Intel is Dead and x86 is Next (Premium)

The x86 instruction set that once defined personal computing is now a boat anchor holding back the PC industry and contributing to Intel’s death. Intel is screwed, but we’ll be OK. Microsoft and Qualcomm delivered a one-two punch on behalf of Arm in 2024, and in 2025, Nvidia and other chipmakers will arrive with the knockout blow that finishes off x86 once and for all.

This is what it’s like to live in interesting times. We’re in what’s called an inflection point, and it is a matter of history that Intel’s power was so vast, and its leadership so shortsighted, that the company kept x86 afloat artificially for many years longer than would have happened in an equitably regulated and truly competitive market. As comes up repeatedly in my ongoing history of Windows on Arm–Part 2 will be available soon–Intel illegally gamed the system for decades by paying PC makers to use its chips instead of those made by competitors, whether they were x86 licensees like AMD or companies using different architectures.

(And it’s still doing it. Back in May, I wondered aloud about the impact of Qualcomm’s latest PC-based Arm chips on Chromebooks, and was surprised when no Snapdragon X-based devices appeared. I learned why in September, when Intel announced the release of its “Lunar Lake” chips at IFA. If you watch the launch event, you’ll notice something odd: Four minutes in, after discussing its advances in “Lunar Lake,” Intel brings out its first partner. And it’s not Lenovo, HP, Dell, or Microsoft. It’s Google. And not just Google broadly: The vice president and general manager of Chrome OS and Education shows up to explain that Google has mysteriously chosen Intel over for Chromebooks and local AI. Curious.)

If dominant Big Tech monopolists have anything in common, and they do, it’s that they sing the same tune: These companies don’t need to be regulated, they argue, as unforeseen competition and market changes will always conspire to topple them in the end. But like all arguments against regulation, this stance is deeply flawed, in this case because it ignores the amount of time required for this disruption to occur, time during which the monopolist is free to continue contorting the market for its needs and, hopefully, finding ways to expand its dominance into new markets.

And so it was with Intel, which was so busy backing giant trucks full of money up to the headquarters of its PC maker partners that it lost the script. There are many ways to describe this period, but here’s one of the more concise. After securing Apple as a customer in 2006 when it transitioned the Mac off Power PC, Intel then fumbled the future by delivering lackluster mobile chips that couldn’t meet the needs of the PC market, let alone the iPhone and iPad that followed. Finally, Apple infamously jumped ship with the Mac as well, transitioning the Mac again in 2020, this time to the Arm architecture, leaving Intel–and x86–in the dust.

It was during this era–the roughly 15-year period between 2006 and 2020–that Intel failed.

It started with Paul Otellini, who became Intel’s first non-engineer CEO in 2005, succeeding Craig Barrett. Otellini won the Apple contract for the Mac, but the rest of his tenure was marked by failures that included massive layoffs and the steady decline of Intel’s chip fabrication facilities. It was Otellini who instituted Intel’s policy of paying customers not to use competing products, the grossest example of the firm’s shift away from focusing on technical excellence. Apple–and then the entire industry–shifted to Arm on Otellini’s watch.

Otellini’s successors continued with similar mistakes. Brian Krzanich was an engineer, at least, but his strategies to push Intel into non-microprocessor chips like those for cellular networking all failed, and he’s widely credited for Intel’s failure to move past a 10 nm chip manufacturing process at a time when TSMC had already surpassed that milestone. Bob Swan, another non-engineer, but with no technical background at all, then succeeded Krzanich briefly, lasting just two years and achieving absolutely nothing of note. And then Intel brought back Pat Gelsinger to clean up the mess.

But it was the compounded failures of this era that foiled the hoped-for resurgence. That the Intel board fired Gelsinger but still plans to plow forward with that man’s strategy, unchanged, says everything you need to know about Intel today and how little it’s evolved. The company is still making terrible decisions, and these decisions lead to one inescapable conclusion: Intel as we know it is doomed. There is no version of this story where Intel emerges on the other side as a formidable chip designer or chipmaker. It’s over.

That said, inertia is a powerful force. Intel is circling the drain and waiting to be picked apart via acquisitions, but x86 won’t disappear so quickly. This is due largely to the aging but sizable installed base and the unique market dynamics that make the PC so lackluster as a business. As recently as 2011, analysts were talking up a future in which PC makers would sell over 500 million units per year. But that milestone never happened thanks to the relentless growth of smartphones and other mobile devices. The PC industry is half that size today, and is it’s now a minority personal computing platform. The long-awaited PC refresh cycle, now delayed two years, may never materialize. We’re in a holding pattern. And it’s all Intel’s fault.

I’ve written many times about Microsoft’s many efforts over many years to get Intel to wake up to the mobile threat it first saw in the early 2000s. Steven Sinofsky highlights the problems the company faced several times in his book Hardcore Software, noting that PC makers were happy to compete with each other for thin slices of market share, ignoring Apple and the Mac, and the broader trends that upended the industry in the wake of the iPhone. That these same PC makers were being paid by Intel to sit still is, of course, astonishing. But in the absence of regulation, Intel was allowed to prevent innovation from occurring elsewhere in the one shrinking market it dominated.

It is perhaps ironic that the situation Intel now faces is the same as what AMD faced in the 2000s, when it was unable to compete with Intel’s chip manufacturing processes. AMD did what it had to by selling off its chip fabrication capabilities, creating a new company called Global Foundries. This seemed like a desperation play at the time, but it saved AMD and, more importantly, allowed it to innovate in ways that had been impossible previously: Freed of its out-of-date manufacturing business, AMD was free to create superior chip designs that could only be built by TSMC and its rapidly advancing manufacturing processes. It was the same model that Apple would soon follow, and to even greater success.

How the mighty have fallen. This past year, Intel was forced to also take advantage of TSMC’s superior capabilities for its most recent x86 chips: After treading water for 10 to 15 years, the company is incapable of producing its own chip designs reliably or at all. The difference between it and AMD is that Intel retained its chip fabrication facilities and all the debt and costs that come with them. And it is now investing an alleged $100 billion to expand into new facilities and improve existing fabs. I write alleged there because Intel won’t live to see this effort to its conclusion: As it continues to miss manufacturing milestones, it will also lose out on most of the promised $11 billion in government subsidies, loans, and contracts, I think.

This is an incredible moment in which AMD, the long-time also-ran that was pushed aside by the industry bully and its once-infinite mountain of cash, could race ahead of Intel and seize the PC market for itself. But that’s not happening. And if you find yourself doubting that x86 is on the verge of collapse, this is where you really need to pay attention. In the wake of its most recent earnings announcement, a quarter in which AI/cloud-based chip revenues jumped 122 percent and almost doubled those from its PC chip businesses, AMD quietly announced that it was restructuring to focus on AI/cloud–the datacenter–and not the PC market. “We are a data center-first company,” AMD CEO Lisa Su said, in an escalation of AI being AMD’s “number one strategic priority” back in May 2023.

So here we are in the waning days of 2024. Intel, which started the year with a market cap of over $200 billion, is now worth just $90 billion, less than half that figure. It was unceremoniously booted out of the Dow Jones Industrial Average in early November, replaced by Nvidia, the AI hardware powerhouse that’s now worth nearly $3.5 trillion and is the second-biggest company in the world, barely behind Apple at $3.7 trillion. Nvidia, like AMD, made a solid business for itself in PC graphics, but its current success–which AMD now wants a part of–has absolutely nothing to do with x86 or the PC market. (And yes, AMD’s AI/cloud datacenter offerings use x86 microprocessors, but they will succeed or fail based on GPU capabilities.)

AI is so big that Nvidia is now worth over 35 times as much as Intel. Think about that for a moment. And think about how the inertia that once benefitted Intel is now working against it as the clock ticks on a resurgence strategy that will never pan out.

In the end, Intel was disrupted. But like all defeats, it came from multiple setbacks, both internal and external. Intel ignored the needs of the market it dominated, failed to expand into new markets, and failed to innovate with its chip making processes at a time in which the rest of the world kept moving forward. It was aided and abetted by PC makers that were all too happy to take the Intel subsidies that made their thin margins slightly less thin. And all this was allowed to happen over far too much time, time during which faster-moving competitors surged forward. And now it’s too late. Too late to save Intel, and too late to reverse the inexorable change that has swept our industry.

Intel also failed where its biggest partner, Microsoft, succeeded. Where Microsoft pivoted successfully to cloud computing and then AI, Intel steered directly into the iceberg. Intel failed at its core businesses so badly that it no longer makes even the best x86 chips, let alone the best PC chips overall. It failed in graphics. It failed in the datacenter. It failed in chip fabrication. It failed in chips of every kind imaginable.

And with this failure, x86, too, is doomed. It is inefficient, complex, buggy, and outdated, everything that Arm is not. What once bound the PC industry together is now a vestigial past, outliving its usefulness. And now that Microsoft and Qualcomm have cracked that nut–Snapdragon X-based PCs deliver levels of performance, compatibility, and efficiency that are so far beyond previous generation Arm chips for PCs that they may as well have been given to us by aliens–we can collectively move forward. With a superior architecture, one that’s tied to the mobile present and future, and not to the desktop past.

This isn’t a slam dunk for Qualcomm, as it will soon have some competition in this space, too. Aside from the lawsuit, Arm Holdings recently announced a more formal push into PC chip designs, and it’s a long-time open secret that companies like MediaTek, Nvidia, Samsung, and even AMD are all working on Arm-based chips for PCs. This diversification will be good for the industry, and healthier than the current market, which consists of a fading monopolist (Intel) and two bit players (AMD and now Qualcomm). But most importantly, it’s good for the PC. This shift will result in new levels of quality, reliability, and efficiency in a product that’s not usually associated with those qualities.

In many ways, x86 can’t die quickly enough, and the real tragedy here is that PC users were denied the benefits of this shift for so long. It is likewise vaguely sad that the PC industry consistently follows and trails behind, and that in making its own shift to Arm on the Mac, starting in 2020, it once again showed us the way forward. It is now up to us, as an industry, to follow Apple’s lead. Intel’s incredible problems this past year finally make doing so both easy, obvious, and guilt-free.

x86 won’t disappear overnight. But it is already irrelevant, and it will disappear.

As for Intel, the future is inevitable, but the details are murky. My bet is on Intel splitting off its non-primary businesses piecemeal and then separating into two companies, one that designs x86 chips for PCs and datacenter, and the other a foundry that builds chips for other companies. Neither will be successful, and the last gasp of the Intel design company will be when it finally and inevitably licenses Arm again. Intel Foundry, meanwhile, will stumble forward as the B-player in a market controlled by TSMC. Perhaps it’s “too big to fail,” and the U.S. government will bail it out, citing national security.

Who cares? Intel and x86 are irrelevant. It’s the end of an era, and that brings a familiar but odd mix of fear and excitement. I get it. But I also refuse to see this as anything but positive and overdue. It’s time to make our peace with this reality and move on.