What’s Old is New Again (Premium)

AI is problematic on many levels. It’s complex and technical, and confusing to, and misunderstood by, even very technical people. It’s riddled with privacy concerns both real and imagined. It’s happening fast, really fast—one hopes, unsustainably fast—with new advances arriving on an almost daily basis. And complicating matters for those of us in the Microsoft community, the software giant has bet the future of the company—for real this time—on AI, to a degree that most should find shocking. This is head-spinning, rapid-fire change, and it’s happening at a company that was perhaps best known for conservative evolution in the Satya Nadella era.

And now Microsoft is coming for our PCs.

To date, Microsoft’s biggest AI advances have been in the cloud, with its Copilot family of services across Microsoft 365 and web, and with its vast array of developer services driving Azure usage ever higher. But what started as a trickle—a bit of background blur here, a lame Copilot sidebar there—is about to snowball into a tsunami that brings AI down to size on the PC desktop. And this is a case where the uniqueness of the PC really comes to bear: Sure, Apple and Google are adding on-device small language models (SLMs) to their respective smartphone platforms, but Microsoft this past week bragged about the over 40 SLMs that will ship inside each Copilot+ PC. That’s local AI at scale.

Because of the current climate—basically fear, uncertainty, and doubt—you’d think that Microsoft would tread lightly. But it is instead pushing forward as aggressively on the client as it did previously in the cloud, damn the bad PR and missed climate goals. And to ease our worries and fears, it is constantly referencing a past we all remember, when Microsoft was the sole overlord of personal computing and ruled by fiat. The good old days, if you will.

Frankly, this is smart. The Microsoft community—Microsoft itself, but also its corporate customers and enthusiasts—are still stung badly by the reduced prestige and influence triggered by the web and mobile waves. You could sense the overly-sensitive defensiveness in all the MacBook Air references in the Monday Copilot+ PC launch event, and while I feel like that over-emphasis was a bit much, one might also argue that it’s overdue: Microsoft ignored Apple and the Mac for too long in the past—I’m looking at you, Mac v. PC ads—after all. As with AI in general, it’s nice to see Microsoft come out firing with Windows and the PC again. It’s been a while.

Anyway, Microsoft’s constant references to the past fascinate me. Which makes sense, given that I wrote a comprehensive history of Windows as a platform in Windows Everywhere and have worried for years about Microsoft ignoring (or worse, undermining) Windows so often.

Here are some obvious examples.

When Microsoft CEO Satya Nadella took to the stage at the start of Monday’s Copilot+ PC event, he referenced the Windows 95 launch, in part because that event had happened at Microsoft’s corporate campus in Redmond. But also because that product milestone remains the feel-good apex for Windows and the PC, and Microsoft is keen to capture at least some of that excitement in this “unbelievable new era of AI.”

Nadella also referenced the Win32 PDC—for the upteenth time—meaning the 1992 PDC, which is often called the NT PDC or the Win32 PDC, as it marked the unofficial start of the Win32 era. He also noted that Win32, .NET, and Azure are the key moments that mark his professional life to date. Win32 (as embodied by Windows) and Azure are absolutely the key markers of the past 30 years for Microsoft. But .NET is an interesting entry in that list because it never lived up to the hype, though its ability to survive over two decades and morph into a sophisticated cross-platform solution is rather impressive.

Nadella also addressed the AI speed issue I noted above.

“These scaling laws [of the PC industry in the past], Moore’s Law, helped drive the information revolution,” he said. “You could say Moore’s Law was probably more stable in the sense that it was scaling at maybe 15 months, 18 months. We now have these things that are scaling every six months or doubling every six months … The effect of these scaling laws is a new, natural user interface that’s multimodal.”

He even went so far as to compare what’s happening now to the dawn of computing. Not personal computing, which dates back to the late 1970s. But what he calls “modern computing.”

“If you go all the way back to even the birth of at least modern computing, 70 years ago, the pursuit has always been about how to build computers that understand us instead of us having to understand computers,” he said. “I feel like we really are close to that real breakthrough … If you think about the new user interface, the universal user interface that’s multimodel that can support text, images, video, both as input and output, we have that. We have memory that retains the important context, recalls our personal knowledge and data across all the applications and devices, and we have these new reasoning capabilities that help us complete complex tasks. We’re entering this new era where computers not only understand us, but can actually anticipate what we want and our intents.”

When you think about hybrid AI—that is, a combination of cloud-based AI and on-device AI—it’s hard not to remember Microsoft’s “better together” messaging of the client-server past. Windows 2000, you may recall, worked great on the PC clients and on pre-cloud servers of the day, but customers saw exponential gains when they adopted both in tandem. Today, we’re making that same leap, and that’s mostly interesting because it requires an advance past the “rich client” of the past, the more sophisticated Copilot+ PC. This is not a thin client, though some of this will work occur in web browsers too.

“We believe AI will be distributed,” he said. “The richest AI experiences will harness the power of the cloud and the edge [client] working together in concert. This in turn will lead to a new category of devices that turn the world itself into a prompt, devices that can instantly see us, hear, and reason about our intent and our surroundings. For us, this vision starts with our most beloved and most widely used canvas, Windows.”

The most fascinating references to the past, to me, are those that conjure up memories of Windows 8. And it’s difficult to know where to land on this, as that product is inarguably the nadir of this history, a horrible moment in time when the team responsible for Windows had all the right ideas but couldn’t have screwed up the implementation more badly. I respect today’s Microsoft owning that history, in a way, and there are undeniable parallels between the work being done with Arm both then and now. Still. It’s a little weird.

“It really comes down to rethinking the core architecture of the PC across silicon, systems, and the end-user experience,” Nadella said, sending my brain flying back in time to the buildup to Windows 8, when that earlier team said they were “reimagining Windows from the chipset to the experience.”

“We wanted to reimagine what Windows could be,” Steven Sinofsky recalls of this time in Hardcore Software. “We knew we would always have x86 PCs and those would always support the existing body of applications, hardware peripherals, and the ecosystem that provided them. There was little life left in the PC ecosystem … Nearly all the innovation that expanded beyond memory, storage, and CPU happened and continues on mobile platforms.”

“Our goal was to infuse Windows with the attributes [I discussed recently in this post] so that new Windows 8 devices and future devices could have the qualities of smartphones and tablets that consumers were experiencing.” Then, as now, the goal was not to target the low-end of the market—”better netbooks”—but to push Windows into a more reliable, efficient, and mobile space.

This work, then as now, would require Windows to run on new silicon, new Arm-based chipsets that are more efficient and performant per watt than the now-legacy x86 chipsets that have, to date, defined the PC.

“To make this possible, we’re introducing a new system on a chip [SoC],” Nadella said, oddly not mentioning Qualcomm. “We’re also modernizing the entire operating system to unleash the power of the SoC and deliver these next-generation applications, models, and experiences that are simply just not possible on today’s PCs. The result is the fastest, most AI-ready Windows PCs ever built.”

In the Windows world, we were introduced to the term SoC when Sinofsky introduced Windows RT, the first port of Windows to the Arm platform. As he describes in the aforementioned book, an SoC package[d] much of the whole board of a netbook into a single chip that specifically includes at least the microprocessor and graphics, resulting in a small and energy-efficient package.” Ignoring the grammar issues, that statement is correct, but it underplays the central point, which is to make the PC more like the smartphone architecturally.

Since Windows 8, traditional x86 PCs have adopted some of the benefits of SoCs, and some x86 chips are technically SoCs. And Microsoft’s work with Arm continued, albeit in unsuccessful form, in Windows 10 on Arm and now Windows 11 on Arm. Qualcomm’s Snapdragon X family of SoCs appears to finally overcome the performance issues of the past, its Prism emulator appears to address the compatibility issues of the past, and so we might argue that only now, 12 years later, has Microsoft achieved its original goals for the hardware side of this port.

But the software goals have changed: Where Windows RT sought to make Windows make sense on what I’d call a device, Copilot+ PC—the platform, which is Windows 11 plus specific NPU-driven AI experiences—is obviously focused on what I’d call a PC. Windows RT was half-hearted—ideally it would have dispensed with the desktop interface entirely and made a go of it with the Metro environment—and I see parallels with Copilot+ PCs. These capabilities are not available to everyone: As with Windows RT, you need to buy a new device, and in this case, that new device is specifically marketed as a PC.

Here’s a curious claim.

“The web was born on Windows,” Nadella threw out on Monday, almost as an aside. That’s not true—Tim Berners-Lee invented the World Wide Web on a NeXT computer—but, again, I think Nadella was going for that positive association with the past vibe: Windows was the vehicle through which the vast majority of people first experienced the web, for sure, and it all started in that heady era in which Windows 95 launched. So this is another link to the past, accurate or not.

There is so much more, but I think you get the idea. And incredibly, everything I mention here occurred in the first 7 minutes of Monday’s event. From there, Nadella handed off the presentation to Yusuf Mehdi, who is another obvious link to the past: A 32-year veteran of the software giant, Mehdi joined Microsoft in 1992 where he—in his words—”played key roles in the launch and popularization of Windows with Windows 3.1 and the industry launch of Windows 95.”

Incredible.