Intel's biggest mistake, Gelsinger is WRONG

[Arthur Hanson]

That is the "Innovator's Dilemma". IBM really struggled with this ... as I believe Intel did also.

I worked at IBM for many years. I was one of the first people to use RS6000 around 1989/1990 for IBM chip hardware analysis. I got a bunch of crap about using the UNIX RS6000 instead of mainframe analysis capability -- because we were designing/supporting chips that went into mainframes. The Executive told me that I was undermining the product that was paying our salary. I told him that I was not going to do a long-term market analysis of the options every time I wanted to do some engineering analysis. (sure he wasn't happy about that comment)

I understand the dilemma that IBM (and Intel with CPU) had because they were making so much profit from the existing products.

I have little to no respect for IBM. I did a job for a super major corporation and thought I quoted a high price with lots of padding. At the completion of the job, the customer showed me the IBM quote at over 500% higher. Also working for a subcontractor, I quit when IBM stood by installing unsafe materials. I had to contact the city attorney as the inspectors were corrupt. The work was torn out. ATT provided the specs on the use of the materials as totally unsuitable and dangerous. I quit my job, rather than install dangerous materials.

 

[Tanj]

The inventor's dilemma explanation is too narrow. Zoom out a bit and you simply discover that IBM or Intel never were players. IBM never stopped making the best mainframes. But it made ECL in its own fab to make those, and by the time it retooled to CMOS it was too late. The microprocessors had germinated far away and blossomed in different fields. To say that IBM was facing a dilemma assumes those fields were ever theirs, but it is a big world and those fields were something new and different, with so many different aspects to be mastered that IBM was irrelevant. Indeed IBM did superbly well with POWER and Z series silicon to stay alive and profitable, crossing the chasms that were within their domain.

For Intel the same with mobile. The reason ARM succeeded wildly was the almost accidental coincidence of near bankruptcy forcing them to sell their IP for use embedded in SOCs just at the time when fabs were ramping up to enough volume to build SOCs for phones. Just like IBM with early CMOS, Intel was so far distant from embedded IP in SOCs that imagining it was an opportunity for them is nonsense. It grew separately. By the time Intel realized those were important there were again many, many aspects of the new markets to be mastered. Those businesses were new, not evolutions of Intel's business.

The one chasm Intel should have crossed and which hurt them in failing, was their first attempt at becoming a fab for 3rd parties over a decade ago. That exposed how lack of scale and inward-looking tools and practices made them unattractive as a foundry, but also should have set off alarms about how far behind they had fallen in so many aspects other than the silicon technology node. It was a chasm that surely should have been visible internally years before the fab failure made it visible outside.

But in general, last years winners will be overtaken from nowhere, not due to some chasm in their own business. Don't think of business like Chess, where all the possibilities are enclosed with the pieces you have. Think more like Go, where the winning moves are in empty spaces often far away, which take time to develop any relation to the earlier moves.

 

[nghanayem]

This is one of the reasons I was dismayed by Gelsinger's appointment as CEO. He is the high priest of x86 CPUs being the center of the universe. I do like the IFS initiative though, so I think he's a mixed bag strategy-wise.

The high priest of x86 thing is pretty self explanatory given what happened with the 486. I want to pick your brain on why you see that as a problem though? Do you think intel should be using ARM for the forest line? Or is it more so the case that you think he is too attached to CPUs and might miss the bus on the next xPU like they did with GPU?

 

[blueone]

The high priest of x86 thing is pretty self explanatory given what happened with the 486. I want to pick your brain on why you see that as a problem though?

There are many semiconductor businesses that have billions of dollars in revenue and gross margins, and Intel has a leadership position in how many of them besides client and server CPUs? None that I can think of. Intel is an "also ran" in several chip businesses, including GPUs, AI chips, FPGAs, networking interface chips (including smart NICs), datacenter ethernet switches (now cancelled, I think), edge computing chips (Xeon D, seriously?), to name a few obvious ones. Furthermore, IMO (which seems to be playing out), all of the biggest cloud computing vendors are going to be designing their own CPUs. Amazon has had a lot of success with Graviton, and the Microsoft CPU just announced today has been an open secret for a long time, Oracle's big investment in Ampere is almost like having in-house CPU development, and Google is probably going to join the fray and announce an in-house CPU too. I think it's actually embarrassing for Google that they haven't announced yet. Server CPUs could become a slowly growing or actually shrinking revenue business for Intel and AMD in not so many years. Then what?

Do you think intel should be using ARM for the forest line?

No.

Or is it more so the case that you think he is too attached to CPUs and might miss the bus on the next xPU like they did with GPU?

All indications are that this is exactly what is going to happen.

 

[nghanayem]

If that is so, why was Andy Grove such a success?

Some nice recordings I stumbled upon:

Chip Making's Virtual Museum - The Chip History Center

In video, Andy Grove talks about the importance of manufacturing in business today at the intel manufacturing excellence conference.

www.chiphistory.org

On The Evolution of semiconductor - The Chip History

In this interview, Andrew S. Grove addresses manufacturing’s evolution: the forks in the road and the lessons learned along the way.

www.chiphistory.org

 

[count]

If you read the innovators dilemma you'll realize the conclusion was that it is the change in business models enabled by technology rather than technology itself that causes the downfall of leading firms. If you look at it this way it's not ARM vs x86, it's the IDM vs Fabless business model.

 

[jms_embedded]

The one chasm Intel should have crossed and which hurt them in failing, was their first attempt at becoming a fab for 3rd parties over a decade ago.

Not to nitpick, but Intel Custom Foundry (2013 - 2018) was at least the company's second attempt in the foundry business.

Their first attempt was in 1981, according to the September 8, 1981 issue of Electronics:

Forging ahead in the foundry business

While one part of Intel knocks on doors for software, another has opened its door to outside chip designers, allowing them to take advantage of its design-automation and integrated-circuit-processing facilities. Though it has been providing the service for six months, the company is just making public the fact that it has set up a silicon foundry in Chandler, Ariz.

Intel is supplying customers with complete sets of design rules for its high-performance MOS and complementary-MOS processes. Also on offer is its scaled-down H-MOS II technology, and even more advanced processes like H-MOS Ill will be added as they become available. Computer-aided-design facilities for circuit layout and simulation will also be provided, as well as services for testing and packaging finished devices.

A manufacturer’s willingness to invite this kind of business usually indicates an overcapacity situation, and Intel’s case is no exception. Indeed, the company says it is setting aside about 10% of its manufacturing capacity for the foundry. However, mixed with this excess capacity is a sincere desire to tap into the foundry business, states manager Peter Jones. “We kicked around the idea when there wasn’t a spare wafer in the industry,” he says of the firm’s long-standing interest in such services.

Referring to more sophisticated customers, Jones explains that “at the system level, their knowledge is greater than ours, so if we don’t process their designs, we’ll lose that share of the business.” The market for the silicon foundry industry could grow from about $50 million to nearly a half billion dollars by 1985. It is all but certain that Intel will eventually enter the gate-array business as well, perhaps by next year. At present, it seems to be leaning toward a family of C-MOS arrays with high gate count.

— John G. Posa

 

[hist78]

Not to nitpick, but Intel Custom Foundry (2013 - 2018) was at least the company's second attempt in the foundry business.

Their first attempt was in 1981, according to the September 8, 1981 issue of Electronics:

Forging ahead in the foundry business

While one part of Intel knocks on doors for software, another has opened its door to outside chip designers, allowing them to take advantage of its design-automation and integrated-circuit-processing facilities. Though it has been providing the service for six months, the company is just making public the fact that it has set up a silicon foundry in Chandler, Ariz.

Intel is supplying customers with complete sets of design rules for its high-performance MOS and complementary-MOS processes. Also on offer is its scaled-down H-MOS II technology, and even more advanced processes like H-MOS Ill will be added as they become available. Computer-aided-design facilities for circuit layout and simulation will also be provided, as well as services for testing and packaging finished devices.

A manufacturer’s willingness to invite this kind of business usually indicates an overcapacity situation, and Intel’s case is no exception. Indeed, the company says it is setting aside about 10% of its manufacturing capacity for the foundry. However, mixed with this excess capacity is a sincere desire to tap into the foundry business, states manager Peter Jones. “We kicked around the idea when there wasn’t a spare wafer in the industry,” he says of the firm’s long-standing interest in such services.

Referring to more sophisticated customers, Jones explains that “at the system level, their knowledge is greater than ours, so if we don’t process their designs, we’ll lose that share of the business.” The market for the silicon foundry industry could grow from about $50 million to nearly a half billion dollars by 1985. It is all but certain that Intel will eventually enter the gate-array business as well, perhaps by next year. At present, it seems to be leaning toward a family of C-MOS arrays with high gate count.

— John G. Posa

Indeed, the company says it is setting aside about 10% of its manufacturing capacity for the foundry.

It's the virtue of the IDM. External customers are always nice to have but not "must" to have.

 

[benb]

I never read "Only the Paranoid Survive" and it has an audiobook now, so I downloaded it, and have listened to the first few chapters. So far, an interesting case study, the 1994 division error in the early Pentium chips. Intel was unprepared for that. The responses weren't paranoid. They might have just been lucky.

One thing that interests me is how the leader in semiconductors changes every so often. Nvidia replaced Intel, like Intel replaced Motorola, and Motorola replaced NEC. It is so predictable, in a way, that I don't think you can blame Intel leadership. They were just at the top of their S curve and had to play it that way. Perhaps Foundry services is a new S curve, and perhaps Intel can leap onto it as the old business declines.

 

[nghanayem]

I never read "Only the Paranoid Survive" and it has an audiobook now, so I downloaded it, and have listened to the first few chapters. So far, an interesting case study, the 1994 division error in the early Pentium chips. Intel was unprepared for that. The responses weren't paranoid. They might have just been lucky.

One thing that interests me is how the leader in semiconductors changes every so often. Nvidia replaced Intel, like Intel replaced Motorola, and Motorola replaced NEC. It is so predictable, in a way, that I don't think you can blame Intel leadership. They were just at the top of their S curve and had to play it that way.

Grove's thesis would argue that intel should have realized it was at the top of the S curve and that they should have found the next S curve and gone all-in to secure the next phase of growth. As for NVIDIA I don't really know what to think. I am not sure if they are still on their first S curve and that only now is the idea of a GPU is fully being realized. Or if NVIDIA is truly the master and they have gone through multiple S curves (3D graphics, Software defined renders, GPGPU, the vertically integrated DC, AI) without ever really missing a beat.

Perhaps Foundry services is a new S curve, and perhaps Intel can leap onto it as the old business declines.

At the very least Pat certainly thinks so. The man idolizes Grove and has called IDM 2.0 intel's second inflection point.

 

[Claudio DD]

Maybe the right kind of paranoia:
* Who's catching up with us and how are they doing it ?
* We've got to obsolete ourselves before other companies do.

was replaced with the wrong kind of paranoia ?
* How do we keep these margins and keep everyone else out of "our" market ?
* Everybody is trying to steal our IP, even our customers so don't let anybody know anything about our products, process, etc.

When you are a near-monopoly, everyone is "out to get you", but you can't let it turn into non-productive and dehabilitating "protect at all costs" rather than "grow new markets" strategy.

as a Friday afternoon post for the historical records:
the "near-monopoly" of Intel was not reached with, how to say, "totally legal" means.
The matter was settled with 1,25 bio$ payment to AMD in 2009.
Everyone can draw his/her conclusion but 1,25bio$ (in 2009) was certainly a quite large amount for a settlement.

 

[blueone]

as a Friday afternoon post for the historical records:
the "near-monopoly" of Intel was not reached with, how to say, "totally legal" means.
The matter was settled with 1,25 bio$ payment to AMD in 2009.
Everyone can draw his/her conclusion but 1,25bio$ (in 2009) was certainly a quite large amount for a settlement.

Intel's sales and marketing people definitely crossed the anti-trust line by paying customers not to use AMD CPUs.

I wish I had the presence of mind to buy AMD stock back then. Note the AMD share price in this article:

Intel to pay AMD $1.25 billion to settle lawsuits

Intel is paying rival Advanced Micro Devices $1.25 billion to squash a legal battle over Intel's sales tactics, a rift that led to antitrust charges against Intel in several countries.

www.nbcnews.com

Today it's trading at $178...