Intel suggests it will wait for new CEO to make critical decisions to fix manufacturing crisis
Intel told employees Thursday it may postpone a decision on how to fix its manufacturing crisis, likely waiting for
new CEO Pat Gelsinger to join the company next month before deciding whether to outsource advanced manufacturing to rivals in Asia.
The chipmaker committed to investors in October that it would make that decision by the time it announced its fourth-quarter financial results, saying that would leave just enough time to make the switch in time to produce the new chips by its target date in 2023. That announcement is scheduled for next Thursday.
But with Gelsinger’s surprise hiring Wednesday – he starts work on Feb. 15 – the chipmaker wants to give him time to weigh in. That’s according to an account of a Thursday all-hands meeting provided to The Oregonian/OregonLive by Intel employees. The company said it still wants to make the decision “as quickly as possible.”
“We expect to make that decision very soon,” outgoing CEO Bob Swan told employees at the meeting on Thursday, “but we’re going to do it with Pat.”
Intel has suffered a succession of manufacturing failures that derailed three consecutive generations of microprocessor technology, most recently with its forthcoming 7-nanometer chips. The resulting delays cost Intel its historic lead in semiconductor technology, along with precious market share and tens of billions of dollars in market value.
Now, Intel must decide whether to admit technical defeat and outsource its leading-edge chips to rival manufacturers in Asia.
It’s a momentous choice that will have
enormous implications for Oregon, home to Intel’s largest and most advanced operations. Intel employs 21,000 Oregonians, more than any other business, and spends billions of dollars every year to equip and maintain its Hillsboro factories.
Intel must make its decision under duress, with competitors encroaching on its turf, marquee customers like Apple choosing to make their own chips instead of using Intel’s, and as investors demand Intel consider selling off its factories.
“We have to deliver better products to the PC ecosystem than any possible thing that a lifestyle company in Cupertino” makes, Gelsinger told employees Thursday. That’s a derisive, if good-natured, reference to Apple and the location of its corporate headquarters.
“We have to be that good, in the future,” Gelsinger added.
Intel declined to comment on Thursday’s employee meeting or its outsourcing plans. It referred to a statement issued Wednesday, in conjunction with Gelsinger’s hiring: “The company has made strong progress on its 7nm process technology and plans on providing an update when it reports its full fourth-quarter and full-year 2020 results as previously scheduled on Jan. 21, 2021.”
Intel has already ceded its historic lead in manufacturing technology to rivals, chiefly Taiwan Semiconductor Manufacturing Co., and any further trouble could render Intel an also-ran for the indefinite future.
In a note to clients after Gelsinger’s hiring, Raymond James analyst Chris Caso said Intel doesn’t have time to deliberate.
“Unfortunately, in order for Intel to implement outsourcing by 2023, decisions need to be made yesterday. Gelsinger’s appointment notwithstanding, we would still view a failure for Intel to discuss a fully developed 2023 outsourcing plan on next week’s call to be a significant negative,” Caso wrote.
“We therefore don’t think Intel has the luxury of waiting until Gelsinger gets into the job to make an outsourcing decision,” Caso wrote. “If the company does wait, they risk falling irreversibly behind.”
An array of problems
Intel lured Gelsinger away from his current job running VMware
with a pay package worth more than $100 million. That’s evidently what it took to pull Gelsinger away from a thriving company and attempt to fix Intel’s problems.
Gelsinger, 59, spent the first 30 years of his career at the chipmaker. He was Intel’s first chief technology officer and one of its leading engineers and top Oregon executives when he left in 2009.
Speaking to employees Thursday, Gelsinger insisted that he’s returning to a company that has its best days “in front of it.” But he will be responsible for rebuilding a business that has lost its edge on multiple fronts:
- Competition: Intel rivals AMD and NVIDIA use TSMC’s factories and have capitalized on the technical advances in Taiwan to leapfrog Intel in key segments. Meanwhile, startups like Ampere – run by former Intel President Renée James – are opening new competitive fronts by developing new chips for the data center.
- Customers: Apple began shifting away from Intel chips last year for its vaunted Mac line of desktops and laptops in favor of processors Apple engineered itself. While Apple represents a relatively small share of Intel’s revenue, its M1 chips handily outperformed the Intel processors they replaced. That carries the implication other companies might be able to achieve the same thing and may go their own way, too. Microsoft, Amazon and Google are widely reported to be developing chips in-house for PCs or data centers.
- Culture: Intel’s manufacturing trouble has been accompanied by upheaval in the top ranks and the departure of respected engineers, from Intel veterans to highly touted newcomers.
- Investors: Intel shed $42 billion in market value on the August day it disclosed its 7nm chips were behind schedule. Under Swan, the outgoing CEO, Intel’s share price barely budged while the broad index of semiconductor stocks doubled.
“From a governance point of view, we cannot fathom how the boards who presided over Intel’s decline could have permitted management to fritter away the Company’s leading market position while simultaneously rewarding them handsomely with extravagant compensation packages; stakeholders will no longer tolerate such apparent abdications of duty,” New York investment firm Third Point wrote in an incendiary note to Intel’s board last month.
Third Point CEO Daniel Loeb called on Intel to consider whether it should sell off its factories altogether, as some analysts have long urged. Separating its research from its manufacturing could make Intel more nimble, ideally leaving it with engineering heft while allowing specialized factories to become a contract manufacturing powerhouse like TSMC.
In his remarks Thursday, Gelsinger said he will continue to integrate Intel’s research and manufacturing.
“When executed well, it has established Intel as a leader in every aspect,” he said. The company’s factories are “the power and the soul of the company,” Gelsinger said, but its business model “does need to be tweaked.”
What Intel has in mind instead, apparently, is some kind of hybrid model in which Intel would outsource only its most advanced chips while allowing time for its struggling factories to catch up and learn how to manufacture the new technology itself.
Its factory setbacks have left Intel choosing from among an array of bad options: whether to outsource its most valuable technology to a rival, to keep manufacturing in-house and hope for better results at its factories, or to dismantle the company.
Low yields
Intel’s latest crisis began in August, when
the company shocked investors by announcing it was a year behind in its forthcoming 7-nanometer chip technology. Intel has previously assured shareholders that its latest chips weren’t suffering the kind of setbacks that plagued its prior two generations of 14nm and 10nm processors.
For decades, semiconductor technology has advanced on a microscopic scale as features on computer chips have grown ever tinier – enabling manufactures to pack more transistors into the same space.
That enabled computer makers to improve performance exponentially while simultaneously reducing costs. It’s a virtuous cycle called Moore’s Law, named for Intel co-founder Gordon Moore.
Intel led that cycle for decades, with engineers in its Hillsboro research factories stubbornly defying the laws of physics as the features on their chips approached the atomic scale. Everyone knew that there would be a limit, someday, to just how small these features could get. But people had been predicting the end of Moore’s Law for years and Intel’s scientists had always proved them wrong.
Things began to go south several years ago with Intel’s 14nm chips. Problems
recurred with its 10nm technology, which arrived many years behind schedule, and once again with the forthcoming 7nm processors. In each case the company suffered low “yields,” meaning that many of the chips that came off its production line had defects that rendered them useless.
Defects are inevitable when operating on scales so small that even a microscopic speck of dust will ruin an entire chip. In normal times, Intel would simply toss the bad ones out and sell the rest. Over time, the company would perfect the manufacturing process and reduce the number of defective chips and improve its profitability.
If there are too many defects, though, Intel has to discard too many chips to make a profit on the rest. For its most recent chips it’s taken Intel several years to get the process right.
Intel blamed delays on its current generation of 10nm chips on being too ambitious in adding new features. It promised a more manageable approach with the 7nm generation but hasn’t explained why development of those processors went haywire, too.
Whatever problems Intel is encountering, though, its rivals don’t seem to be running into the same roadblocks. TSMC has steadily moved ahead with each new generation of chip technology.
Pressing its advantage, TSMC said Thursday it will spend up to $28 billion to expand its production capacity --an astonishing 60% increase. Even if Intel can fix its factories, it may not be able to match what TSMC is investing in its own future.
‘A proven leader’
The cupboard isn’t bare, though. Intel indicated Wednesday that its 2020 sales will top $75 billion, up nearly 5% from last year and well ahead of its forecasts at the beginning of that year. PC demand was strong as more people work from home during the pandemic and the data center industry remains robust overall.
Intel introduced a slate of new processors for PCs this past week, new chips that it claims provide significant upgrades in performance and power efficiency.
And inside Intel’s Oregon factories, technicians report the company is busily installing new manufacturing tools. It’s not apparent to them that Intel is hedging its bets or preparing for a major upheaval.
There’s no chance that Intel will walk away from its Oregon investments, or make significant cuts anytime soon. The company is two years into construction on a multibillion-dollar expansion of its D1X research factory at its Ronler Acres campus near Hillsboro stadium.
To make investments on that scale pay off, Intel needs to keep those factories humming. But if Intel decouples its research from its leading-edge manufacturing it will inevitably diminish Oregon’s essential role within the company, which could lead to a long-term decline at its Washington County campuses.
Intel’s Thursday deadline for an outsourcing decision was self-imposed but not arbitrary. Sending that work to offshore contractors will require years of work to coordinate the transition and reserve manufacturing space. Intel indicated this past week that it needs the new chips on hand in 2023, whether it makes them itself or buys them from a foundry like TSMC or Samsung.
Investment analysts are split over whether Intel should keep making its own processors or send advanced manufacturing to Asia. But there is broad agreement that Gelsinger brings engineering and leadership skills to Intel that Swan, a finance professional, simply didn’t offer.
“Bob Swan, while a solid manager, was not the right person to lead a manufacturing turnaround at the company,” Susquehanna Financial Group analyst Christopher Rolland wrote in a note to clients Wednesday. “We applaud the board’s decision to bring back Gelsinger, a proven leader with real experience in chip architectures and manufacturing, to push the company in a new direction.”