IBM's Fragile Chip Business

[iPC]

Originally posted by gregorypierce
It doesn't matter how many customers you have - it only matters that those customers ask for a lot of volume. If you only had one customer and they asked for 50 million units a quarter, I doubt you'd care about (or have time for) other customers at that locale.

Please RTA and my post again. Maybe then you wuill understand the hybrid nature of the plant. As a previous reader posted, this article has squat-all to do with Apple. Someone probably did a search on google for "I.B.M. Fishkill" and found this articel... http://www.google.com/search?hl=en&ie=ISO-8859-1&q=I.B.M.+Fishkill&btnG=Google+Search Notice most of the 1st page is from this site...? I think that resolves the signifigance of this discussion.

 

[mainstreetmark]

"obligatory 'Man with Wafer' photo. "

Haha... no kidding. Those guys love to pose with those wafers. At the plant, they should have one of those lifesized photos, except with the head cut out so you could stick your own head in there and get your picture taken with it.

 

[SubGothius]

NO chip-fab manufacturing problems at IBM!

This article, as with others on the same subject, is about the economics of production, and terms like "production" and "efficiency" and "yield", as used in this particular context, do not refer to the technical fabrication of chips in the manufacturing process. These articles do not even hint or imply in any way that IBM is having any problems nor technical glitches whatsoever with the actual fabrication or manufacturing of chips, nor with engineering even newer-generation chips to produce.

Unlike Moto in days of yore, there have been no reports of any inability to design/engineer chips that work to intended specs, no reports of any inability to fabricate those chips to spec, no reports of any inability to fabricate those chips in usable quantity nor of unusably flawed chips coming off the line.

If anything, IBM at Fishkill can produce a lot more chips than promsed or even needed, certainly as many chips as IBM is contracted to make, up to the plant's full capacity, and therein lies the only so-called "problem".

Simply put, the plant has barely opened for business and started production, in the middle of a sour economy at that, and already the analysts are whingeing that IBM doesn't have enough clients contracted to keep the place cranking out chips 24/7/365 at 100% capacity. That's the only "production problem" -- not that IBM can't produce enough chips to meet demand, not that they can't make quality chips that perform as intended, not that they can't engineer advanced chips, nevermind anything about fatally flawed chips -- just that they currently don't have enough client orders to recoup IBM's investment as quickly as Wall Street would prefer.

FACTS:

• IBM can produce the kinds of chips they promise.
• IBM can produce the quantity of chips they promise.
• IBM can produce the quality of chips they promise.
• IBM can design and engineer the chip advancements they promise.
• None of these statements have been challenged in any way by recent articles about the economics of production at IBM.

Sheesh...

 

[blakespot]

It would seem that perhaps IBM's fabrication yield difficulties stem from the sacrifice of wafer after wafer of silicon for use in the obligatory "Man with wafer" PR photos. Has anyone done that math?


blakespot

 

[jaedreth]

Yield issues

Yield issues are *demand* issues, not the ability to supply demand.

IBM has a fantastic record, and has lots of clients to make chips for. However, IBM would like to produce far more volume that is being requested. Now that the 2GHz chips are such a hot item for G5, Apple may request a lot more supply, which will make IBM happy.

Jaedreth

 

[nitz]

Der Wafermann/Dönermann

Der Wafermann seht aus wie der Dönermann oder? Vielleicht sind sie verwandt?

 

[SubGothius]

Re: Yield issues

Originally posted by jaedreth
Yield issues are *demand* issues, not the ability to supply demand.

In the economic sense of "production", yes; however, in the technical manufacturing sense of "production", IIRC the term "yield" refers to what proportion of chips on a wafer are usable and stable at what clock speeds, e.g. 95% total yield per wafer (only 5% fatally flawed chips), with 25% yield testing stable at 2GHz, 30% yield at 1.8GHz, and 35% yield at 1.6GHz (not real-world figures, just made up to illustrate the concept). I suspect a lot of the misguided uproar over so-called "production problems" at IBM Fishkill stems from people simply mistaking the various articles' discussion of economic issues for being fabrication issues (using similar but unrelated terms) that, as a close reading of the articles in question reveals, plainly are not the actual case.

IBM has a fantastic record, and has lots of clients to make chips for. However, IBM would like to produce far more volume that is being requested. Now that the 2GHz chips are such a hot item for G5, Apple may request a lot more supply, which will make IBM happy.

Connecting the chip-fab sense of "yield" with the economic sense, this also explains why Apple may consider offering a dual-1.8GHz G5. That is, they underestimated demand for the dual-2GHz G5 and overestimated demand for the mono-1.8GHz G5, leaving a surplus of 1.8 chips, and insufficient 2.0 chips to meet initial demand quickly. They could just crank up their order for more 2.0 chips to meet demand, but that would exacerbate the surplus of 1.8s, as only some chips on each wafer will certify as stable at 2GHz, others at 1.8, etc. Rather than just throw away the surplus 1.8 chips and eating that cost, they can simply drive up demand for the 1.8s by offering them as duallies, thus drawing down the 1.8s' surplus and bringing production supply of both 2.0 and 1.8 chips more in line with actual demand -- that is, bringing chip-fab yields in line with optimal economic yields.

 

[NeXTCube]

Re: Der Wafermann/Dönermann

Originally posted by nitz
Der Wafermann seht aus wie der Dönermann oder? Vielleicht sind sie verwandt?

Entschuldigung. Ich weiss nicht, was der Dönermann ist.

(Nitz is saying that the Wafer man looks like the Dönermann, and is asking if they are related. Sadly, my knowledge of German isn't good enough to figure out what/who the Dönermann is. Of course, when I find out, I'm going to feel like a dumba**, since I probably already know, but don't know that I know!)

 

[edStar]

Didn't the analysts say Apple was dead in the water .. a couple of years back??

 

[Cappy]

Originally posted by edStar
Didn't the analysts say Apple was dead in the water .. a couple of years back??

It's all one big circle.

Analysts talk about the companies. Consumers talk about the analysts. The companies talk about the consumers. It's all good.

 

[IDTheft]

I talked to my psychic today and she said that Apple is to start regaining market share again with the new G5. IBM is going to have so much capacity due to a hyper efficient plant that they will practically be giving away the chips just to keep the plant running at half capacity.

The subsequent low price, high preformance of the G5 will drive enterprise, university and consumers to embrace it.

In fact, IBM is going to deliver the G5 3ghz ahead of schedule and then surprise everyone with a G6 June of 2004, running at a whopping 4Ghz eventually topping out at an insane 8Ghz.

As for the PB? Well, Apple will endow the next version with a revolutionary MRPU short for Mind Reading Processing Unit....Oh Lord, I need a life

 

[legion]

SubGothius has it right... the issue is a lack of demand for the chips from the customers. IBM can produce the amount easily. Think of it this way: in a normal retail business you have a certain cost of overhead for real estate etc, not enough customers buy the goods (sour market), you've still have to pay the overhead and that's where issues in demand come up and you end up with a losing business.

As for IBM's analysis of the market and setting up clients beforehand-- the planning stages in a company as large as IBM and shifting product goals takes place many years before lining up clients. This is why, in general, when you need to react quickly to market forces small businesses are the best (you don't need to be quite so predictive and your assets aren't tied up.) All IBM needs (much like the article states) is to make more right choices than wrong ones.

As for IBM vs Intel in computer-centrism, IBM is still the leader. The financial record of IBM is the backbone and even through ups and downs, it's always came out a leader. From a customer on the street basis, Intel looks like a giant, but it's not the same for the larger computer industry.

One thing about Fishkill is that IBM is planning on using 970s in their own machines (sans Apple ROM and OS.) It's another case where they are trying to increase the need for the chip because Apple is still a minor player and that's the only way to make production streams profitable.

As for who's going to be at 90nm first, Intel is already there. Dothan (the next Pentium M chip) has already been produced in large scale numbers for release to OEM customers. They are only waiting until September to place them in front of retail consumers in order not to sabatoge their price-point strategy. IBM has already publicly conceeded this.

Another note, since I've seen this posted, is that the R&D group, even if it's associated with semiconductors, isn't part of the semiconductor production business unit. This means that the production of chips for customers doesn't have to pay for the R&D, as some have asserted, and the R&D unit is profitable as it stands today. This division is very beneficial for IBM because they sell a lot of their tech to companies like Infineon and TSMC which allows them to make a profit and continue being a leader in the research field (without having to bank on outside customers.) This has been the case even before IBM started looking to become a general semiconductor producer.

 

[NeXTCube]

Infoworld's Tom Yager on the Fishkill plant

In a Tom Yager column a couple of issues ago, he talks about the Fishkill plant...

Honestly, I can’t convey the awe I felt when I toured the facility. It looked like a sci-fi movie set, or an artist’s animated mock-up of how chips might be made in 20 years. But as impressed as I was watching the robotic carts gliding on tracks overhead and platters spinning inside huge machines, I looked around the joint and saw money. Not the money IBM spent, although that sum is dizzying. I saw all the money this fab will make when it’s completely operational. IBM is building the Death Star of the semiconductor industry.

Yager, who is Infoworld's test center director, has become quite the Apple-booster lately. He's recently mentioned Apple favorably in the workstation business, the WWDC, and has even switched to using a 17" Albook.

 

[billyboy]

Originally posted by gregorypierce
It doesn't matter how many customers you have - it only matters that those customers ask for a lot of volume. If you only had one customer and they asked for 50 million units a quarter, I doubt you'd care about (or have time for) other customers at that locale.

The easiest way to commit suicide in business is rely on one customer. Fishkill wouldnt have got a dime in investment unless its risk was spread. The more customers they have the greater the chance of one or some of those companies expanding and dragging IBM orders up. Sounds like Apple´s in good company at IBM.

 

[Phil Of Mac]

Re: IBM's Fragile Chip Business

Originally posted by Macrumors
The article also contains the obligatory Man with Wafer photo.

Thank God! All good news comes with Man with Wafer photos!