Firing Pat Gelsinger doesn't solve Intel’s problems

Despite Intel’s recent woes, I didn’t expect to see CEO Pat Gelsinger joining 15,000 or so of his colleagues being shown the door. Gelsinger is a storied engineer and business success who laid down an exhaustive rescue plan when he took the helm of the beleaguered chipmaker in 2021. It was never going to be a quick fix, given the company’s long legacy of missteps. Gelsinger may be the public face of Intel’s current malaise, but the problems started long before his tenure and will likely keep going.

How Intel got here

Gelsinger was tasked with addressing almost two decades’ worth of bad decisions, all of which have compounded. Intel became an industry-swallowing behemoth as one half of the Wintel alliance, producing chips that went hand-in-glove with Microsoft Windows. The vast profits that flowed from this partnership meant there was an institutional reluctance to look too hard at new business ventures that could distract from its golden goose, still going strong all these years later.

In 2005, then-CEO Paul Ottellini turned down the chance to make the iPhone’s system-on-chip. It would have been easy for Intel, since it already made XScale ARM chips for mobile devices. You could find an Intel ARM chip inside popular phones like the BlackBerry Pearl 8100 and Palm Treo 650. A year later, it would sell XScale to Marvell, believing it would be able to shrink its x86 chips to work on smartphones. The first Intel Atom handsets showed some degree of promise, but the Snapdragons of the day — produced by considerably smaller rival Qualcomm — beat them pretty easily.

At the same time, Intel was working on Larrabee, its own discrete GPU platform based on the x86 architecture. Despite several years of marketing bravado and suggestions it would “kill” AMD/ATI and NVIDIA, Intel axed it in 2010 in favor of bundling integrated graphics into its regular processor products. The decision would hand the bulk of the GPU market to NVIDIA, making it the go-to name for gaming, supercomputers, crypto and AI, posting quarterly revenues of $35.1 billion on November 20.

Could Intel have foreseen the meteoric rise of AI? Maybe not. But Reuters reported former Intel CEO Bob Swan turned down the chance to invest in OpenAI in 2017. It was looking for a hardware partner to reduce its reliance on NVIDIA, offering a generous deal in the process. Swan, however, reportedly said he couldn’t see a future for generative AI, and Intel’s data center unit refused to sell the hardware at a discount.

Intel’s core strength was in the quality of its engineering, the solidity of its product and that it always kept close to the cutting edge. (There are parallels to be drawn between Intel and Boeing, both of which are watching their reputation for quality erode in real time.) Sadly Intel’s bread-and-butter business hit the skids after the company failed to produce 10-nanometer chips by its planned 2015 deadline. The company’s famous “tick, tock” strategy of launching a new chip process one year and a refined version the next ground to a halt.

These issues enabled Intel’s competitors to step in and steal a march, harnessing more modern chip architectures. AMD, which held a little over 10 percent of the chip market for much of the 2010s, has seen its market share double in the last few years. The biggest beneficiary, of course, was TSMC, the Taiwanese chip factory that has become the envy of the world. Even if Intel controls the bulk of the x86 processor market, it’s TSMC that makes the chips for Apple, Qualcomm, NVIDIA and AMD, among others. Intel, meanwhile, was saddled with an older chip manufacturing process that it couldn’t use to catch up with its rivals.

The Gelsinger doctrine

Gelsinger was as close to an Intel “lifer” as you could imagine, joining the company at 18 and rising to the position of Chief Technology Officer by 2001. In 2009, he left Intel to become COO at EMC and held the position as CEO of VMWare for almost a decade. After taking the reins at Intel, he laid down a detailed plan to mastermind its glorious comeback.

Step one would be to separate Intel’s design and manufacturing business into two distinct entities. With one eye on US subsidies through the Biden administration's CHIPS and Science Act, Gelsinger pledged to build two new chip factories harnessing the same EUV (Extreme Ultraviolet Lithography) technology used by TSMC.

Gelsinger was also determined to reestablish discipline in Intel’s chip business and get back to the “tick, tock” structure. Unfortunately, the production delays that had been building up since 2015 meant that Gelsinger’s target was just to get back to parity. In the interim, Intel would also get TSMC to manufacture some of its newest chips which, while costly, would help address any concerns the company was lagging even further behind.

Nobody had any doubts as to the size of the task facing Gelsinger, but there was plenty of room for optimism. Gelsinger was humble enough to accept Intel couldn’t simply stay on its current course, and had to embrace its new status. He proposed Intel could grin and bear the short-term pain for the company's eventual benefit. If it could build for the future, harness its rivals to keep it in the game and restore faith in its processes, Intel would emerge from this as the winner. All it needed was for nothing to get worse.

Things got worse

At the end of October, Reuters reported Gelsinger made a colossal faux-pas when speaking about TSMC. The CEO was quoted saying “You don’t want all of our eggs in the basket of a Taiwan fab,” and that “Taiwan is not a stable place.” This offended TSMC to such an extent that it ended a discount Intel had taken advantage of for years

Sadly, Gelsinger’s desire to restore discipline to the chip division would also backfire, with the latest Core processors blighted by voltage instability issues. Intel was forced to extend those chips’ warranties, which came at an additional cost it couldn't really afford. In August, it posted a loss of $1.6 billion and pledged to cut 15,000 employees in an attempt to right the ship. But it was forced to post the biggest quarterly loss in its history three months later, losing $16.6 billion, albeit much of that tied to revaluing company assets and paying for the layoffs. Worse, Intel’s new production process, 18A, reportedly failed crucial tests ahead of its 2025 debut.

Perhaps the lowest point in Intel’s year was when its stock price fell low enough that it became a takeover target. Rumors suggested Qualcomm was potentially eyeing a takeover while others indicated ARM had made inquiries about purchasing Intel’s product unit.

Where does this leave Intel?

The New York Times reports Intel’s board grew frustrated with Gelsinger as his rescue plan was “not showing results quickly enough.” But Intel wasn’t going to hire Gelsinger in 2021 and suddenly bounce back in 2024. Building large and complex chip factories isn’t easy. Nor is getting thousands of engineers to solve difficult problems around chip yields. And obviously reversing a slide that started in 2015 was never going to happen overnight.

Intel’s board is presently looking for a full-time successor to Gelsinger but it’s hard to see what someone else would do differently. After all, the company still needs to build those factories in order to own and control its future, and it still needs to fix its processes. Unless, of course, the next CEO is going to be told to just stanch the bleeding and keep the money rolling in. Even in its deeply-wounded state after a few bad quarters, Intel is still the biggest name in the x86 chip world and will keep making money for years to come.

You could easily imagine Intel’s board sitting around, prioritizing a few years of healthy profits at the cost of the company’s long-term future. It can keep selling modified versions of its existing desktop chips, ceding the technological leadership to AMD, Qualcomm and others. There’s probably a decade or two of big industrial clients who would be happy using Intel processors for their hardware for as long as they’re still using Windows. Perhaps that would be fitting given how big and ossified Intel has become, admitting that it can’t move fast enough to evolve.

It’s likely that scenario won’t be allowed to happen given Intel’s broader role in the global tech space. Even if the incoming administration criticized the CHIPS Act — Intel is still set to be its largest funding recipient — having a domestic manufacturer of Intel’s scale will be an asset few sane governments would allow to fall. But just switching CEOs won’t suddenly fix the company’s big, hard-to-solve problems. It wasn’t Pat Gelsinger who screwed up power design for Raptor Lake, nor did he pass on the opportunity to make the iPhone CPU all those years ago. The TSMC stuff, he can own that, but while a CEO sets the direction of travel, he can’t micromanage every process in a company of Intel’s scale. So whoever replaces him will have the same big stack of issues to tackle, knowing that the board’s patience will be even shorter this time out.