Former Intel CEO Craig Barrett on saving Intel

[tomatoma]

Everybody can have easy 20:20 hindsight from the retirement corner pontificate about this or that. Once you've walked a ton of miles in any of the players shoes is it really that easy to separate what got you to where you are and what you need to do next that is so obvious now? Unless you've lived it deeply first hand to appreciate what drove the culture creation, the personalities and the decisions that enabled people to rise to their position of power to make those decisions, to easy to just state the obvious mistake.

The look back is always obvious, looking forward very hard sometimes to see what could be the real pivot changing direction.

I agree with you very much, sometimes you can look at the past
Let's look ahead

 

[Daniel Nenni]

But Didn't Craig kinda assist in ruin Intel during his 35 years? And now at 85 years old he is gonna fix Intel? He needs to spend some time in the semiconductor trenches to better understand how the foundry business works, absolutely. Do that and that and you can claim to have vision into the future. The past just wont get you there...

 

[DanX]

The interesting question looking forward is what happens beyond 14A - what is the the main driver as the economics of the next processes fall solely on TSMC, as Intel and Samsung fall away ? Seems to me the real drivers for greater density from future nodes are client-side inference and data center inference / training, meaning general processing is good enough as a supporting function to AI, today. Both of those are going to be driven by energy efficiency and memory requirements of the next generations of models.

No. Process node advantage is too big.

 

[tomatoma]

The interesting question looking forward is what happens beyond 14A - what is the the main driver as the economics of the next processes fall solely on TSMC, as Intel and Samsung fall away ? Seems to me the real drivers for greater density from future nodes are client-side inference and data center inference / training, meaning general processing is good enough as a supporting function to AI, today. Both of those are going to be driven by energy efficiency and memory requirements of the next generations of models.

The only thing I can say is, is it possible to endure the huge demand with only TSMC? What will happen in the absence of competition? Is
No matter how much you try to expand the improvement, it will take several years to actually start production and reach the customer's hands.
And if some kind of bubble demand decreases during that time, it will be a big burden for TSMC.
There is no absolute in the semiconductor industry, I think so
I don't think this AI bubble will last forever, and from a later history, the AI boom that is happening now may be treated as an initial investment.
After the bubble bursts, it may be a relatively gentle model change or demand.

 

[blueone]

But Didn't Craig kinda assist in ruin Intel during his 35 years? And now at 85 years old he is gonna fix Intel? He needs to spend some time in the semiconductor trenches to better understand how the foundry business works, absolutely. Do that and that and you can claim to have vision into the future. The past just wont get you there...

I don't think age has anything to do with it. Craig was and perhaps still is a brilliant materials scientist, I suspect he was in the 99th percentile in understanding what it really took to manufacture chips. But computer engineering didn't seem like a strong point for him. This has always been my concern about the IDM model for chip companies. I've never run into anyone who was brilliant in chip manufacturing and at computer science and computer engineering.

 

[KevinK]

Was he arrogant, or simply naive?

I watched the Otellini video last night - seems like even in 2012, Otellini was naive:
* Recognized that the mobile was now changing the dynamics of the semi industry but was still convinced Intel could enter the market.
* Highlighted Intel-branding as a key differentiator for entering mobile late with x86 (hah!)
* Repeated the mantra of an alleged Intel 5 year advantage in process when Intel was at 22nm and TSMC was just beginning early silicon production at 20nm. His quote was “the next big thing will be on Intel transistors”.

 

[hist78]

Recognized that the mobile was now changing the dynamics of the semi industry but was still convinced Intel could enter the market.

The first iPhone was released in June 2007, and by 2012 it had enjoyed several years of explosive growth. In this 2012 interview, Otellini mentioned that Intel had created several reference smartphone models for non-US markets. I believe he and Intel were trying to apply the same strategy that had helped Intel dominate the PC industry, but it was a problematic approach.

In the PC market, Intel’s OEM customers, such as Dell, HP, Lenovo, Asus, and Acer, used Intel’s reference designs to develop their own final products. While this approach could save time and money in bringing products to market, it also made it difficult for them to differentiate themselves from competitors. As a result, their profitability suffered; most have endured single digit net profit margins for years.

The smartphone industry, however, was filled with strong and capable players enjoying rapid growth and healthy profit margins by using either in-house designed processors (Apple, Samsung, Huawei) or those built by Qualcomm and MediaTek. Why would they give up their independence and profitability to follow Intel’s lead?

Source: https://www.bankmycell.com/blog/how...o date, the iPhone,the iPhone revenue in 2022.