Can AI Save the PC Industry? (Premium)

Microsoft and its chipset and PC maker partners offer bullish predictions about a future in which AI innovations drive a new wave of PC upgrades. But the PC market is in freefall right now, and it’s only getting worse. Can AI save the PC industry? Or is it already too late?

It’s not looking good right now: The PC industry has now suffered 8 consecutive quarters—two full years—of year-over-year sales declines, during a pandemic-era boomtime into a digital Great Depression. And there’s precious little in the way of a silver lining here, despite the cautious predictions of future stabilization growth from analysts at Gartner and IDC and elsewhere in recent quarters.

And that makes this past quarter’s news so much more horrible: Hardware makers sold 66.25 million PCs in the third calendar quarter of 2023, a decline of 7.4 percent when compared to the 71.15 million units sold in the same quarter one year ago. (As always, these numbers are averages of those provided by each firm.)

Now, the glass-half-full crowd will point out, correctly, that PC sales had fallen an even higher 17.25 percent in that year-ago quarter. And fair enough, but the general trend here is that the expected market stabilization keeps getting pushed back: We were told that the sales shortfalls end by this point in time and then perhaps to expect some modest growth in the years ahead as corporate upgrade cycles kicked in again.

The post-pandemic boomerang effect was bad enough—it’s still ongoing, after all—but the promise of a white knight riding to save the industry in the form of AI makes today’s dull reality even harder to swallow. The dream is compelling: Thanks to a coming wave of AI innovations, PCs will suddenly become even more important, and consumers and businesses alike will upgrade en masse in ways they have not done since the rise of the Internet and the release of Windows 95 in the mid-1990s.

I like this dream. It’s nostalgic and overtly positive, and it harkens to a bygone again in which Microsoft reigned supreme. But I also realize, pragmatically, that those days are over. And even Apple, the darling of the consumer world, no longer sees lines of customers snaking around the corner from its retail store on product launch days. In the modern era, we all buy more products than ever, but we do so behind closed doors and those products are shipped to us, at home. Oh, there are still lines. They’re just lines of trucks on the highways.

And maybe more to the point, AI as a technology will lift all boats and there is no reason to think that its impact will incommensurately impact the PC industry. Back in March, I noted in When Everything is AI, Nothing is AI (Premium), that when AI becomes table stakes, as is happening, the playing field remains unchanged. Back then, I was concerned about Bing’s position in the market vis-à-vis Google Search, and it’s fair to say that my argument was sound, since Bing’s usage share hasn’t moved an iota since it introduced AI capabilities via Bing Chat. Even Microsoft CEO Satya Nadella was forced to admit that he had been wrong about the impact AI would have on Bing while testifying in U.S. v. Google.

So why on earth would we believe that AI will reverse the PC industry’s slide? Is there anything special about the PC that makes it particularly ideal for AI workloads or capabilities?

I don’t believe there is.

Instead, I think that AI will benefit users on PCs just as it will users on smartphones and other devices, and that the addition of AI will do nothing to change the relative positions of these devices in the market and in users’ lives. Again, this is about table stakes: If everything is powered by AI in the future, then these products will simply continue to compete on their own merits. Each product will be improved, yes, but their positions relative to each other—unit sales (market share), usage share, whatever—will remain the same. People aren’t suddenly going to put down their smartphones and use PCs more often.

This doesn’t diminish the impact of AI on the PC, necessarily, though it’s alarming to me here in October 2023 that we still don’t have a clearer understanding of how AI will transform the user experience and how, exactly, coming hardware upgrades like NPUs (Neural Processing Units, essentially a way to hardware-accelerate AI tasks in the same way that GPUs do for graphics) will improve matters in the slightest.

That surprises me. This year has been a heady rollercoaster ride of AI explosions, not just in the Microsoft ecosystem but across the industry. But if we look specifically at what we know about AI on PCs, the story is more sober than I’d like. PCs are just one way that users can and will access Microsoft Copilot features like Bing Chat and Bing Image Creator, as is the case with third-party solutions from OpenAI, Google, and others. The additional Copilot capabilities in Windows 11 were weak when announced and have yet to improve. Only the capabilities in Microsoft 365 Copilot are at all fleshed out and substantive. And those will only be available to commercial customers (businesses) at first and will be opt-in because this service will cost an additional $30 per month per user on top of the normal monthly cost of Microsoft 365. Yikes.

Coming into last month’s Microsoft special event, I expected more clarity. I predicted that Microsoft would dramatically expand our understanding of what was coming, and when and where it was coming. But we got very little of that. The AI features in Bing Chat/Edge, Windows 11, and Microsoft 365 commercial were not expanded on in any way, and we only vaguely learned that there would be a Microsoft 365 consumer AI offering of some kind sometime in the future, details TBD. More specifically, we learned nothing new about PCs and NPUs, at all.

And that’s curious. To date, Microsoft has exactly one AI-based feature that it can tout, called Windows Studio Effects, and until this past week, it was only available on niche Arm-based PCs like the Microsoft SQ3-based Surface Pro 9. (The new Surface Laptop Studio 2 also has an NPU.) But background blur on webcams is common these days, AI hardware or not. And we need much more. We need that AI killer app.

I know that AI has the potential to make us all more productive. It can help users who are not familiar with certain things—complex products like Excel or PowerPoint, creative solutions like image creation and editing, whatever—accomplish things that to date have required expertise, training, or assistance from the outside. But again, table stakes. If we are all x percent more productive on PCs in traditional apps, and x percent more able to edit photos and videos on phones, what changes exactly? The PC is still a PC, and it’s still used for those tasks where a big screen and hardware keyboard and mouse are more efficient. But the smartphone is still the center of our digital lives. And PCs still represent work.

As an old-school PC fan, I view the smartphone era from two perspectives. On the one hand, this pocketable and more personal device stole away users’ time and engagement, and most people now spend more time with their phones than they do with all other devices combined. But on the other hand, the smartphone era helped the industry focus the PC on those productivity tasks for which it is best suited. (Windows 8 was a stupid attempt to counter that change which failed for all the right reasons.) The PC is used for “less,” I guess, but what it’s best at is important.

And as an old-school PC fan, I’ve also watched warily as the “post-PC” crowd sought to not just diminish the PC but to replace it. Over a series of generational improvements, Apple transformed the iPad from a consumption-only, unnecessary extra device into a new kind of computer that’s so versatile that it can straddle, sometimes awkwardly, the consumption/creation divide, and, for some, replace the PC. Over in Google land, the Chromebook has emerged as a modern take on the PC, one that isn’t just cloud-based but rather cloud-centric, but also improved over generations with more powerful capabilities and Android app compatibility.

Neither has struck the death blow to the PC, at least so far. But the aggressiveness of the makers of those platforms should give any Windows fan pause. And it’s fair to say that these devices, which are are simpler and less expensive than PCs, but are also now imbued with many of the same properties that make PCs so useful, are a threat. And here’s a wake-up call: These devices are further along on the AI wave than is the PC. The chipsets that drive Android and iOS/iPadOS devices have had NPU hardware (by whatever branding) for years. And those platforms have been using that magical combination of hardware and software to make users’ lives better for years as well. The PC, as always these days, is behind.

This is what we’ve gotten so far. Intel has released its first NPU-integrated chipset for PCs in Meteor Lake. AMD has released exactly one NPU-based CPU model. Qualcomm has been doing NPUs for years, and it’s on the verge, we hope, of a NUVIA-powered Windows on Arm performance breakthrough that could finally take that product line from niche to mainstream. PC makers are rallying around NPUs and this coming AI era, but then of course they are, their literal survival is riding on this upgrade cycle. And then there’s Microsoft. The company that kicked off the 2023 AI explosion, perhaps impulsively, when it revealed Bing Chat back in February and tried to make “Google dance.”

Well, the sleeping giant has awoken, Microsoft. Giants, really, since Google is not alone is responding to the Microsoft threat with an incredible array of new products and services of its own. And here again, I keep coming back to table stakes. (And I hate that term.) So it is on Microsoft to ensure that its client platform, Windows, doesn’t just meet the bar but raises it on AI capabilities. And it just isn’t addressing that challenge, as was demonstrated by the lackluster September event.

As a Microsoft watcher, this puts me in familiar territory as I often have to look forward to the next event, to the next year, or to some future time when maybe, just maybe, Microsoft will do something material to improve this platform. We have key milestones each year we can point to, like Build and Ignite, and we have one-off events that come out of nowhere like the February Bing event and the September special event. And I must now point to November, and Microsoft Ignite, and the rumors of Microsoft AI silicon. And say, maybe. Maybe this will be time.

But it probably won’t be. So maybe it will be CES 2024 and all the NPU-powered PCs. Maybe it will be Windows 12, that AI-powered product I predicted way back in February. Maybe, maybe, maybe.

The reality of today’s PC market is a sobering and ongoing reminder that talk is simply that. And that we have yet to see a single compelling reason, a killer app, to believe that AI will do anything at all to rescue the PC.

And I want to believe. I really do. But that’s where I am right now.