← The Directions on Microsoft Briefing Podcast
Loading summary
Transcript39 lines
- [00:00]
A
Foreign.
- [00:20]
B
Okay.
- [00:20]
A
Welcome to the inaugural 3 Smart Guys podcast. A little presumptuous there, but sponsored today by Directions on Microsoft, your authoritative source of information on all things Microsoft enterprise software, cost management and licensing. Today's topic is Microsoft's challenges for 2026. But since this is our first one, allow me to introduce ourselves. George Gilbert. Wave, wave, George, that's me, Peter o'. Kelly. Say hi and wave.
- [00:49]
C
Hi.
- [00:51]
A
I'm Barry Briggs. We've been friends for something like 30 years or so, ever since we worked together at Lotus Development, subsequently for Microsoft. And today I'd say we represent the three geographic poles of high technology. George is located in Silicon Valley. Peter is still in the Boston area, and I'm in Seattle, just a few miles away, as a matter of fact, from Microsoft headquarters. So, George, say a few words about yourself to introduce.
- [01:21]
B
We actually go back collectively more than
- [01:25]
A
35 years, so I'm trying not to sound that old.
- [01:32]
B
I first worked with Peter on Lotus DBMS Tools, which we actually never shipped, but that introduced me to client server computing, data management, then onto Lotus Notes. And I did a short stint at Microsoft before I went to software banking, and I've been an industry analyst after that.
- [01:58]
A
Peter, say a few words about you.
- [02:00]
C
Okay, I want to say at the outset, thanks for inviting me to join this discussion. I'm excited to be able to participate. For a brief summary of my background, I've been primarily focused on data management and collaboration software since the early 1980s. I'm currently freelance data modeler. At the outset, I should say my relationship with Microsoft is a little bit complicated. I worked for companies that were leading Microsoft competitors for about a dozen years, wrote about and consulted with Microsoft as an industry analyst for several years, and also worked in Microsoft consulting services for a couple of years. So I've been focused on Microsoft for a big part of my career, inside and outside.
- [02:42]
A
Yep, I'm Barry Briggs. I'm a former developer on Lotus 1, 2, 3, Lotus Fellow, and I'm what they call a serial CTO. And my last CTO gig was as CTO for Microsoft's own IT unit for about five years. And today I'm VP of Research for Directions on Microsoft. Let's dive in, gentlemen. At Davos this week, Satya Nadella, who I have claimed repeatedly is the philosopher king of high technology, and you've heard me say it more than once, he's continuing to maintain his optimism about the future of AI, and he said some interesting things at Davos. He said, first of all, that it's a great leveler because everyone has access to it, whether it's in the developed north or the Global South. And also, interestingly, I found this point particularly provocative that in the future, companies will competitively differentiate by simply adjusting their model weights. In other words, the weights in the models that your company's agents leverage and use represent your unique value. What do you guys think about that? And Peter, I'm going to call on you first because it's in our pre discussion you were somewhat more vociferous about Satya's talk at Davos.
- [04:07]
C
Okay, great. So here's my overall perspective having watched both the World Economic Forum interview and the latest all in podcast interview that was at the same location. So my high level take on this and we can come back to model weights in a moment. But if you are a believer in the most optimistic AI projections and concepts such as fungible infrastructure, token factories, infinite minds, presumably most of them non human minds, if that all collectively seems like the next step for humanity, you're probably very enthusiastic about the views that he's espousing. And it's clear that BlackRock CEO Larry Fink and the Microsoft alumnus and current Trump administration AI and cryptos are David Sacks among the interviewers there are in the superfan club. But I have to imagine that for many of Microsoft's enterprise IT focused customers, this is going to be a bit bewildering. So Nadella seems to implicitly consider the traditional Microsoft products somewhat tedious relative to the bright AI future he's trying to bring to fruition. And again, if I were enterprise Microsoft customer today, I would be concerned if it looks like Microsoft considers its current customer base to be useful, most as token factory funders and perhaps inadvertent data sources. So to be clear, I'm very impressed with the latest AI advances, especially by Anthropic and Google, and I think we'll talk more about this later in the discussion here. But I also suspect that a lot of people in the directions on Microsoft community are wondering if Nadella's sort of quasi utopian AI centric future is highly aligned with their customer concerns.
- [05:57]
A
Yeah, George, do you think Satya is like way out too far, in fact too far out in front of the customer base?
- [06:05]
B
I can see steps leading in his direction. I come at this from a slightly different angle. My take is that the age of procedural software is fading and that in the age of AI you program it with data and so like to I think we're going to see a sorting of vendors and software products themselves. There's Two categories I like, I conjure with it. It's going to sound like a weird analogy, but if you remember the last James Bond film, the scene where Raimi Malek is the villain is shooting at Leah Seydoux's character, the younger version of her. She's below the ice, and he's above the ice. I think the software industry is gonna sort into categories where above the ice is where you're providing products that help customers use their data to train models about how their business works. And below the ice is. Is anything in the infrastructure, typically procedural software that supports those platforms and tools. And I think the differentiated space is above the ice. And so just to put that in more concrete terms, in Microsoft's emerging product line, the Fabric IQ and the Work iq. Fabric IQ is the ontology builder on the fabric data platform. And Work IQ is the old Microsoft graph, I think, dressed up. Those are ways of, as Satya describes it, modeling the people, places and things in your world and then using that to train models. So I can see where he's going with his claim that in the future, you differentiate with, you know, the weights in your models, and I think they're trying to build products that help you train those models with the data about how your business works. So that was a long way of getting to saying, I kind of agree with where he's going.
- [08:26]
A
Yeah, I sort of feel the same way that we're sort of building these sedimentary layers of logic and, you know, in different forms. Back in the day, we built gates and then we built assembly language, and we're sort of working our way up into a whole new stack now. You know, which leads into my next question, which is, you know, I think, you know, we. We'd all agree that the world changed on November 30, 2022. I mean, that date is sort of etched in my mind forever. That's the date that, of course, ChatGPT was foisted upon the world. And Microsoft, really, for the first time in its history, really was a first mover in that space. They went all in on AI, and they've typically been a fast follower. If you think about GUI and the Internet and cloud and so forth. They've typically been following others, but they've tried to be a leader in the AI space. It's an immensely competitive space these days. George, do you think they can maintain their leadership and their momentum, or are they going to fall behind their competitors?
- [09:34]
B
Well, I think they made one huge, hugely consequential choice in 2022 when there were not that many GPUs around, like Nvidia's capacity to build them via TSMC wasn't that great and Microsoft didn't have that many. They had to choose of three sort of broad product lines which they were going to gen AI Enable first, it was Bing, Office or Azure. And I think they chose Bing first because they wanted to draw Google into the fight and make it a disruption to Google's business model, not the technology model. They knew they were never going to unseat Google as king of search, but I think they felt that if they could disrupt the business model that the $100 billion a year in after tax profits that Google was making on search, they could begin to squeeze that and Google then would not be able to subsidize quite as lavishly their Google Cloud platform and the workspace. In other words, it was a way of trying to squeeze the cash cow that Google had that Google was using to fund to go after Microsoft's real businesses, which was, you know, the productivity apps and Azure.
- [10:59]
A
When you say hugely consequential, I'm sort of tempted to hear the word mistake.
- [11:06]
B
Well, no, because I don't disagree with doing Bing first, but they did Office second as opposed to Azure. I don't know for sure, but I suspect it goes back to Gates's original dream of having Office be this programmable platform for building apps and that the last time that really worked was before the web came along when they had Visual Basic for applications. And I suspect he wanted, you know, this was a reincarnation of that dream. And in hindsight, the first killer app for gen was software development. And if they had put, put their GPU capacity on sort of remaking all the Azure tools. Not GitHub Copilot, not just GitHub Copilot, but every part of the software development lifecycle. And if they devoted their capacity to that, I think it would be a different company today.
- [12:05]
A
Peter, we've seen lately Gemini from Google and Opus 4.5 from Anthropic. These are pretty superior Chatbot and LLM tools. Is Microsoft falling behind, do you think, and can they catch up?
- [12:24]
B
Yeah.
- [12:24]
C
Well, let me come back to the context for your broader question there and then I'll come and I'll address the part about catching up. I do think then the bigger picture. Microsoft was shrewdly opportunistic in leveraging the split between Elon Musk and Sam Altman. That was early in the days of OpenAI. That led OpenAI to be very aggressive about partnering with Microsoft Because Elon Musk reneged on his funding commitment. And I think anybody who's a Microsoft shareholder should be very thankful for that because it's been a real boost for Microsoft's market cap over the years. I do think in terms of being a first mover, that they were a first mover among their direct and most strategic competitors. But I think OpenAI would argue that they were. And OpenAI, which is now increasingly competing with Microsoft, they would assert that they were the first mover in terms of where it goes from here. I think it's too soon, too soon, too soon to tell if Microsoft's going to be successful with leveraging AI beyond the infrastructure layer. There's no question that they're going to be a leader in that. They do seem to be placing some win, place and show bets because they're also in with a multibillion dollar investment and partnership with Anthropic because as you noted, Anthropic is seems to be moving swiftly ahead in AI for software engineering. But ultimately I think the aqua hiring of inflection and that brought the Google DeepMind co founder Mustafa Suleiman to Microsoft or he's now the CEO of Microsoft AI, I think that's going to be most pivotal and he has said lots of interesting things about whether or the extent to which Microsoft aspires to directly compete in the frontier models. So I think I'd say watch this space. Too early to tell.
- [14:14]
B
Sure, sure, sure.
- [14:16]
A
Both of you have been really vocal speaking of infrastructure versus applications, about the new applications that are emerging and that frankly are addictive to use, like Notebook LM and others that essentially seem to be disintermediating the knowledge Worker ui. And you know, that's a space that Microsoft has claimed ownership of and has owned for years and years and years. You know, Peter, you were an early and dedicated Notebook LM user. Do you think the Office UI is in jeopardy?
- [14:57]
C
I think it is in jeopardy of potentially being pushed back a little bit, like demoted to more of a supplier role for individual components, for instance. And this is something that we collectively have been talking about for a lot of years. But it's what is beyond the traditional model of productivity applications. When you consider word processing, for instance, was really starting as a print centric thing and taking advantage of PCs as virtual typewriters. There have been many attempts, including several by Microsoft to move beyond that to more of a dynamic, free flowing, distributed sort of gig jam or now copilot pages, if you will. And it's not that the traditional productivity Applications for their respective spaces go away. They're just not at the top of the user experience. They're more components. A topic we've talked a lot about over the years. So I do think that there's a risk here. I mean, there's also an opportunity. If Microsoft, of course is successful with Copilot pages and similar kinds of canvas oriented approaches, they may be able to add more value for their customers. But I would say at this point, with NotebookLM being a very fast evolving leading indicator, I think that Microsoft has some challenges trying to leapfrog that.
- [16:15]
A
Yeah, George, I want to ask you this, but I want to preface it by saying I did a presentation a few weeks ago and I had Gemini basically build it for me. And I'm curious if you think that that's a shot across Microsoft's bow, or if that's like a serious attempt to basically compartmentalize Office and just make it one among many.
- [16:43]
B
Well, there's so many angles on this one that we've talked about in the past. First, I would, I would point out that when you have a platform shift, the incumbent does have a grace period to respond. And that grace period depends on how deeply entrenched they are. And Barry, as you've pointed out, Office is not, or whatever we call it now, Microsoft 365. It's not just a collection of tools. It is a productivity platform and it's deeply embedded with it in terms of like, you know, information management and security. But I would say that this, this grace period is like in, in, in the first transition, I don't think the original, the traditional GUI goes away entirely. I think it gets complemented in the early part of the transition just the way like when the web first came out, for instance, you know, we put a gui, we put a browser version of the GUI on traditional, on prem apps. Later the apps were completely reinvented with a new delivery and operating model. We called it SaaS. But I don't think the original, the first transition with the productivity tools means they need to get replaced. I'm a huge fan of NotebookLM. I think it's awesome. But I do think Microsoft has quite a grace period. And I would point out that even if they started too early, like if the programmability of both the Office tools and the models they were using were both limited, I think part of why they tied up with Anthropic is because they, Anthropic has a whole new take on what an agent is where you don't just have a JSON call As a, you know, for a, for a tool invocation, you know, you have a programmatic invocation which gives you much more flexibility. You avoid the context window. The point is, I think, I think even if they've dug themselves in a hole in terms of technical debt, they can either get out of that hole and start over quickly or I think the tools that they're using now, if they start using Lycanthropic as the engine, I think they'll get a lot farther. It's just, you know, I suspect. But if they don't show something by the end of this year, I imagine they're going to start testing the patients of their customers.
- [19:29]
A
That's right. It's really interesting to think about the grace period and you know, how long that'll be and you know, given how entrenched Microsoft 365 is. One other aspect about that, and we talked about this, is that when we think about Microsoft 365, it's really in two parts, right? There's the applications like Word and Excel and PowerPoint and so forth. And there's a platform which is Purview for compliance and data governance and Defender for security and Entra ID for directory. And that's also something that's going to be really hard for other companies to displace. That said, I think that's a good news for Microsoft. On the other hand, Microsoft seems to be putting all the new features and all the new capabilities in the E5 tier of Microsoft 365, which is the most expensive one. And it's getting more expensive, I won't say every day, but regularly. Let's shift topics slightly and talk about the so called AI bubble for a second. One piece of evidence everybody points to, we've all seen it, is that there's just an incredible amount of capex being committed to data centers. And as you know, Microsoft spent something like $14 billion and probably more just for two built for purpose AI data centers in Atlanta and Wisconsin. And those are, at least from what they say, are specifically devoted to AI training and inference and in particular one client. OpenAI two questions here for you. Do you think we're really in a bubble? And if not, will these data centers have an roi? And when and what happens if they don't? Who wants to go first?
- [21:24]
B
I'll take a crack at this because as a, as a former securities analyst, I'm very sensitive to having lived through the bubble on Wall Street. As a securities analyst, I'm sure that
- [21:37]
A
was a pleasant experience.
- [21:39]
B
It actually Was it was a good time to be, as you remember, Saturday Live, you know, baseball been very good to me. Wall street was all right. On a more serious note, I think you start off with the demand side and there are three pillars of demand for Gen AI and the first one was the consumer agent, let's just call it ChatGPT is the prototypical one, but then Gemini and all the other follow ons and that one, because it's mostly free, there's a lot of demand. And the limitation there is on how much funding the vendors can get. OpenAI has to raise yet another round. They're supposedly seeking 100 billion, just 50 billion in the Gulf states, 10 billion from Amazon and who knows how much more. But the point is I would put that leg as the most unstable of the three. But you know, there's still, it's. The funding has not dried up and typically what dries up funding is actually tighter money and we're not in that environment yet. So I'm not worried about that. In other words, there's plenty of risk capital sloshing around. The second one is developers. There seems to be an infinite demand for tokens because we're just building these developer agents. First it was cursor on anthropic, but now like Claude Code, there's 30 million developers and we can add a lot of productivity to them, giving them God knows how many tokens and then that's just for the programmers. And there's a $300 billion software industry or more and that has to get remade. It's as if we were going to rewrite everything to go from hierarchical databases to relational databases. In other words, that Sequoia article about, you know, there's no revenue. They're missing the fact that if the entire software industry is being rewritten and then the one that, the one that where, where we probably will hit a wall soonest is the enterprise because they have to really remake the operating model, the technology model like as well. I'm sure at some point we'll talk about the data and how they, how they operate and the business model that goes with it. In other words, that is a heart and lung transplant to get this stuff adopted in the enterprise. So my take is there's plenty of demand in the near term, but the enterprise, which will ultimately really float the demand boat for a decade or two to come, that's going to be harder to come by. I think so, yeah.
- [24:54]
A
So the demand seems to be continuing to rise.
- [24:57]
B
Peter, what do you think?
- [24:58]
A
Do you think it's going to continue like this for a long time to come or is it going to trail off?
- [25:03]
C
Yeah, I have to say in many ways I think we're just getting started. And I realize that this coincides with a lot of people heading into the trough of disillusionment from the Gartner point of view saying like the hype on this has been overwhelming now for several years as you noted since things went big in late 2022 from a consumer experience. But I think there are several things that are just coming into place now. I would point within the last week, for instance to Google personal intelligence and the architecture that's behind that. It's not just extending Gemini with search access to Gmail. There's some really deep architecture inside there that I think is also going to be very useful for enterprise accounts. And of course there's looming in the horizon we'll hear about later this year. What's going to happen when Apple takes Gemini and brings that into the next generation of she who Must Not Be named. Of course all of our devices will go off if we say it. So I think that and like just yesterday Anthropic published its latest constitution for what it's doing and getting a lot of interesting attention as well. So I do think in many ways we're just getting started. I do think though like going back to your comment George about the not really fun for everybody bubble 1.0 for the Internet when that popped all of the fiber that was in the ground remained valuable even though it didn't have an immediate ROI that was hoped for at that point. But so I do think at an aggregate level we're like I said, just getting started and that the infrastructure investment is not going to be wasted money. But I do think we are going to start seeing a division between winners and losers and I think there are going to be some epic losers in this space. I think there will be like a Netscape 2.0 kind of trajectory for at least one player.
- [26:51]
B
I think we both have, we all have in our head an idea of who that might might be. Who's it most at risk?
- [26:58]
C
Yeah, lots of moving parts though. And you know again from from a directions on Microsoft perspective I would say the biggest question is going back to the the priorities aspect of it. So Microsoft is making this huge infrastructure investment. They are absolutely competing and complementing on the frontier models and competing very aggressively for the strategic partnerships in there. The real question for me is they will have a positive return on investment for the infrastructure build out. But again, how much is that going to change the corporate DNA and ultimately for its next 50 years, how closely is that going to be aligned to traditional Microsoft customer priorities? Right.
- [27:45]
A
I think you hit on something that changing the corporate DNA is something that, you know, it's going to take years, if not decades to happen. That said, my own perspective and I think of the three of us, I'm the most optimistic about it is that, is that it's going to happen. And I think we'll still be in a growth mode through 2026 for sure. And one of the reasons, I think that there are two reasons, one is I saw an interesting article by JPMorgan Chase actually, and they said that comparing the AI build out to the telecom build out, telecom build out was marked by excess capacity. And we have not seen that in AI yet. I mean, all the capacity that's being built is being used. So I think that's, I think that's a good leading indicator for now. And the other one I'd say is a lot more qualitative and I call it the amazement quotient, which is that when there's a new technological revolution,
- [28:44]
B
you
- [28:45]
A
start up a new application or run new service and you step back and go, holy cow, that's amazing. And I still see that from time to time with AI. You know, when I first used Notebook lm, I stepped back and was amazed. When I first used Gemini Pro to generate images and generate my presentation, I was amazed suno for doing music and so on. So I don't think we're out of the age of amazement yet. And AQ is the amazement quotient still pretty high? Guys, it's been a real pleasure chatting with you. Thank you. Thank you. Thank you so much. Thanks to our sponsor, Directions on Microsoft. If you need a reliable information source about anything Microsoft Enterprise, whether software, services, costs, licensing, come visit us at directions on Microsoft.com and if you like this inaugural podcast from the three of us, if you have suggestions or topics you'd like us to cover, our next one is going to be about data. We touched briefly on that and all the problems associated with cleaning up the enterprise data state and getting it ready for AI. But if you have suggestions too, drop us a line in the comments. We'd love to hear from you. Thank you very much.