Intel CEO Says Company Has Fallen From 'Top 10' Semiconductor Firms, 'Too Late' To Catch Nvidia in AI (oregonlive.com) 51
Intel CEO Lip-Bu Tan told employees this week that the company has fallen out of the "top 10 semiconductor companies" and that it's "too late" to catch up with Nvidia in AI training technology.
The remarks came as Intel began laying off thousands of workers globally, including 529 in Oregon and several hundred others in California, Arizona and Israel. "Twenty, 30 years ago, we are really the leader," Tan said during a conversation broadcast to Intel employees worldwide. "Now I think the world has changed. We are not in the top 10 semiconductor companies." Tan said Nvidia's position in AI training is "too strong" and that customers are giving Intel failing grades.
Intel's market value has dropped to around $100 billion, roughly half its value from 18 months ago, while Nvidia briefly hit $4 trillion on Wednesday. Tan said Intel will instead focus on "edge" AI that operates directly on devices rather than centralized computers.
The remarks came as Intel began laying off thousands of workers globally, including 529 in Oregon and several hundred others in California, Arizona and Israel. "Twenty, 30 years ago, we are really the leader," Tan said during a conversation broadcast to Intel employees worldwide. "Now I think the world has changed. We are not in the top 10 semiconductor companies." Tan said Nvidia's position in AI training is "too strong" and that customers are giving Intel failing grades.
Intel's market value has dropped to around $100 billion, roughly half its value from 18 months ago, while Nvidia briefly hit $4 trillion on Wednesday. Tan said Intel will instead focus on "edge" AI that operates directly on devices rather than centralized computers.
CPU architecture doesn't matter (much) (Score:4, Insightful)
The x86 architecture is not the problem here. Architecture almost doesn't matter on the user side, we have compilers and operating systems for all of them. I hop on an Aarch64 (ARM) or an x86-64 server and I sometimes forget which I am on (thankfully uname tells me). The ecosystem is a solved problem.
It is important to have an architecture that your own and that you can control. AMD kind of puts a kink in that for Intel but there has always been plenty of room in the market. Looking at ARM, they are geniuses at licensing their architecture. They have a very different business from Intel and take far less risk under their model.
Things where Intel could have improved would have been aiming at the higher margin equipment in the data center. Improving their solution for multi node fabric. They just kind of settled on Omni-Path and InfiniBand, and stalled on keeping the development going. Then spun it off as a different business. You'll see this a lot with Intel; they'll initiate some technology, decide it's not part of their core business, and let it stagnate in an underfunded external business. Time and time again when given the opportunity to grow their business, Intel put its foot down and said: No, we're a CPU company!
I think Intel will be remember as a technologically successful company, but a poor business. With a self-defeating management hamstringing their growth repeatedly for decades.
Market Cap Worship (Score:2)
Re: (Score:2)
Which direction do I have to point at now when I pray for divine tech guidance?
I'd say towards Santa Clara (Nvidia) with a nod over your shoulder to Redmond (Microsoft). But stay tuned.
Re: (Score:2)
It's not just market cap. nVidia's net income (profit) is higher than Intel's entire revenue. nVidia's revenue is rising year on year, while Intel's is shrinking. And Intel had massive losses last year, unsustainably large losses.
Re: (Score:2)
your username is reminiscent of the FLCL cigarette "never knows best"
I assumed that would be pretty obscure reference, but I guess it struck a chord with lots of people since it even send to have merch:
https://www.google.com/search?... [google.com]
Re: (Score:2)
Has anything really changed? (Score:5, Interesting)
And will those "edge" AI devices have to sell for $1000 each in order for the engineers who design them to be rewarded and promoted within the company?
That has been Intel's doom for the past three decades. The employees working on high-end products with high margin sales get all the attention and promotions. The ones working on high-volume, low margin products go nowhere. That's why Intel has made so little headway in anything beyond enterprise products in recent years.
This new "focus" will go nowhere unless Intel's corporate culture is vastly different than what is was when I last dealt with it.
Re: (Score:2)
Re: (Score:2)
Yep, decades of arrogance and scams paid off! (Score:2)
I am sure they always had good quarters, until they did not. With some actually working strategic planning and some curbing of the greed, they would still be up there.
Re: (Score:2)
The suits got rich and that's all that matters.
Re: (Score:2)
Looks like it. Yes. Time to make sure these people cannot do so much damage anymore. It costs society wayyyy too much.
That ain't it, Chief (Score:2)
Tan said Intel will instead focus on "edge" AI that operates directly on devices rather than centralized computers.
This is precisely the wrong approach. No one wants AI on their device. It makes them expensive and slow.
Re: (Score:3)
No one wants AI on their device.
I do. I want AI on my device, or not at all. I absolutely do not want somebody else's computer to vacuum up as much data about me as they feel might be relevant, mull it all over, and re-process my data for that company's other business purposes -- or for my data to be handed over to some third party because the AI company is getting sued for copyright infringement. I would rather, and do, live with less and slower use of AI, and with only using models small enough to run on my own computer, instead of u
Re:That ain't it, Chief (Score:5, Informative)
Also, a large part of that device market is Apple. Apple designs their own chips and gets a large amount of capacity at TSMC to manufacture them. Intel failed to become Apple's foundry partner, which was ultimately due to Intel's inability to learn from TSMC. Intel can be a foundry for its own chips, but making someone else's is too hard an engineering and business problem for them to figure out. There is a lot of room in the foundry market for someone to take share away from TSMC, but Intel failed to capitalize on it.
Intel has made wise decisions NOT to compete in markets and focus on a single purpose. They got out of the memory market in the 1980's because they realized they couldn't compete with Japan. Why not? They knew how to make memory, but the Japanese companies made it better and cheaper. They took the same play with Apple and TSMC. It shows that TSMC is smarter, better managed and way ahead of Intel in most ways that matter.
Re: (Score:2)
TSMC is ahead of Intel in every way that matters for a fab.
Intel has failed at 18A, which we know because they claimed they were going to offer foundry services with that node and aren't. The only plausible reason for why is that it wouldn't be cost effective, either due to how much support they would have to provide, how poor yields are, or both. Now we're supposed to wait for their next process step to see how great they are.
Well, they aren't, and they haven't been for a long time. Intel's only secret sau
Re: (Score:2)
They did get all of the ASML EUV machines they could in 2024 and expect those to be running 14A process by 2027. 14A may be too late for Intel, and if it fails like 18A then yeah what's next for their foundry business?
Re: (Score:2)
Yeah, people really hate translators, image search, and things that make their selfies look good.
Re: (Score:2)
For robotics you definitely (should) want it on the device. Might be a market.
Re: (Score:2)
Jensen might have actually been one of the few people who could have saved Intel from itself.
I think corporate success is rarely tied to top management. There are exceptions, but corporate success is usually a collective effort with leadership coming from multiple sources. The people at the top are figureheads, but its the collection of people around them that determine the outcomes. Bad top leadership can do a lot of damage but creating success not so much. Intel is simply following the normal pattern of all living things. It got old and didn't keep up with the times, falling back on its earlier s
Re: Intel patent settlement (Score:2)
A new CEO would of course have brought his own team with him.
Re: (Score:2)
A new CEO would of course have brought his own team with him.
A new CEO would have brought, at most, a handful of people to a company that has over a hundred thousand employees with a shared culture and literally thousands of leaders within the group.
If any company deserves to fail... (Score:4, Insightful)
Time for New Leadership! (Score:2)
Sounds like a quitter! Nobody like a quitter! Time to roll up your sleeves Lip-Bu and put in the long hours!
great job CEO (Score:4, Insightful)
This is a great message for the CEO to give to investors. I'm inspired to throw some retirement money at intel stock. Nope. nope;
Intel is a sleeping giant (Score:1)
MBA's are the problem (Score:3)
Re:MBA's are the problem (Score:5, Informative)
It's easy to blame MBA's generally when in reality they have lacked effective leadership regardless of the leader's backgrounds for many years. Good leaders who can foster the right kind of high end high volume innovation in tech are actually extremely rare, and this has very little to do with MBA vs non-MBA. MBA is a generic title (kind of like Woke is used to describe movies when you can't actually articulate what is really wrong with a movie) for people who don't understand how a leader impacts the company. Paul Otellini got his MBA long before he headded intel and he undeniably took Intel through it's greatest growth period pulling intel out of the lull that it was in during the end of the Barrett years where all they were capable of doing was pushing clock rates incrementally higher.
Yes it was the MBA that drove Intel to innovate again, ask yourself why that doesn't fit the narrative and then you get to the core of my point: It's not MBA vs non-MBA. CEOs are selected and chosen based on their history. The kind of question you need to ask yourself is what kind of behavioural change do you need to drive. Just because you're an engineer doesn't mean you can drive technical innovation, and likewise just because you're an economics major with an MBA doesn't mean you can only focus your company on the coming quarterly profit statement.
Intel's history is actually a textbook case for *not* simply throwing the word MBA around and looking more into details. That said I don't think Geisinger (who didn't have an MBA) was given a fair go at undoing the damage caused by Swan (a short term finance driven guy at heart with an MBA), but he undeniably ran the company during a period of absolute disastrous response to major quality control issues that developed under his leadership.
Tan... I honestly don't know what to make of the guy, no idea if he'll pull Intel out of the rut it is in, but he's a science guy in background.
Re: (Score:2)
They lacked effective leadership when the MBA's wrangled the rains of leadership. That was 2 decades ago. And they cared more about enriching themselves than building the company.
When has Intel innovated recently? They are at least 10 years behind NVidia who released CUDA almost 15 years ago. I said it then that NVidia would overtake Intel. I was laughed at, but I have witnesses.
I understand that MBA is a generic term, but it certainly applies here.
Intel couldn't even name their processors correctl
Re: (Score:2)
When has Intel innovated recently?
That is kind of my point. Intel for the past 5 years was *not* run by someone with an MBA or finance background. The term MBA doesn't define the ability to innovate.
I understand that MBA is a generic term, but it certainly applies here.
it doesn't. MBA doesn't mean anything. If you want to look at the damage Swan did, it wasn't his MBA, it was the fact that was at core a purely finance business guy. MBA is just a post grad title - amounting to little more than some glorified schmoozing at a business school. Engineers have MBAs. Accountants have MBAs. Lawyers have MBAs. Innovato
Re: (Score:2)
Re: (Score:2)
Smart because words are used correctly? No I'm not arguing semantics here. The problem with comments like yours is that you are very specifically vilifying the wrong thing. Actually the problem is you're vilifying anything at all. There is a use for CEOs who have a finance focus as well (just not at Intel - that wasn't what they needed at the time). But by saying "MBA" you're perpetuating the myth that the MBA degree is universally bad.
Again, no different then calling movies woke. I don't feel smart, but I
Not too late (Score:2)
Re:Not too late (Score:4, Insightful)
Nvidia is charging what the market will bear, which leaves plenty of room for slower but cheaper AI hardware.
No, it doesn't. Slower but cheaper isn't actually cheaper, because it means you need a bigger building and more electrical power to get the same result.
Re: Not too late (Score:2)
"What's stopping Intel from putting 128G/256G RAM into their comparatively slow GPUs?"
Lack of demand and high memory prices.
Large memory alone is okay for running the LLMs, but not for training.
Re: (Score:1)
Re: (Score:2)
I believe training is also memory bound.
It is from my understanding, but if you put it on lesser GPUs, then you'd be processing bound again. Also, the GPU has to support higher-bandwidth memory to move the needle there, not just have more VRAM, because as you quote the memory bandwidth is a factor.
Re: (Score:1)
Re: (Score:1)
I'm the leader (Score:2)
Intel once held a commanding lead and money was rolling in. Did they spend it on innovation or on stock buybacks ?
oh [calcalistech.com]
Maybe it's worth a buy (Score:2)
When buzzwords are supposed to drive sales (Score:2)
Anybody could see that such a business model dependent on fancy stickers and buzzwords would be destined to fail.