Wednesday, May 28, 2008

I Worship in the Temple of Christensen

"I Worship in the Temple of Christensen!"

Back when I was doing my MBA, my International Marketing professor (HarlemBoy, the nickname he choose for himself) used to say as a catchphrase: «I Worship in the Temple of Walmart». In that same term I was taking General Management classes (you may call those Advanced Strategy),and I loved that phrase so much, that I adapted it for myself, and began saying there: “I Worship in the Temple of Christensen”.

You see, that refers to Mr. Clayton M. Christensen, DBA (Doctor in Business Administration, a REAL PhD) and best-selling author. I actually read one of his books “The Innovator’s Solution” before I entered the MBA. That advanced knowledge, and my firm belief in Christensen got me an A+ in General Management class (the Holding Fast Case, in case you are curious). I wholeheartedly recommend reading his books and articles, you can become a convert too.

Every time Christensen talks, I listen. So, I was very surprised when Christensen talked about the « The New Economics of Semiconductor Manufacturing ». I read the article, and something got me thinking:

« Competition is shifting toward a new playing field. Now what matters is making a large variety of products, each product in small volumes and each perhaps for only a short time. Examples of these growing markets include cellphones and MP3 players, which are subject to trends in fashion. Then there are the thousands of chips that are increasingly finding their way into our homes, offices, automobiles—and into every nook and cranny of our lives.

You often hear executives in the semiconductor industry sighing for the next great vehicle for industry growth, like the PC in the 1990s and the minicomputer before that. Well, perhaps the next killer application won't be one thing but rather scores or hundreds of things, none of which require the raw performance that only the biggest, most technically advanced fabs can provide. Perhaps what the next wave of killer apps requires is a new business model, made possible by such things as TPS.

Throughout history, business models that reduced the minimum effective size of factories have transformed entire industries. Steelmaking was transformed by the minimill's ability to efficiently produce small batches of steel, business computing by a succession of ever-smaller machines starting from mainframes for payrolls and ultimately leading to the personal computer, and photographic film processing by fully automated one-hour film-processing machines, which were then replaced by digital photography. Because these transformations offered customers entirely new ways of doing things—rather than simply making the existing model work a bit better—we call them disruptions. The agents of disruption are invariably business models (although these models often come with a new technology wrapped inside).»

That got me thinking. Then I read this blogpost from Mr. Rahul Sood « AMD Breakup ». Here Mr. Sood argues that AMD should split the Fab Business from the µProc+Chipset+Video business. At first I thought “This is Madness! Madness, I tell you!” But then, I heard Christensen’s voice… I mean, I really heard his voice in this podcast: «Spectrum Podcast: Q&A With Harvard Business School's Clayton Christensen »

And here, The Light was made. Basically, One of Christensen’s points is that the semiconductor industry is going through a commoditization/decomoditization phase, things that before were differentiators become commodities, and what was a commodity before, becomes a differentiator (too confusing, read his book, he is a DBA, I am just an MBA).

Of particular interest to us is that, for the bulk of the market (about two nodes behind the state of the art) the capacity to design chips is becoming a commodity, because libraries of modular components are readily available, but now the capacity to produce small batches of wafers with very fast turnaround times is the differentiating factor, because the process in very interlocked, one may say proprietary. His advice to investors:

«Go around and buy a bunch of these fabs […] then as you get the [TPS] system going you can price your products 15~20% over the market price, because you are fast».

Also, remember that at 32nm, with 450mm wafers, there is a huuuuuuuge number of chips that can be manufactured per wafer. Christensen admits that there is a segment of the market (like state of the art X-86 µProcessors) that will need to stay in the forefront of Moore’s law, but for the first time in history, for the bulk of the market, we can go back a couple of nodes and Just Do It Right.

Please bear in mind that I am not the only one listening to Christensen. Many people in the investment community do. So, ¿What would I do if I were Hector Ruiz? Not exactly what the rumor mill is telling you. Rather, here is my recipe:

1.) Make a partnership with some investor (Chartered? TSMC? Some European Player? Some Chinese bit player with no access to such advanced technology?). Give them 45% of the Dresden Fabs right now, for a much needed cash investment. Assure them steady business from AMD (let’s say, for 10 years or more), and full ownership at a later date. Retain operational control.
2.) Do step (1) in a way that is compatible with the covenants in your patent agreements with Intel.
3.) Use the proceeds to finance your Fab in the US. When that FAB is Up-And-Running, transfer the rest of the shares to your partners.
4.) The Fab in the US will be your main Fab, will be your Fab for advanced µProcessors and GPUs (or Fusion, whatever that means). 32nm with 450mm wafers means plenty of chips. Probably, will allow AMD to serve 40% of the Advanced X86 Market.
5.) Let your partner in Germany manufacture yesteryear CPUs, GPUs and Chipsets.

That way the boys and girls at AMD can survive. Will there be anyone interested in grabbing those Fabs? YOU BET!

Hey Hector, if you need more details, you can hire me. Or Christensen. Or both of us! (or just leave a coment)

Salud! Howling2929

Some Clarifying notes:
· If you already heard the podcast, you know one of the companies exploring this is Intel, in FAB 17. They have so many Fabs that for the 32nm nodes and beyond, they will have excess capacity. That is part of the allure of Classmate PC, or Atom, increase the demand of TRANSISTORS (not µProcessors). As usual, Jon Sokes has a great article about it « Beyond the BlackBerry crowd: life in a post-32nm world ».
· Is AMD exploring this? We do not know. Besides, the application of the TPS in the Fabs at Dresden is a problem of the one who buys the Fabs, not AMD’s.
· The investors from Dubai got their 6% in AMD as an investment, in the same sense that my MBA was an investment. They are learning first-hand the Ins and Outs of the semiconductor business from a global player. If the company itself turns a profit later on or not is secondary. Please remember that not every day there is an opportunity to enter such an important company in such a big way, so if the heavens give you lemons, you make lemonade! I guess it would have been impossible for them to buy such a big stake in Intel, Philips, IBM or Samsung at that particular point in time.
· Chinese companies are sitting on 130µm technology (more or less, depending on the foundy). And European regulations are more lax towards China than the US’s, I bet many Chinese foundries would LOVE to get their hands on Dresden, even at a premium.
· Samsung acquiring AMD was my scenario of choice, now is in the second place. I meant to write an article about that, but there is no point now.
· I was thinking on doing an article about Intel’s AMT. Is not Active Management Technology, if that is what you think, but searching for it in the Net is not easy). Does the respectable audience want it? Let me know.


Eddie/Chicagrafo said...

What makes you think the guys from Dubai got a board member?

I checked,,51_306_571,00.html

and there was no indication of any board member ties to Dubai.

Otherwise, the guys from Dubai just lost their money, I think they just got duped

(This is me, eddie/chicagrafo, just not authenticated)

howling2929 said...

Sorry, my bad, Dubai did not get any seats on the board. I corrrected the article to reflect that, and left the comment to acknowledge the blunder. But, if latter on they get a chair, well you read it here first ;-)

Christine said...

Fans of Clayton Christensen will love the following programs he has done and they are still available on demand. Previews of both are below

SPARKS said...

“That way the boys and girls at AMD can survive. Will there be anyone interested in grabbing those Fabs? YOU BET!”

Not so fast.

The Dresden Fab is geared up for 65nM SOI production. Although a great process on paper, it has proven, in fact, to be more expensive and less than successful in practice. (Barcelona’s high thermal’s due directly to leakage issues.) Partially depleted Silicon On Insulator is far more expensive to manufacture than the more successful and practical Bulk Silicon process. Intel’s Hafnium gave the Bulk process a new and extended life.

Therefore, it is not as attractive as you think, unless of course someone got it for pennies on the dollar.

As far as AMD on 32nM and 450mm, this is a stretch. They haven’t worked the bugs out on 65nM yet, let alone 45nM which they have yet to demonstrate.

Further, there are three companies who are sharing the bulk of 450mm tooling R+D, Samsung, TSMC, and INTC. The agreement was reached about a month ago, AMD was conspicuously absent.

The cost of 450 tooling is so high that it may just take a few companies with very deep pockets to set it into motion. That’s a commitment in BIG dollars, NOW, for a 2010 investment, which AMD can ill afford in their current financial position.


Eddie said...

Sparks, to tell you the truth, I really didn't like Howling's article, but he led me to think about something that is increasingly concerning me:

I am an investor in Intel, I am ok with the performance of the company and the share price, but I wonder if it is truly a good investment: Looking at Nehalem, with its four cores and 8 thread capabilities, even assuming this time hyperthreading will work, I wonder whether the vast majority of consumer needs such super computing monster.

I should know, 'cos I've worked in supercomputing, this is more or less my conclusion: The only driver for more computing power is graphical realism, because accurate physics is avoided in games by game designer tricks, artificial intelligence is just to hard to do at the scales of game publishers, so, there only remains graphics.

Howling points to this deceleration of computing power demand as an opportunity to explore other business models such as low computing power processors that are designed much faster and produced very agile to satisfy shifting market dynamics, and I think that's an important idea to think about. Then, it follows that economies of scale may turn less important, and thus fabs such as AMD's may be sold for uprising agile companies.

But still, just like you do, I see far too many problems with that. To begin with, AMD is pretty much a classical processor company

SPARKS said...

Eddie- First, please allow me to say I agree with you on all points. Secondly, and more importantly, I believe you have hit key points in the future of computing. Don’t underestimate the power of your writing.

The past has been riddled with example of companies, individuals, and people who asked “what does the average guy need with all this power?” Even Gates himself didn’t think 32 bit was a necessary step when INTC transitioned to 386!

It isn’t the power, it is the evolution. You said it first, it’s AI.

Intelligent transportation vehicles will protect a bad driver from himself. They will not allow themselves to be stolen, and will bring fault checking to new heights. Ask yourself this. If the two shuttles we lost had an AI computer on board, is possible the outcome would have been different in both cases? Those DARPA cross country races INTC subsidizes every year are no joke. The military is very much aware of the power of AI.

Would you feel more secure if the wife and kids were in an automobile that would circumvent the multitude of hazards one faces on the road, such as collision avoidance, foul weather, GPS, etc? How about preemptive maintenance BEFORE a break down occurs?

Do you want to make millions? A truly integrated home with current wireless solution could run HVAC, sophisticated yard maintenance, intruder alarms with facial recognition, leak detection, CO levels, room sensors, fire detection, all in control by home AI. The principals can be applied to endless examples. Frankly, our P.C.’s are powerful enough to do today. We are just not smart enough to compile software to make it happen--------yet!

What we need, now that we are approaching the high computation power, is to have someone or a group to flawlessly write a sophisticated algorithm to give the machine LIFE. This is what we’re working for and this is why we need small faster and better, daily.

INTEL has got the power and resources to do it. They are poised to do it en mass. You cannot spell Artificial Intelligence, without i-n-t-e-l. This is the holy grail of the twenty first century, and it WILL happen as you so nearly pointed out.

“Good afternoon Eddie. You’ve received 3 emails and 4 phone calls. There are two bills that need to be paid. Your current checking account balance is $3500. Would like me to pay them? The lawn sensors are detecting some dry areas would you like me to turn on the sprinkler system?”

“Yes pay them, and water the lawn, thanks HAL”


SPARKS said...

If you guys think I was over estimating the cost of 450mm wafer production, read this, it ain't pretty.;jsessionid=LSWSTP1Q2FO02QSNDLRSKH0CJUNN2JVN?articleID=208402622


Chicagrafo said...

Sparks: I don't view AI the same way you do:

AI has all the benefits you mention, but more computing power is not a requirement.

More computing power helps because it enables a tiny bit of laziness in the development of software, but that's all.

What needs to happen for the process you describe is a radical transformation of the business of software. What you describe is not happening already only because there are too many impediments for users to use software in creative ways; to begin with, today it is impossible to connect pieces of software that individually do all of the things you are describing not due to incompetence of hte users, but because some of the software is proprietary, obfuscated to impede independents to do non-anticipated things, there are the severely broken systems of "copyrights" and patents.

I have confidence that the world is awakening to the fact that the excessive protection it grants to "idea traffickers" is inducing the opposite the protection wants, it impedes innovation and invention. On the other hand, the practical advantages of Free and Open Source software for both producers and users are such that in effect FOSS is displacing proprietary software, so, civilization is improving from that side, it just needs a little help from improved laws to allow the transformations to happen.

Coming back to Intel, Intel is in the position to benefit from monolithic monopolies in Software and Hardware, it is a beneficiary of the problems of the existing system of patents and "copyrights", in a world in which users could freely mix and match their software, they wouldn't have to buy computing power at any premium. Also, for users free to mix and match software, it would be more important to acquire very specific computing power, and the agility to produce those processors will become an essential competitive advantage, I think that's the root of the idea Howling tried to convey.

SPARKS said...


Brilliant, that was excellent.

So if I follow you correctly, the hardware is ready and if the hodgepodge of software could be brought together without the lawyers charging in, we then could have AI.

I one knew someone who worked with mainframes. He was as quiet and unobtrusive as anyone could get. He was a complete gentleman, kind and good natured, I shall never forget his kindness. God, rest his soul.

He once told me he DREAMED in code! This guy was a human compiler! There are times that I wish I had the genius to understand how to talk to the machine the way he did. If I did, I would work alone, damned the lawyers, and have the machine talk back to me.

After I got it all together, I’d let the lawyers sort it all out.


Chicagrafo said...

There is an observation I invite you to make: The successful Free/Open Source Software projects out there are successful because some of its users actually contribute back, either code, adaptations to other products, documentation or fellow user help. That way, through voluntary work they manage to have the best operating system, the best web server, and the second best virtualization out there, among others.

So, this tells you that the problems of AI would be within the reach of voluntary work if it were not because of pre-exisitng impediments in the form of inventions that are not inventions and "copyrights" that in reality are the opposite, prohibitions to copy (share, adapt, modify, document) independently.

And again, the computing power we take for granted today is more than enough for the interesting applications based in AI you mentioned.