Intel's biggest Misses AI and Mobile?

[hist78]

Let’s not forget if IBM hadn’t picked 8086 x86 Intel might never have become Chipzilla and it could have been Zilog or Motorola.

But those past glories do not help Intel in 2024. Intel needs to reevaluate its IDM business model in today's world.

 

[hist78]

There were few business as lucrative as what Wintel had for twenty years. That situation and durations is why you have the Intel today with its trouble and challenge.

For intel the x86 monopoly as well as the profit margins were so obscene for so long the culture around both product and process and manufacturing innovation became so skewed. I could argue it is now so imprinted in the senior leaders, processs and procedures it will take a strong outside leader and an existential moment to drive change. Clearly the existential moment has come but not clear the leadership is changing inside.

"For intel the x86 monopoly as well as the profit margins were so obscene for so long the culture around both product and process and manufacturing innovation became so skewed. I could argue it is now so imprinted in the senior leaders, processs and procedures it will take a strong outside leader and an existential moment to drive change. Clearly the existential moment has come but not clear the leadership is changing inside."

Do you think that without changing the IDM business model, the changes and needed talents you mentioned will be possible at Intel?

In my opinion, a strong, visionary, and capable outsider leader may be less likely to join a troubled IDM like Intel. The CEO position at today’s Intel is incredibly challenging. He or she must navigate serious conflicts of interest between Intel’s design/product divisions and Intel Foundry. After managing so much internal turmoil, I am not sure an Intel CEO would have much time or energy left to engage with external customers.

 

[blueone]

I'll take a stab here -- you can learn a lot about a company's culture from the outside.

Intel isn't pivoting very fast (mobile, graphics, AI, data center, fabrication), there are interviews of with ex-Intel employees that tell a lot (Jim Keller's discussion a few years ago about a defeatist culture in the fabrication area was eye opening - "Moore's Law is dead"), and the amount of innovation actually making it into products has been historically low (for Intel) until recenlty (not perfect but improving).

Jim Keller's view is valuable. As for product innovation, one can easily argue - without knowing anything about Intel's culture - that if it isn't a general purpose CPU, Intel isn't committed to it. I actually hoped Gaudi would break the mold, but the jury is still out.

Corporate cultures tend to be defined at the top*, and the last two CEOs before Pat certainly weren't superstars. Pat has his hands full, and jury is still out there. (*though older companies tend to have mature processes that also influence culture over time).

Agreed, though IMO, knowing PG, he is unlikely to change Intel's culture for the positive.

It was interesting reading the marketing book on the early days of Intel (forgetting the name) that talked about Operation Crush and how the Intel organization worked under Andy Grove. Starting with Otellini until Pat, Intel didn't seem to be really hungry for winning new business.

I agree again, especially about the really hungry part. Bob Swan appeared to be interested in diversification, but certainly Gelsinger doesn't appear to be. Though I'll say again I completely agree that IFS must be the next frontier for Intel.

None of this precludes the awesomeness of the engineers, architectures, and techncial staff at Intel (or individual leaders). I think we're all certain they have a lot of superstar talent, and people really dedicated to the mission. But the management culture by and large over the past decade looks .. like it could do better.

Agreed.

 

[hist78]

Required reading for everyone!

How Intel Missed the iPhone : The XScale Era

Challenging the conventional narrative on Intel's smartphone miss

thechipletter.substack.com

Intel was lucky and made their luck. Clearly in the last 20 years a lot of things came to ruin them.

To rise again the first thing you need is the Board of Directors, the CEO and the senior leaders or the ELT as they are called to really realize where they are and what the they need to do. But that isn’t enough the ELT need to get into the details and force a culture pivot in addition to this strategy change!

Pat may have a strategy but it isn’t clear that Intel has the culture or execution mindset to execute competitively against Apple, Nvidia, AMD, Qualcomm or the super scalars on design /product. On the technology/ manufacturing side I think they have no clue to how behind they are on culture. There is no corner on architecture, tools or chemistry and physics but on culture and execution will be why they fail.

It is always as a great innovator said it is easy to connect the dots looking backwards. I’d say it’s less than 50:50 this IDM2.0 IFS is going to fly. As noted above it’s about culture and listening to Pat and watch his body language it ain’t looking good is it ?

It's a great article. There is one error though:

"One can imagine Otellini’s surprise at the iPhone’s launch in 2007. Steve Ballmer may have laughed at the iPhone soon after its launch, but almost everyone else knew how important it was.

There was perhaps a dawning realization in the months after the launch that Intel’s best chance of building a series of SoCs for future iPhones and their competitors - XScale’s cellphone products and the staff that had designed them - was heading out of Intel’s doors."

Actually Intel sold XScale division to Marvell and completed the transaction in 2006, well ahead of the first iPhone 2007 launch. It's not something like "was heading out of Intel’s doors". They already left in 2006!

 

[samwilde]

You can bet it took a long time to develop the iPhone. If Apple reached out to Intel to make the CPU it would have likely happened before Intel sold XScale.

Just the time alone to port Darwin to ARM and turn MacOS X into iOS could have easily been a couple years. Probably two years or more. If Apple contacted Intel it would have likely been in 2005 or earlier.

 

[hist78]

You can bet it took a long time to develop the iPhone. If Apple reached out to Intel to make the CPU it would have likely happened before Intel sold XScale.

Yes, the discussions or negotiations between Intel and Apple likely took place as early as 4 to 5 or 6 years before the launch of the first iPhone in 2007. This raises the question of whether one of the untold reasons Intel didn't secure or pursue the iPhone contract was its lack of interest in continuing with the ARM platform. It's possible that Intel had already decided to move away from ARM much earlier than its 2006 sale of the XScale division to Marvell.

 

[benb]

Paul Otellini got the job of Intel CEO in 2005. The first thing he did when he got control was to axe 10% of Intel's workforce to improve their profits and raise the stock price. That was when they sold XScale to Marvell.

I wanted to add a few comments here, because I think 2006 was the pivot point for Intel in many ways. They had both 8 inch and 12 inch fabs at this point, and put 100% of their efforts into 12 inch, and closed all the 8 inch fabs that year. There was a huge loss of culture, talent, and institutional experience. When Intel eliminated the operator layer of fab operations, which wasn't needed and didn't exist in 12 inch fabs, they lost something important, hands-on troubleshooting know-how, which you might trace to Intel's later troubles transitioning from 14nm to 10nm.

Contrast TSMC. TSMC kept many of the 8 inch fabs, spinning some into Vanguard. But they kept operating in any case. They found businesses for the 8 inch fabs, novel new businesses, that maintained positive cash flow, and kept the experienced 8 inch people engaged with the 12 inch people, which were usually at the same sites. The 8 inch experience of troubleshooting problems, the hands-on knowledge, can't be easily duplicated in highly automated 12 inch fabs. And those 8 inch fabs with their low costs remain the best way to produce certain commodity products for automobiles.

 

[samwilde]

For Intel back then, as an IDM, it made no sense to keep older fabs around. It was a better deal to just sell off the tools and use that money to help buy them a new leading edge fab. They probably could use the older node to make chipsets or something but would that even have been profitable? Eventually as the chipsets grew more complex the older fabs would have been useless anyways.

The switch to 300mm was with the 65nm process node. And Intel had no issues with either that or several processes afterwards. When Intel released 22nm tri-gate in 2012 they basically beat everyone else in the industry to the FinFET era.

Intel's fumbling, conceding density to TSMC, started around the 10nm process.

 

[benb]

The ability to keep the talented but less educated people employed at the same site for decades has a compounding effect. Intel still has some sites like that, but it's still basically a cut off at about 2006, that's about as far back as they go. And these people are indeed quite old, and getting ready to retire, so essentially irreplaceable.

 

[samwilde]

The more senior you are at Intel the more stock you get. Those are the people who run the company. A lot of them were interested in keeping the dividend going for as high and as long as possible. Delaying the factory upgrade as much as possible was one way of doing this.

But that is a gross oversimplification. You have to remember the initial EUV machines had abysmal productivity. The light sources were weak as hell so the exposure times needed to be higher. Intel was also a major customer of Japanese lithography gear and was not interested in moving to being an ASML exclusive client while they could still push upgrades to their Japanese machines. Which were much easier and cheaper to upgrade than ASML gear from what people at Intel told me.

So Intel tried pushing their DUV gear as far as possible with multiple patterning.

I think this was not a big deal all by itself but their 10nm process was just too big of a leap versus the earlier process. And lithography was not exactly the main issue from what I heard. I think the main issue was them using cobalt in the process to replace tungsten. But this time I heard that from people not in Intel. So I cannot be 100% sure of what the problem actually was.

This article pretty much explains what I heard about the 10nm issues.

https://www.eetimes.com/intels-10nm-node-past-present-and-future/

I am not a processes guy so a lot of this stuff just goes way over my head.