, those lines of communication. Because a simple report like that can give you the way to ask, you know, ask questions to the Azure geniuses that have PhDs in azurology. You can, you know, you, you'll have something that you can go to the table and say, hey, what is this? How do I address this? How do I, you know what, you know, what can you tell me about, about this, you know, spending trend? Is this something that's increasing, et cetera? You can have those conversations. So the best part, what is it that they say? Performance measured is performance improved. So start by measuring it.
- [15:39]
A
Okay. Anything else you would like to add for customers, either about the whole capacity constraint issue or the metering, anything that we didn't cover that you think they should definitely pay attention to?
- [15:52]
B
And know, I think just that we're at an inflection point, right, like, and right at the beginning of it, right? It's just starting to really take a bite. And you know how it is when you make a big plan and once you're into the plan, you don't look at the plan anymore because you're executing the plan. Everything's going, you know, you're going, you're going 100 miles an hour. There's always that time where you want to, like, put the brakes, step back, take a look at this, you know, from a, from a, from, from a fresh lens. This is probably a good time to do that with your Azure consumption, especially if you've got a Mac renewal coming. You know, even if It's a year, 18 months off, like I said, it's a good time to start kind of reassessing where you are. Because think about it, you sign that commitment, you say you're going to spend, you know, X number of dollars if the infrastructure that, that you planned on being there isn't going to be there, then that's a challenge, right? Like that, that, you know, okay, next time you negotiate that Mac number, you need to build that risk into that Mac number, because what happened? Because Microsoft's not going to let you off the hook if you don't hit that number, you know, or will they? Like, how do you, how do you account for this? This is, we're all working through this right now. We're all developing best practices. So, you know, think about it, talk to your peers about it, post on Reddit, whatever, you know, but, but let's, let's all work through this new problem together so that, you know, because you know, nothing, nothing is worse than you have that beautiful plan, you're building that beautiful workload, whatever, and you hit some stupid constraint like this, I'll close like this. I think there's a historical parallel here. And I'm going to Say, feel free to shoot me down in flames on this. It's fine. But this is my personal opinion. So this is not represent, Donald. But remember back when the DOT bomb happened, right back in, at the turn of the century, like there was all this E commerce stuff and all this amazing things that were happening on the Internet and then all crashed and burned in 2000, around 2001. Why? My opinion, the network wasn't there. Right. The broadband Internet in every household just wasn't strong enough to support that much transition to the Web. But if you go, you know, fast forward 10 years and all the stuff from 1999 actually happened pretty much the way that we wanted it to. It just the bandwidth wasn't there at first. Right. I kind of feel like we're in that same exact spot right now. Right now we have power constraints, we have things that are bottlenecking. And what's going to happen? Is this bubble going to burst? Even if it does, fast forward a couple of years, we're going to be. I think there's a historical parallel there. So right now is a good time to really reassess what you're doing, how that's going to survive and scale in the long term. Because I don't think these bandwidth constraints are going to be getting any better until somebody shocks us all by inventing cold fusion or they put a data center on the moon or some crazy thing happens that goes. Or quantum computing across multiple dimensions, who knows? But something's got to remove this bottleneck the same way broadband Internet removed the bottleneck of the previous boom. So just my opinion, but I think there's a parallel here. Always look at history because there's guidance for the. For the present in the past.
- [18:59]
A
Yeah, that's true. All right, Lane, well, thank you as always for all the great information that you shared today. I would like to remind our listeners they can find lots more coverage of all things Microsoft related on directions on Microsoft.com thank you so much for listening. If you have questions, comments, or any topics that you would like to hear the directions analysts cover in one of these podcasts webcasts, please do not hesitate to contact me via X or BlueSky. Directions on Microsoft also is on LinkedIn, so make sure you follow us there and give us a follow at DirectionsMSFT on X or directions on Microsoft on BlueSky for all of the latest Microsoft Enterprise product and licensing information. Thanks again,
- [19:50]
B
Sam.