While 2024 was a banner year for some chip companies – notably Nvidia and Broadcom – Intel had a year it would rather forget. Another CEO was forced out, it lost market share, and its stock dropped 60% during a year that was strong for semiconductors. So much went wrong that some industry watchers proposed the unthinkable: a takeover by private equity.
How did this happen? Not overnight. Analysts who spoke with Network World said these problems have been brewing for decades, and multiple CEOs failed to fix them. Once a stable company, Intel has seen four CEO departures in 12 years: Paul Otellini (2005-2013), Brian Krzanich (2013-2018), Bob Swan (2019-2021), and Pat Gelsinger (2021-2024) have all been shown the door for different reasons.
As Intel languishes, a company that used to make graphics cards for video games (Nvidia) is now synonymous with artificial intelligence, and a competitor left for dead a decade ago now (AMD) has a market capitalization that’s more than double that of Intel. For Intel to regain the dominance it had 20 years ago would require a catastrophic foul up by both AMD and Nvidia – something not likely to happen with their current leadership teams. That said, there is room for growth, and here are the steps Intel needs to take.
Step 1: Get another technologist CEO
Just as AMD and Nvidia have electrical engineers as their CEOs, Intel needs one as well (and it had one in Gelsinger). Its next CEO needs to be able to talk shop, understand the technology at the deepest levels, and be business-savvy.
“That CEO pick is going to be important because they have to change around the culture. They have to be believable. They have to understand the market, and they have to understand semiconductors. That makes it a very short list,” said Alvin Nguyen, senior analyst with Forrester Research.
“When they get a new CEO, are they able to look at these technology inflection points that come up? And, you know there’s going to be something after AI, as well. Are they going to recognize those technology inflection points and make the right decision? I’d hate to see them get somebody who is just short-term thinking and misses out on that,” Nguyen said.
Step 2: Stay the course on IFS
Intel is on a multiyear mission to reinvent itself as a foundry service and compete with TSMC and Samsung. If there is a knock on Gelsinger, it’s that he didn’t make it abundantly clear to the board and to investors that this is a long-term project that will require time and investment before a payoff. The investment can pay off if the board and shareholders are patient, but pushing Gelsinger into retirement is not a sign of patience. And they have to be, argues Patrick Morehead, president of Moor Insights & Strategies.
“What Intel has to do is execute on the strategy that it has set out on for the last three years, and that’s to get competitive in foundry and get more competitive in products. And I think on the foundry side, what I’ve seen on the 18A process is good so far,” Morehead said.
The 18A semiconductor manufacturing technology uses a 1.8 nanometer process node. It would be well ahead of the best that TSMC can manage for the next two years, and it would give Intel a clear competitive advantage over TSMC and Samsung. “In the end, that’s what it’s all about. It’s getting 18A to the appropriate yield and defect density rates to be able to make a profitable business, not only supporting the Intel design company but also for external partners,” said Morehead.
Intel has six months to go before it hits high-volume manufacturing on 18A for internal products, and then another year before it goes into high-volume manufacturing for third parties.
Step 3: Split
Gelsinger was trying to create a foundry business as part of Intel. But Nathan Brookwood, research fellow with Insight 64, says the company needs to split itself in half and spin off the foundry business, just like AMD did in 2008. Up until 2008, AMD made its own chips. It then decided to sell off its foundries to an Abu Dhabi investment company under the name GlobalFoundries. Today, AMD is thriving, and GlobalFoundries is the third-largest fabrication company in the world, behind Samsung and TSMC.
“I’ve been talking to a few people over the last week or two, and the general belief is that the company’s too far gone to remain a single enterprise, and it’s going to have to be split” into a product company and a foundry, Brookwood said.
Step 4: Clean house
News of Intel’s plans to cut 10,000 jobs is not enough, argues Bob O’Donnell, president and chief analyst of TECHnalysis Research.
“From what I gathered in speaking with Intel folks, there were fiefdoms and projects going a number of different ways, and some of those were, I think, ultimately, some level of distraction. And because of that, you didn’t necessarily have everybody on board internally,” O’Donnell said. Gelsinger wasn’t aggressive enough in dealing with those distractions, so the board decided it needed to eliminate the distractions and get everybody focused on the core stuff, he said.
There is plenty of room to cut, Morehead argues, even if most of the employees of Intel are at the foundry, not in chip design. Still, he notes that Intel probably has twice as many people on the design side per revenue as AMD, maybe even more. “The company has to learn how to work more efficiently with less people,” he said.
And along with cuts must come process overhauls. “I worked at Fortune 500 tech companies for over 20 years, and I’ll tell you that just cutting heads does not make you more efficient. You have to redo the way that you work, right the processes that you have. This is just a fact,” said Morehead.
Step 5: Get an AI strategy in order
Intel’s messaging around AI has been unclear and unfocused. Is it Gaudi? GPU Max? Xeon? Altera? Meanwhile, AMD and Nvidia have well-known product strategies: Blackwell and Instinct, respectively. Intel needs a clear strategy because right now it doesn’t have one.
“As much as I liked the ‘AI everywhere’ pitch that they did during the event in April, they need short term success with the GPUs to show that they’ve got something that’s competitive in the market, while Nvidia pretty much owns it,” said Nguyen.
Brookwood argues that Intel has to make an even more basic decision. “Intel needs to figure out if AI is going to be the major transition for the next generation, the next decade, and then they have to figure out how to catch up. Or maybe they can find something beyond AI, or AI plus, that they can seize and pioneer in order to establish a new product with a better product position,” he said.
Step 6: Develop a coherent GPU strategy
If there is one area where Intel has repeatedly failed, it is in its attempts at a high-end GPU. Its low-end GPUs, embedded in desktop and mobile CPUs, do an adequate job. But the high-end discrete market dominated by Nvidia has escaped Intel for multiple attempts now, and Intel needs a good GPU story for its data center efforts.
Morehead notes that Intel and AMD both pursued supercomputer contracts with the national labs (Argonne, Lawrence Livermore, and Oak Ridge) run by Department of Energy. AMD won multiple contracts over Intel because it had high precision, high floating point designs in its Instinct line, to which Intel had no equivalent.
“I do think that’s a failure not only to understand what your customers want, which is a highly programmable, low precision, high TOPS GPU for training, but also potentially a testament to resource allocations and speed of execution,” he said.
However, that may change later this year with the release of Falcon Shores, where Intel will merge its Gaudi AI accelerator with the GPU technology of Ponte Vecchio. But it really needs to perform if Intel is to keep up with even AMD, let alone Nvidia.
Source:: Network World