Just finished "Only the Paranoid Survive", and have a few thoughts about current day Intel

[Xebec]

I finally finished Andy Grove's book "Only the Paranoid Survive". It was enlightening to me when it talked about Intel's culture around constructive criticism and providing forums for challenging ideas. I'm not sure how it was in practice as I never worked there - but some aspects of Intel's culture of the 1990s sounded very refreshing, and explains a lot of their success at the time...

The central theme of the book is about "10X inflection points" which is where market forces change drastically enough that a business needs to address it's core business completely or else suffer great losses. With this in mind, it appears there have been several 10X inflection points that Intel doesn't seem to have properly responded to at all:

- The Rise of Mobility (~2008+), Intel's response was too little too late, and they're still paying for this in terms of limited market TAM, and scale of in-house foundry services.

- "AI" (~ 2022+), Intel's response is confusing at best (multiple semi relevant product lines), and they're not leveraging their software strengths here like they should be. They're not leading anything here.

- The looming end of the cost part of Moore's Law ($/mm2 cost decreases appear to be flattening) - Intel flip flopping on IDM vs outsource is the key point here - IMO they would (have been/) be better off sticking to one or the other. By changing strategies multiple times they're diluting the value of either choice.

- The fading relevance of general purpose CPUs (https://semiwiki.com/artificial-int...-warning-foretold-todays-architectural-pivot/) - Intel gets a partial pass here as they have finalled delved into discrete GPUs, and they appear to be adding accelerators galore to their Xeon chips, but I don't think they're doing nearly enough given how many others are "making" instead of "buying" (Intel).

(I do credit Intel for "dealing" with the end of Dennard scaling though in 2006; Tick Tock, and a focus on perf/watt was a good pivot).

..

I know LBT is focusing on some basics right now - improving fiscal and engineering discipline, and customer management.. but I think it's the next steps (strategically dealing with these inflection points) that are going to make or break Intel. They appear to have a lot more work to do than they will have available workforce..




P.S. None of this was written by AI, all hallucinations, spelling errors, logical fallacies, and grammar failures are mine alone.

 

[XYang2023]

I finally finished Andy Grove's book "Only the Paranoid Survive". It was enlightening to me when it talked about Intel's culture around constructive criticism and providing forums for challenging ideas. I'm not sure how it was in practice as I never worked there - but some aspects of Intel's culture of the 1990s sounded very refreshing, and explains a lot of their success at the time...

The central theme of the book is about "10X inflection points" which is where market forces change drastically enough that a business needs to address it's core business completely or else suffer great losses. With this in mind, it appears there have been several 10X inflection points that Intel doesn't seem to have properly responded to at all:

- The Rise of Mobility (~2008+), Intel's response was too little too late, and they're still paying for this in terms of limited market TAM, and scale of in-house foundry services.

- "AI" (~ 2022+), Intel's response is confusing at best (multiple semi relevant product lines), and they're not leveraging their software strengths here like they should be. They're not leading anything here.

- The looming end of the cost part of Moore's Law ($/mm2 cost decreases appear to be flattening) - Intel flip flopping on IDM vs outsource is the key point here - IMO they would (have been/) be better off sticking to one or the other. By changing strategies multiple times they're diluting the value of either choice.

- The fading relevance of general purpose CPUs (https://semiwiki.com/artificial-int...-warning-foretold-todays-architectural-pivot/) - Intel gets a partial pass here as they have finalled delved into discrete GPUs, and they appear to be adding accelerators galore to their Xeon chips, but I don't think they're doing nearly enough given how many others are "making" instead of "buying" (Intel).

(I do credit Intel for "dealing" with the end of Dennard scaling though in 2006; Tick Tock, and a focus on perf/watt was a good pivot).

..

I know LBT is focusing on some basics right now - improving fiscal and engineering discipline, and customer management.. but I think it's the next steps (strategically dealing with these inflection points) that are going to make or break Intel. They appear to have a lot more work to do than they will have available workforce..




P.S. None of this was written by AI, all hallucinations, spelling errors, logical fallacies, and grammar failures are mine alone.

The strategy is AI which is obvious enough at this point.

 

[benb]

I read it recently too (there's an audiobook). While this book is certainly better than some corporate "How I Did Its" (thinking of Jack Welch's Straight from the Butt), it fails to address the titanic shift that was apparent even in the 90s, which was:
- The Stickiness of Foundry (1990s to present): Intel failed to perceive how Foundry would achieve huge economies of scale which would breach their technology moat eventually. In 2004 when Motorola, once the leading chip maker worldwide, spun out Freescale, and the fab operations continued but scaling development did not, Intel should have known, for certain, that IDM would constrict their options more and more over time. If that wasn't sufficient, then in 2009 when AMD spun out Global Foundries, that should have definitely prompted action. As Foundry gained huge economies of scale, IDM began to lose economies of scale, slowing investment in new fabs, slowing design innovation/new nodes (since both depended on new fabs). This resulted in loss of market share and eventually economic losses. While Intel did nothing, Samsung began foundry operations in 2005. Qualcomm produced their first Snapdragon chip with Samsung in 2007. Intel Custom Foundry (Intel's half-hearted Foundry 1.0) started operation in 2012.
- Intel Converted Apple (2006): This was a huge positive development for them, and may explain why they missed the other things in the 2000s (mobile, Foundry/Fab-Lite, etc).

A few other relevant points of history:
-INTC maximum price: $74.88, August 31, 2000. So Intel has been stuck in a rut for 25 years.
-Only The Paranoid Survive published: 1996. It was 25 years after Intel produced its first microprocessor (4004, in 1971). This was 9 years after the founding of TSMC (1987). TSMC first major customer (Apple) occurred in 2011 though.