Ask Paul: May 1 ⭐️

Happy Friday! It’s Labor Day here in Mexico, but there’s no rest for the wicked, so I’ve pushed myself off an accounting and financial reporting cliff, and the bill has come due. You’re welcome to join me for the ride.

“Jim Cramer says the market’s biggest winners all have one thing in common”

They’re not Microsoft

🪟 There can be only three, or maybe four

helix2301 asks:

Paul, the $13.2 billion Microsoft pulled in with more personal computing is all Windows operating system, Xbox, Surface devices, and Bing search advertising. If you were to guess how much of that 13.2 billion percent wise is Windows, Xbox, Surface, or Bing advertising revenue. Do you think Windows is still the biggest part? I can’t picture Surface or Xbox being a huge part. I do know a lot of businesses using Bing maps API because cheaper than Google maps, but even that can’t be a billion-dollar business.

The lack of transparency in Microsoft earnings makes all this difficult to assess. But here’s what I know.

Back when Microsoft was still fighting antitrust concerns about its proposed acquisition of Activision Blizzard in mid-2023, I wrote an article describing how this would impact the financials. This was 2023, so the most recent data at the time was the first calendar quarter of that year and then the three previous quarters. Basically, Activision Blizzard was averaging a bit over $2 billion in revenues a quarter. So for a company the size of Microsoft, with over $207 billion in revenues over those same four months, the change was minimal, or a bit under $216 billion total. But for More Personal Computing, which was averaging a bit under $14 billion a quarter, or $55 billion or so for that time period, the change was much more significant (by percentage), with the two combined delivering about $63 billion.

Tied to the question below, Microsoft changed how it accounts for revenues from Windows in August 2024, which is after the overview noted above and after the company successfully acquired Activision Blizzard. That’s notable because most Windows-related revenues come from enterprise licensing, not PC sales, and with the accounting change, revenues that Microsoft used to attribute to Windows Commercial products and cloud services are no longer part of More Personal Computing. And so the basic math we might have used in 2023 to describe the various parts of More Personal Computing, which were likely out of date at that time, are now completely different. In short, we have to guess.

(Tied to that, too, the account change combined Surface into Windows in what is now reported as Windows OEM and Devices. Surface revenues could skew this.)

The long-time assumption, because this was once known to be fact, is that about two-thirds of Windows revenues came from businesses. My guess is that it’s closer to three-quarters now, but let’s not muddle this with too many variables. Pre-Activision Blizzard, I would have argued that two-thirds of More Personal Computing revenues could be attributed to Windows, with most of the remainder coming from Xbox/Microsoft Gaming and then small amounts from Surface and Bing/advertising.

But with Windows commercial revenues now being attributed to Productivity and Business Processes (which houses Microsoft 365), that means that the Windows revenues in More Personal Computing have fallen by what I will guess is roughly two-thirds. If you look at this business in the current quarter, there is about $13 billion in revenues to account for. Pre accounting change, I might have attributed two-thirds of that, or $8.6-ish billion, to Windows. But Windows revenues are now smaller, one-third smaller by my guess, or about $3 billion. So the remainder would come from Xbox/Microsoft Gaming and whatever else.

But that would mean that Xbox/Microsoft Gaming made at least $6.5 billion-ish dollars, if one assumes—there are many guesses there—that it is somehow earning two-thirds of the remaining revenues. Is that possible? Aside from the obvious, like Xbox/Microsoft Gaming revenues falling this past quarter, I feel like it is. In that 2022/2023 accounting, More Personal Computing was averaging about $16 billion a quarter. Slice of two-thirds of that, and you get $10.6 billion. And two-thirds of that is $7 billion. So it does add up.

It’s important that I communicate again how many guesses are in this. But just eyeballing this using out-of-date math and understandings of businesses that are now in no way transparent whatsoever, I arrive at these numbers for the quarter:

• Windows and Devices – $3 billion to $4 billion in revenues
• Xbox/Microsoft Gaming – $6.5 to $7 billion in revenues
• Search advertising – $3 billion to $4 billion in revenues

That latter number feels high to me, but it’s what’s left over. And while the Windows number likewise feels small, it doesn’t including Windows commercial revenues. So the total is likely still just south of $10 billion there. But this also doesn’t account for Surface. It’s a small business in many ways, but it’s also hardware and each device is expensive. So there may be $1 billion or more that needs to skew back into Windows and Devices. But this is good enough for horseshoes and hand grenades.

To actually answer your question, I believe that Xbox/Microsoft Gaming is now the biggest part of More Personal Computing by revenues. That said, it’s not profitable, and Windows is hugely profitable. I have no particular feel for Search advertising profitability, but I do think the issues in Xbox/Microsoft Gaming are serious enough that More Personal Computing is either unprofitable or close to it. The addition of Activision Blizzard revenues (and profits; that business was always profitable) isn’t enough to overcome the loss of Windows commercial revenues.

Almost anyone with accounting knowledge could easily throw any number of wrenches into these guesses, I’m sure. But that’s where my head is at, and I would love to see hard numbers on all this. That’s not the Microsoft Way.

“Key takeaways from Musk’s testimony at OpenAI trial”

Most obvioulsy, he’s a jackass

🔢 Accounting for complexity

OldITPro2000 asks:

Do you expect Microsoft to change their financial reporting to lump all of Windows together? Right now they split three ways between Windows Server, Windows via 365 licensing (so Enterprise/Education), and Windows OEM. With the OEM numbers projected to be poor the next quarter I wonder if combining those lines is in the back of someone’s brain over there.

As noted above, Microsoft changed how it reports Windows revenues in late 2024, with Windows commercial revenues going into Productivity and Business Processes, the home of Microsoft 365. This makes sense, since Windows commercial is a Microsoft 365 offering now, and it’s sold and serviced according to completely different terms from the consumer/individual purchases that typically occur with a new PC purchase. However, I suspect that Microsoft only felt comfortable making this change after acquiring Activision Blizzard, as those additional revenues would help offset the loss in revenues from Windows commercial. Also, it helps fuzz up the math since if it had just removed Windows Commercial without adding Activision Blizzard, we’d have a better idea of how many revenues at least Windows Commercial makes/made. We can’t have that. (The addition of AI revenues in both Productivity and Businesses Process and Intelligent Cloud helps further mask where money is made and lost, of course.)

This feels cynical to me, but it’s important to view every accounting change at Microsoft in terms of them hiding the truth. And that’s really all this is. When Microsoft announced the accounting change, I wrote that More Personal Computing revenues would be “significantly lower year-over-year (YOY) moving forward.” And that’s what we’ve seen and, to your point, what we will see even more clearly over the next few quarters. In the post-earnings conference call, Microsoft CFO Amy Hood said that More Personal Computing revenues for the current quarter would be roughly $12 billion, compared to $38 billion for Intelligent Cloud and $37 billion for Productivity and Business Processes.

Think about that for a moment.

There was a long period of time when these three business units were averaging roughly $11 billion a quarter, and they were all very even. But things have skewed towards the two largely cloud- (and now AI-) based business units since then, and the delta is growing dramatically. In the same quarter five years ago, Microsoft earned $41.7 billion in revenues, with three business units delivering similar revenues/percentages of the total: More Personal Computing was at $13 billion (31 percent), Intelligent Cloud was at $15.1 billion (36 percent), and Productivity and Business Processes was at $13.6 billion (33 percent). They were close.

Now look at how much that will change in the current quarter if Hood’s estimates are correct. Total revenues will be about $86 billion. More Personal Computing will have $11 billion of that (13 percent), Intelligent Cloud will be $38 billion (44 percent), and Productivity and Business Processes will be $37 billion (43 percent). The cloud-based businesses are going up by percentage, and More Personal Computing is going way down, with extra harm courtesy of the component shortages and, hopefully, a temporary condition. But it will always be down going forward compared to the other two. Unless they change the accounting again.

Will they?

I don’t believe so. The accounting change they made in 2024 still makes sense. Long-term, I’m sure the hope is that Windows and Xbox can both rebound, and if Microsoft can ever get over its money-losing hardware businesses, the entire unit would be hugely profitable, as Activision Blizzard usually was. This isn’t predicated on Microsoft dropping Surface and/or Xbox console hardware, but that would put it over the top. The former is more likely than the latter, of course, especially short-term. Microsoft has already committed itself to two more years, almost, of pre-release Xbox console buzz and then we’ll see what happens.

That said, further accounting changes will make sense someday, and it’s not difficult to imagine Microsoft reorganizing the business because of AI and all the infrastructure investments. Today, this is mostly about hiding all the losses inside its two highly profitable business units, a sort of “fake it until you make it” strategy that can’t persist indefinitely. I always felt having three top-level business units was overly simplistic for a company of Microsoft’s complexity, but that’s by design to obfuscate the numbers. And I’m sure Microsoft would argue that because AI is everywhere in the stack, making a standalone AI business unit doesn’t make sense.

What would make sense to me is a differentiation between infrastructure that’s used for third parties, first-party services, first-party software, and, because they can’t seem to rid themselves of this for now, first-party hardware. Or whatever, the important bit here is figuring out a profit/loss for infrastructure that serves customers and separating that from anything first-party.

Whatever happens, Microsoft is a completely different company than it was even five or ten years ago, and those changes will continue. What used to be mostly a high-margin software/services company is transitioning into what I feel is a bigger but lower-margin business of incredible complexity that includes real estate, leases, hardware in datacenters, and whatever else in addition to the things we still think of as Microsoft. In time, the cost and value of those new things will outweigh the cost and value of the first-party software and services. At which point Microsoft becomes IBM, essentially, albeit much, much bigger. It’s not a terrible outcome. Unless you’re the technology enthusiast who’s here for the stuff that Microsoft, Wall Street, and the world no longer care about.

“Spotify wants to help listeners better identify AI-generated music”

And not a tool to just eliminate AI-generated music? Curious

💩 They all do it

S100 asks:

You wrote previously about the problems with Lenovo’s consumer Vantage software in terms of ads/upsells etc. How is the equivalent software on HP, Dell or other systems you have tested on the consumer side? I assume business systems are better in this regard?

Yes.

All the major (and minor) PC makers have these types of utilities. On HP it’s called the HP app (previously MyHP). And they all have their own stupid AI chatbots and whatever other nonsense. This is a decades-old problem that’s tied to two issues specific to the PC market, its low margins and the misplaced desire on that part of those at PC makers who want to also add value and not just be the company that ships a device and then never interacts with the customer again unless there’s a problem.

In short, this is the dark side to the obvious benefits of having competition and more choice, which leads to endless sales and these companies driving to undercut each other. The alternative has historically been monoculture in the form of Apple and the Mac, and the higher prices and whatever limitations one might see or perceive there. And then Linux as a fun little wrench in everyone’s plans, or alternatives like Chromebooks and mobile platforms now.

To be fair to PC makers, it’s a tough business. Microsoft sells them a Windows license that they can include on their PCs, but then it washes its hands of any issues the end user may have, requiring the PC makers to provide support. The hardware sales are already incredibly low margin, with minor exceptions for premium solutions like PCs for gamers or creators. And a single support incident could eradicate any profit on a PC. So they have tried various means over the years to increase profitability through crapware (which often involves a per-PC payment from the software makers), value-add subscription services often tied to some form of support, and whatever else. We can debate individual bundled apps and upsells, of course, but at a high level, PC makers are just trying to be profitable. To us as consumers, some of the tactics are not ideal.

So it’s no coincidence that the PC maker’s alternative to switching to a Mac is to just spend more money on a premium PC of whatever kind. The experience is almost always better. Premium consumer PCs will still come with whatever crapware and upsells, of course, but it’s usually not as bad as that seen on the cheapest PCs. And business-class PCs are almost always devoid of crapware. Lenovo is explicit about this with ThinkPad, but not with other product lines. And while HP bundles its own Wolf security solutions on its business PCs, it at least doesn’t dump McAfee on customers. Any individual could buy a ThinkPad or an HP EliteBook, or whatever business computer, and have a reasonably or truly clean experience. It will just cost more.

Leaving aside the MacBook Neo, which is really aimed at people who do not use a computer for anything taxing or don’t need one every day, I’ve long felt that the MacBook Air was a great value at $1000 and up (maybe $1100 and up these days). And that equivalently priced PCs are likewise usually in the same class from the perspective of being clean-ish and with little or no upsells. This is the true cost of a thing coming through to the consumer. At this level, PC makers can afford to behave more like Apple.

With the MacBook Neo, Apple has sort of flipped the script, but if you’re someone who actually uses a computer every day, that thing would need to be upgraded more frequently. You’d be better off with a MacBook Air, but Apple benefits either way. One MacBook Air every 5 years or one Neo every two or three is all the same to it. Plus, Apple does have upsells now. Witness the iCloud storage plans, Apple One plans, and all the other hardware you’ll want to buy in that ecosystem. Cha-ching. You pay one way or the other. What price one’s soul?

“Outgrown Your Starter Record Player? Here’s How to Shop for a Quality Turntable”

It’s called a Compact Disc player, you’ll love it

🔮 Reading the room

spacecamel asks:

Your article on Ubuntu planning to put AI into their OS made me compare it to the messaging that Microsoft did not do. Ubuntu seemed to focus alot more on that these are just features of various programs in the OS and the fact that you will be able to opt into them if you want them. Considering the linux user crowd tends to be more anti-AI that the general population, I think the message went over better than I expected and is probably what Microsoft should have targeted in the first place. Do you think we will see better messaging on this in the future?

I’m still quite taken by this announcement, and I thought enough of it that I mentioned it to my wife, who doesn’t care about technology but does use AI regularly and much more than I do. Ubuntu Desktop is a lot more complex than say, the Firefox web browser, but at a high level, their respective approaches to adding AI to the products are similar. Both will offer a kill switch of sorts, though in the Ubuntu case it’s more of a series of kill switches because of that complexity.

But both also correctly acknowledge that some AI functionality–language translation in Firefox’s case or speech-to-text/text-to-speech for Ubuntu–is too obviously useful to let it get caught up in some anti-AI rhetoric. So the way they’re handling this, in each case, seems good to me. Those that want AI will get it, those that don’t can turn it off or never install it, and those who really want to fine-tune which AI features they use and which they don’t can do that.

That said, when I read the Ubuntu announcement, I had the same feeling I had when Microsoft announced Recall at that Copilot+ PC event two years ago: They don’t understand the trust-based can of worms they just opened. In Microsoft’s case, it’s like they have no idea how untrustworthy they are to so many customers, and not fully explaining all the work they had done to secure Recall was a huge mistake. The Ubuntu announcement was much more self-aware, but the Linux audience, as you point out, is much more aggressively anti-AI in general. As with Microsoft, I feel like it could have better predicted the fallout.

What I told my wife would happen is what I then saw within 24 hours: Ubuntu fans are already debating which distribution they’ll change to, which distributions will have a strict no-AI policy, or whether there will be new distributions that are essentially Ubuntu minus all AI. A sort of UbuntuDebloat, if you will. As predictably, Canonical quickly responded to some queries indicating that one will be able to install future Ubuntu versions with no “explicit AI” features enabled during setup. Just like Microsoft later detailed the security controls in Recall and then even later pretended to address the nonsense concerns about them. But remember, the audiences are different. Windows is more mainstream, and Linux is more technical.

Ubuntu is in an interesting place. It’s more popular/used than all other Linux distributions (perhaps even when combined), and so it plays a unique role. I feel like its announcement was mostly on point and responsible, and to be fair to Canonical, there are some details even it’s not sure of yet because it’s so early. This is a kind of feeling out period, and I’m sure it will rethink some things if the feedback on them is overwhelmingly negative. But the Linux community is what it is. We have a lot of old-timers in the Windows community (like me) who are set in their ways–the continued silliness over whatever individual app or printer or whatever not working in Windows 11 on Arm is a classic/tragically misguided example–but the Linux guys are a hundred times more opinionated and uninterested in change.

The thing I really respect here, though, is just the acknowledgment that this is happening with or without us and that we need to deal with it one way or another. If your goal is to further divide the Linux community and then, by pretending that AI isn’t useful or important, to further relegate Linux to the sidelines, then yes, by all means, ignore AI. But if you realize as Canonical that you are a leader that needs to lead by example, what it is doing feels both smart and correct at a high level. Again, there will be some adjustments here and there in time.

But ultimately, the central premise of Linux will be maintained. If you want AI to whatever degree, Ubuntu and others will offer that, but they will also give users the choice to not use it.

What Microsoft will always get wrong (or “do”) is that it will at times put its own strategic requirements above the needs of its customers. Some of this is perception, like removing a Copilot icon from Notepad but leaving all the AI writing features in the app. But some of it is real, and can take the form of opt-in vs. opt-out. For example, it’s incredible to me that Recall was going to be enabled by default with an opt-out and hugely important that Microsoft reversed that decision. But there’s no extreme swing where all AI is opt-in Windows, that’s not happening because AI is a way for Microsoft to differentiate. And so the thing to watch here is how Microsoft handles adding AI features to Windows down the road. It will be “better” than the initial approach, but it will rarely truly satisfy the desires of technical enthusiasts, those who are largely anti-AI.

We won’t have long to wait. After months of Patch Tuesday updates with no major new features this year, the May update, in preview now, includes the dreaded AI agents on the Taskbar feature that unfairly got Pavaan Davuluri in hot water late last year. The pushback on that little feature will be an interesting test. And we’ll soon know who does more poorly, Microsoft … or us.

“AI’s public perception problem”

… is that sometimes the perception is true

🐉 Enter the Dragon

zhackwyatt asks:

I am looking to buy a Snapdragon X2 laptop this year — the are just coming out. But I was curious if you had opinions on the laptop manufacturer’s themselves. Which have the best support, which have the least crap-ware, which are the most reliable, etc. Or are they all pretty comparable and brand shouldn’t be a consideration.

I don’t have broad enough experience with Snapdragon X2 to know how much this has changed yet, but one of the more notable aspects of the first wave of Snapdragon X Elite-based laptops almost two years ago was how nearly identical they all were internally. PC makers differentiated their laptops with various design choices, but the day-to-day usage was largely identical.

In the intervening years, Qualcomm expanded the Snapdragon X lineup with X Plus, X Plus 8-core, and X chips, expanding availability to less expensive PCs. And with Snapdragon X2, it’s launched with three tiers, one of which, X2 Elite Extreme is new, upmarket, and will absolutely have a battery life tradeoff for the additional performance it provides. But I haven’t experienced that yet.

What I have experienced is what I saw with the Snapdragon X Elite-based Surface Laptop 7 and base-level Snapdragon X-based OmniBook 5 from the previous generation: Though the processors couldn’t be further apart from each other on the product chart, the day-to-day experience of just using the laptops is incredibly similar (and excellent). Both are still my favorite two laptops. That’s been true, so far, for the Snapdragon X2 Elite-based Yoga Slim 7x and the base-level Snapdragon X2 Plus-based IdeaPad Slim 5x.

But that’s also just two PCs. I need to spend more time with them and with other Snapdragon X2 laptops before I really know where things are at.

Per a question above, all Snapdragon X2-based PCs are Copilot+ PCs, meaning they are premium PCs, and so the experience will generally be pretty good or better. (The IdeaPad has more crap than the Yoga, for example, but both are consumer PCs.) But you will have the best experience with a business-class PC, and I have yet to see a business-class Snapdragon X2-based PC.

My favorite three PC brands are Surface, HP, and Lenovo, but it varies by model/brand too. Surface PCs are generally excellent, though I understand why some might want to avoid them. With HP, the higher-end OmniBook (X, Ultra) and EliteBook models are just about always excellent. And Lenovo has ThinkPad and Yoga, though, again, the IdeaPad Slim 5x I’m reviewing seems good. I don’t review ASUS laptops typically, but I am very much looking forward to the ASUS Zenbook A16 waiting for me in Pennsylvania. Acer and Dynabook (formerly Toshiba) are also usually good, especially the premium models, with some quirks.

Given my experiences with Surface Laptop 7, I’m curious to see what Microsoft does with Snapdragon X2 this year. I’d like to see more ThinkPads based on Snapdragon X2. The HP stuff will always be excellent. I may have a buying decision this year, too, but I’m waiting to see more. The X2 rollout has been surprisingly muted and slow given all the build-up. I don’t know why.

“OpenAI explains why ChatGPT suddenly loved goblins”

The company is literally run by goblins