Sunday, December 15, 2024

Intel CEO takes his leave as ambition meets reality

Must read

Comment Pat Gelsinger is out as Intel CEO, cutting short his nearly four-year crusade to revitalize the beleaguered chipmaker.

Gelsinger’s “retirement” as CEO and departure from the x86 behemoth’s board of directors is immediate, having gone into effect on Sunday, though Intel didn’t disclose the decision until today.

The abruptness of Gelsinger’s departure and lack of a succession plan suggest it probably wasn’t voluntary.

“I think the board made a decision or overruled a decision that Gelsinger thought was a bad one and he was out — likely related to the splitting Foundry from the design company,” Patrick Moorhead, chief analyst at Moor Insights and Strategy, told The Register. “Something happened in the last week.”

“I think people are missing the speed at which this went down. They backdated the announcement. He retired on December 1; it goes on December 2, right? He’s not on the board. He’s not an advisor… There is a hard separation between the board and Pat,” he added.

While Intel would never admit it, it wouldn’t be the first time the board had lost faith in its chief exec. Former CEO Bob Swan’s time at Intel ended rather abruptly in early 2021 amid pressure from activist investors and mounting pressure from competitors. His departure, of course, made way for Gelsinger’s return as CEO that same year.

In his stead, CFO David Zinsner and Michelle Johnson Holthaus, who runs Intel’s Products division, have been named interim co-CEOs while the board searches for Gelsinger’s replacement. Board member Frank Yeary, meanwhile, will take over as interim board chair.

Intel’s prodigal son

Gelsinger’s return to Intel in January 2021 came as the chipmaker faced mounting pressure from rivals – most notably AMD, which has been steadily stealing market share from the x86 giant – as well as activist investors unhappy with its financial trajectory.

His appointment was heralded as a turning point for Intel with many lauding the decision to put a proper engineer rather than a bean counter at the helm. Better yet, Gelsinger knew the chipmaker well. Prior to his time as COO of Dell EMC and later CEO of VMware, he’d spent nearly three decades at Intel where he helped architect the venerable 80486 before rising through the ranks to become the company’s first CTO.

At the time, Gelsinger’s engineering background seemed like it was just what the doctor ordered as many of Intel’s biggest hurdles were technological. Its 7nm process tech – since rebranded as Intel 4 – had been severely delayed by design flaws disclosed in mid 2020. At the same time, Intel was struggling to roll out its 10nm process tech, which wouldn’t see widespread release outside the notebook space until early 2021.

Also, AMD was already shipping its second-generation of Ryzen processors based on TSMC’s 7nm process tech and Apple had just launched its M1 SoC on the Taiwanese foundry operator’s 5nm tech.

And while Intel’s 10nm transistor densities may have been closer to TSMC’s 7nm tech, public perception was that while Intel was the market leader, its technology was beginning to lag that of its rivals.

Ambition meets reality

Gelsinger aimed to change this while simultaneously charting a new and ambitious course for the chipmaker.

A little over month after taking over, the newly appointed CEO announced plans to open the company’s fabs to contract manufacturing with the formation of Intel Foundry Services and invest more than $20 billion in a pair of leading edge manufacturing plants in Arizona.

Up until that point, Intel was one of the very few companies designing and manufacturing its own chips. The rest of the industry was busy gravitated toward a fabless model, with companies like AMD and Nvidia designing with industry-standard tools with hand-off of manufacturing to the likes of TSMC and Samsung.

In the midst of a pandemic-fueled semiconductor shortage, Gelsinger saw an opportunity to challenge TSMC and overtake Samsung as the second largest foundry operator.

During his tenure, he announced more than $100 billion in planned investments to massively expand Intel’s manufacturing footprint across the US, Europe, and the Middle East. At the same time, Intel began development of next-gen manufacturing processes called Intel 20A and 18A, which were expected to close the gap with TSMC and Samsung’s 2nm process nodes.

Unfortunately for Gelsinger, realizing these ambitions wouldn’t be easy. He was essentially starting from scratch with foundry being of limited utility until 18A and the fabs to produce it were ready in 2026.

Despite Gelsinger’s grand vision, he never could quite escape the perception that Intel was falling behind.

While he was free to chart a new course with Foundry, he was burdened by a product roadmap too far along for a fresh start and (we suspect) overcoming that momentum proved more challenging than he’d bargained for.

Arguably, the most embarrassing example of this was Intel’s infamous 4th-Gen Xeon Scalable processors, better known as Sapphire Rapids. The chips were originally slated for release in 2021, but repeated setbacks and poor yields pushed its release back to early 2023 and even then, bugs in the product forced a brief halt in shipments for some SKUs.

However, it wasn’t just Intel that suffered as a result of the delays. An HBM-equipped variant of Sapphire Rapids was, alongside Intel’s Ponte Vecchio GPUs, slated to power the Argonne National Laboratory’s Aurora supercomputer.

With a peak performance of nearly 2 exaFLOPS, the machine was expected to overtake the AMD based Frontier system at Oak Ridge as the United States’ most powerful publicly known super. Unfortunately, the system fell short of expectations achieving just over an exaFLOP of double precision performance in the Linpack benchmark this spring. By this fall El Capitan’s debut meant it would never claim the number one spot.

This, it seems would be the final hurrah for Ponte Vecchio as, around the same time, Intel reportedly began sunsetting the product. Amid the AI boom, Intel shifted focus to its Gaudi3 accelerators in the hopes the chips would drive half a billion dollars in revenues in 2024. Those revenues, Gelsinger would later admit, would not materialize this year.

Intel’s client division wasn’t without controversy either. Most recently Intel acknowledged a defect in its 13th and 14th-gen desktop CPUs which caused degradation and instability of the parts.

Not long after, Intel announced its next-gen desktop CPUs, the first to use its all-new Intel 20A process tech, would instead be built by TSMC. Following Intel’s earlier decision to manufacture its Lunar Lake mobile CPUs at TSMC, only a small number of products are still built in house.

We’ll note that very little of this was actually Gelsinger’s fault. Product development for new silicon takes years and work on these chips almost certainly began well before he took over as CEO.

The collapse

The same can’t be said of Intel’s ailing foundry unit, for which the real reckoning came in early 2024 when Intel officially split off Foundry as a standalone unit and released revised financial reports revealing that the unit had bled $7 billion in 2023.

In the quarters since, Intel Foundry has posted more than $11 billion in operating losses. Investors clearly felt lied to. The precipitous decline of Intel’s share price — currently down 47 percent since the start of the year – has spurred multiple class action suits from investors who claim Gelsinger and Zinsner misrepresented the outlook of its foundry division.

However, we’ll note these losses really shouldn’t have come as a surprise to anyone. While Gelsinger was busy building Intel up to be a foundry giant, its product division was shifting ever more of its production to TSMC.

Gelsinger’s insistence that production would begin to return home in 2025 with the launch of Clearwater Forest in the datacenter and Panther Lake on client side, it seems faith in Intel’s ability to follow through on its promises has been lost.

Just weeks after Intel posted a record breaking $16.6 billion loss in Q3, the largest in its history, Nvidia supplanted Intel on the Dow Jones Industrial Average.

Gelsinger for his part has taken drastic action to get Intel’s finances under control, having previously announced plans to lay off more than 15,000 staff — or about 15 percent of its workforce — by the end of the year, cut capital expenditures by 20 percent, and end quarterly dividends beginning in the fourth quarter.

Gelsinger attempted to assuage investors by spinning off Foundry as an independent subsidiary with its own board, a move he argued would bring in new sources of capital for the ailing business.

The announcement also saw Intel scale back its foundry expansion and “pausing” development of its €30 billion fab project in Magdeburg, Germany and $4.6 billion assembly and test facility in Wroclaw, Poland.

The delay will no doubt cost Intel any subsidies it’d hoped to claim under the European Chips Act, as we don’t anticipate the EU will fund delayed fabs. Without the funding, we suspect that “pause” may in fact be permanent.

Intel’s uncertain future

With Gelsinger out of the picture and no clear successor on the horizon Intel’s future remains an open question. What we do know is whoever Intel finds to fill Gelsinger’s shoes has one heck of a knot to untangle.

For some, the obvious course would be to cut Intel’s losses, abandon Foundry, and fully embrace TSMC as rival AMD has. As it is, Intel is already outsourcing production of many of its current-gen products.

However, spinning off or even selling Foundry will be far more challenging than proponents of such a plan would have you believe.

A spin-off would please Intel’s investors, eager to get the blood off the books. But it’s hard to see how a company now burning more than $5 billion a quarter is supposed to survive on its own even with Intel as a guaranteed customer. The reality is Intel Products is Foundry’s only customer of consequence, and that’s still not enough to turn a profit. It’ll stay that way at least until 18A reaches volume production, which isn’t until late 2026.

The design tools necessary to harness older Intel process nodes either don’t exist for outside customers or are limited to less attractive, legacy nodes.

Further complicating the matter is the fact Uncle Sam has a vested interest in Intel’s success. As we’ve previously discussed, Intel is the only domestic supplier of leading-edge process technology, arguably making it the most important chipmaker in the country with regard to US National Security policy.

Because of this Intel has been awarded roughly $7.86 billion in CHIPS Act subsidies to support development of domestic fabs and is set to receive another  $3 billion to establish a secret enclave for the development of leading edge chips for the US government.

Taking this cash comes with its own restrictions, namely that Intel must retain 50.1 percent ownership or voting rights in the foundry unit if it is ever spun off. What’s more, any spinout would require Intel to remain a Foundry customer.

And so while Gelsinger will no longer be in the picture, he’s made it exceedingly difficult to walk away from his foundry dream. ®

Now read: With Gelsinger gone, who benefits from an Intel break up?

Latest article