AMD Considering Getting Out of Fabrication Business 229
mytrip writes "2007 has not been kind to AMD, but it's surprising to hear rumours that they might be considering outsourcing chip fabrication. Analysts are predicting that AMD will try to cut costs by moving some fabrication elements out of the company by early next year. 'One Citigroup analyst is predicting a "transformational move" that would result in AMD's lower-end CPUs being manufactured by a third party and possibly selling off part or all of its Dresden, Germany facility. Another report from Goldman Sachs outlines the investment firm's belief that the company will leave manufacturing completely in the hands of third parties.'"
What a Busines (Score:5, Funny)
Like Show Busines...
Re:What a Busines (Score:5, Funny)
Re:What a Busines (Score:4, Funny)
Busines? (Score:3, Informative)
You know, I even did the good little
I'm not trying to sound like a jerk, but come on editors, this is basic stuff here.
Re:Busines? (Score:5, Funny)
AMD Considerin' Gettin' Out o Fabricatin Biznes, Yo!
Re:Busines? (Score:5, Funny)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2, Funny)
I hate to say it... (Score:5, Insightful)
Why? Because the main reason that no one but AMD can curretnly compete is because of the hight cost of the fab's... If third party fabs, capable of producing transistors the size that Intel makes, start springing up around the world we will probably see other design companies come out of the woodwork and start producing innovative and competitive chip designs.
If Via, for example, could produce chips in a 65nm fab in reasonable volumes... they might compete for the laptop market.
It may not be the best move for AMD, but for the buying public it should encourage innovation and competition. Which ultimately benefits everyone.
Re: (Score:2)
From someone who adores FPGA, I would welcome the possibility of getting custom chips fabricated at a reasonable rate. I have some ideas I'd love to see in silicon, but the cost is just horrendous. Guess it'll sit in VHDL till this day comes.
Re: (Score:3, Interesting)
Re: (Score:2)
Re: (Score:2)
Re:I hate to say it... (Score:5, Interesting)
It may not be the best move for AMD, but for the buying public it should encourage innovation and competition. Which ultimately benefits everyone.
Don't even kid about it. It's a path that once taken will be very hard to revert for AMD. Before you know it they'll outsource the rest of their fab, then sell their design to someone, and all that will be left, is a patent troll.
Last time when we discussed AMD's poor financial performance, I critized a guy who said we should buy AMD to support them, or the future may be quite grim, with Intel (being de facto complete monopolist on the x86 market) raising prices and stagnating.
When I read THIS article, I gotta say, that fear makes me think more like this guy and I'm suddenly feeling the need to buy AMD chips for the hell of it. I know it's wrong.
I always suspected that if they continue performing badly, IBM could consider purchasing them and entering the market of x86 chips. Both companies have worked together for a long time and share lots of technologies, some fab and many processes and design decisions.
Thing is, I didn't expect AMD to begin falling apart by itself, by selling some of its fab business. If they continue trying to minimize their losses by destroying themselves in this way, soon no one will want to have anything with them at all.
What a sad fate.
Re:I hate to say it... (Score:4, Interesting)
It's called voting with your wallet. It's pretty much the only thing you as a single consumer can do to affect these large companies, the other being to spread the Word. Not only is it the right thing to do, it's your responsibility to consider who you are buying from.
Re: (Score:2)
AMD has no products to effectively compete and is bleeding money because they couldn't come to market quickly enough with an effective counter measure.
Ditto for ATI, under the AMD banner, same problem... they just haven't kept up with the competition.
From being an AMD
Re:I hate to say it... (Score:5, Interesting)
I just built a new AMD rig however for two reasons.
Firstly, At the very low-end price point, I found AMD still performed better for that price range. I bought a Brisbane dual-core proc for $59. It overclocks unfairly well, and the peformance I get out of it is insane given the price. I haven't dared really push it over the edge, but consider just the latest review off NewEgg.
"This chip's a little beast, I've got the combo running stable (prime 16 hrs) @ 3106 (9.5,x326 @ 533 htt 3x) on air! "
Again, we're talking a measly $59 USD.
Secondly, it seems AMD got Intel dead to rights on their anti-trust suit. Several vendors and partners have offered credible evidence, and Intel is claiming their IT department deleted all pertinent email that would be the nails in said case. Again, they sound guilty as sin. I will not financially support such a company.
Even if Intel offered slight performance increase for the money (which isn't the case here) I wouldn't buy their product.
Re: (Score:2)
The issue is: given the ongoing price cuts based on the AMD/Intel wars and the new Core Duo architecture, I would have willingly bought a faster CPU for my mobo or one with better virtualization support
Re: (Score:2)
Or, rather, it worked easily up until Core. Now that AMD is no longer top performance dog it's harder to justify a purchase beyond supporting a company flag.
Now you have to decide whether the increase in benefits of an Intel processor is worth, well, supporting Intel.
T
Re: (Score:2)
Corollary: either you have convictions, or you don't. Your actions demonstrate which is the case.
Uh huh... (Score:3, Informative)
AMD made sure that (of all things) NetBSD ran on the K8 architecture 18 months before engineering samples were available.
What do you want from them? They just bought ATI and frankly their software development team was a mess, AMD is just beginning to untangle that.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
I doubt they sell their designs. Xilinx seems to be doing ok as an IP firm. They certainly seem to sell plenty of FPGAs (odds are there will be at least one or two Spartan FPGAs in some product in your house/apartment right now). Should also note that a small little company call
Re: (Score:2)
Re: (Score:3, Insightful)
Most third party fabs run in the 20-35nm range where flexbility is higher and diffrent products can be released simultaneously.
I would prefer to see AMD stay in this and create an arm of their business that fabs and outsources fabbing to companies like SIS, NVIDIA etc who could benefit from larger dies
Re:I hate to say it... (Score:5, Informative)
In reality, there are already plenty of third-party fabs out there. For instance, TSMC. And they have a 65nm process and that's what ATI's new 2000 HD series is manufactured on. So AMD (which includes ATI) is already manufacturing a lot of chips through a third-party. Even more than that, the current lowest end AMD processors, the Geode family, which is being used for the OLPC is also already manufactured by a third-party.
The only contention in this story is that AMD will be moving more low-end manufacturing to third-parties. The highest-end CPU's really have to be manufactured by the company itself. Not only does AMD have to stay as close to the bleeding edge as possible but they also have to have control enough to add certain devices or change certain design constraints. The change in volume to a TSMC or other third-party manufacturer from moving over some of AMD's manufacuting would not affect their bottom line or cost very much at all.
In addition, there are plenty of companies making various chips for all kinds of purposes. The limiting factor for new entries into the general purpose processor business is not the fab technology . A company can find the few million to make the masks and start making runs but the number of engineers they would need to compete with a design from Intel or AMD is enormous and would take years. In addition, Via could make a chip at 65nm right now if they wanted to but they don't have the partners or the platforms or market for those chips so they're not going to do it.
So while I'm looking forward to the day when there can be lots of players in the high-performance CPU business, the day is not here yet and this rumor, even if it were true, would do almost nothing to bring it closer.
You forget patents (Score:2)
Both Intel and AMD seem to keep going through many wild gyrations that don't seem to make long term sense. For instance, both got into the mobile CPU business (Au1000 and XScale) and baled out.
However, the change to outsorcing fab does mak
Re: (Score:3, Interesting)
Re: (Score:2)
Assuming a 70W difference (more than what Intel claims [zdnet.com]) in TDP at 100% utilization (unlikely) and a 10c kWh (residential prices are lower, not to mention corporate rates), that works out to 50*0.001*24*365*0.1 = $43 for a year.
Even if you double that to account for extra air-con costs, you're still far from it. The cheapest core-duo is around 200$. The above calculations are for high-end CPU, which cost hundreds more.
In reality, CPU utilization average is generally nowhere near 100% over a whole y
Re: (Score:2)
Of course the highest end parts all have higher TDP's, but if you're going for raw performance and don't care about price, you'll be getting a high end Core 2 anyway, at least for the time being.
Re: (Score:2)
Re: (Score:2)
Re:I hate to say it... (Score:5, Insightful)
We need more competition, not less. If there was ever a time when subsidies were a good idea, this would be it.
Re: (Score:3, Insightful)
Yes, by all means use our tax dollars to support failure. That will really help!
Re: (Score:3, Interesting)
The processors that came out of the Digital's collapse (UltraSPARC V, K7 and K8) were great processors but paled in comparison, especially considering the relative power of other CPUs on the market at the time (200MHz in 1992!).
What else out there is both powerful and elegant? All I see around me are multicore monstrosities.
Re: (Score:2)
Re: (Score:2)
The Alpha cost huge money to produce and ran hotter than hell (I used to have a DEC Alphastation, whee. I hear you can get them for scrap costs now, but who wants one of those? They're antiques now) whereas Hammer is [relatively] inexpensive to make and has very low power consumptio
Re: (Score:3, Informative)
The market hasn't lost any major vendors yet...
A lot of people seem to be getting really bent out of shape considering that all that has really happened is that a few analysts have speculated that AMD might continue doing something that it has already been doing for a while- i.e. outsourcing more of its low end chips to third party fabs.
Re: (Score:3, Insightful)
You don't suppose that a few tax pounds might have made it quite unnec
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Capacity-limited AMD just BUILD a fab (Score:4, Informative)
Outside of the last couple quarters, AMD's biggest problem has been production capacity. As in, they can't make enough chips, their market share is artificially capped, and as big players like Dell sell more AMD chips others are having a hard time buying enough.
That is NOT a problem you solve by becoming fabless. The already have foundry deals with e.g. Chartered, simply to provide some flexible extra capacity. It CAN NOT replace their current capacity with foundry deals, much less expand it. Being Yet Another TSMC Customer is not how you maintain your position as a top cpu maker.
The way you solve a capacity problem is by building another fab, which is what AMD just did. They built a whole new fab abutting the existing fab in Dresden, to the tune of $billions. $Billions that comes largely in the form of debt. You can't undo that by selling the fab because like a car the equipment begins to depreciate immediately. The only way to recoup that investment is to build parts in that fab and sell them. Now some analyst is saying that AMD is going to dump the fab, abandon that investment as a wash, and essentially give up the ability to have more than a pitance of marketshare while still carrying all the debt for building the fab? That's a great way to shore up the financials!
Utterly. Retarded. Analyst.
But I repeat myself.
As long.. (Score:2)
Queue up years of a true Intel monopoly. (Score:5, Interesting)
Awesome news! Next up, Torvalds indicted on murder charges when a mailing list discussion gets so heated he sticks a pointer straight through a face? Netcraft confirmation of BSD's death? Ron Paul is assassinated as republicrats cheer in the streets?
Re: (Score:2)
Re:Queue up years of a true Intel monopoly. (Score:5, Informative)
Chartered Semi just signed another tech partnership with IBM, Samsung, Infineon, and Freescale. This one [charteredsemi.com] goes down to 32nm.
UMC and TI are working on 32nm together [digitimes.com], too.
Fujitsu, although not especially known for fabbing chips for third parties, is working on getting down to 32nm as well. They do some fabbing for others now.
In any case, this story at Fabtech [fabtech.org] gives a much more reasoned and insightful look at the issues. They says it's likely AMD will outsource lower-end CPUs and continue to outsourc emuch of the GPU business as ATI already did. They may ramp up more outsourced work to Chartered than they currently do, and may share some fab space at Dresden and in New York. That's a far cry from going fully fabless.
Re: (Score:2)
Re: (Score:2)
The PPro is alive and well in the form of Core 2, of course with some incremental changes every generation. P3 was a damn good chip considering performance per watt, which is why it was used as the basis for Pentium M, which in turn was further developed into Core. Meanwhile, the P4 was a damn stupid chip that should never have been released, if only for environmental reasons.
Re: (Score:2)
Works for NVIDIA (Score:5, Informative)
There are lots of companies who only do fabrication, just as there are many other fabless semiconductor companies. With process shrinks occuring as quickly as they are today, it makes a lot of sense to let someone else (or several other someone elses) deal with the cost of developing fab facilities capable of the latest and greatest process size.
Re: (Score:2)
If the two aspects of AMDs business aren't lining up then that's terrible but it seems like they were doing a good job for a while, fab5 and 6 were coming online when the 754 and 939 were doing well.
Re: (Score:3, Interesting)
Re: (Score:2)
These companies that run 65 or 45nm plants for fabless semiconductor companies can run them non-stop at full capacity and never have to worry about exess inventory, demand, et al. They just fill their quotas for various manufacturers.
The biggest problem these companies face is taking more orders than they are capable of producing therefore their clients face production delays. I'm sure you've h
Re: (Score:2)
AMD going fabless would just be another way to shoot them
Dumb question (Score:2)
Re: (Score:3, Interesting)
Re: (Score:2)
Not quite. Lotus has had significant input [wikipedia.org] in the design of the Tesla Roadster. They're not just assembling them. Other 'outsourcing' ventures by car companies (e.g. convertibles being built by the likes of Karmann) have been similar.
The Roadster was developed in collaboration with Lotus Cars.
Lotus supplied the basic chassis technology from its Lotus Elise. Tesla engineers designed a new chassis with this technolo
Re: (Score:3, Insightful)
This isn't entirely offtopic, because I'd say that that may be the approach AMD is taking here. Lots of companies can make silicon chips with ever smaller features, its just a matter of time and money, and AM
Re: (Score:2)
Re: (Score:2)
Okay, so let's say the plants are spun off as an independent entity. Well, this other company will still need to spend money building and upgrading plants, or else AMD will be stuck with a stagnant supplier. Where would this other company get the money to do this? From the money it charges AMD to manufacture the chips, which AMD gets fr
Few companies have fabs (Score:2, Interesting)
Re: (Score:2)
depends on the state of the fabs, and the market for the chips that fab is capable of making.
For example. There might be an after market for a chip, but mot enough of a market to spend a billion dollars on a fab. 200 million might be worth it.
AMD might cut the buyer a deal on the price, but in return reserve the right to some fab time at a reduced rate.
I am not surprised (Score:5, Informative)
-AMD is behind in the laptop market, which is growing at a staggering pace. -Intel has as extreme cash flow, and therefore more room for mistakes. -The marketing team at Intel has been doing a better job than its counter-part. -Intel is ahead of schedule. In the meantime, AMD is behind. -AMD recently purchased ATI. It is not necessarily a bad move, but it cost them tons of money. To make things worse, ATI is behind schedule and also behind its only competitor, nVidia, which means less money for AMD. -AMD shares are currently falling.
I can only hope that I am wrong but I would definitely not buy AMD shares today.
Re: (Score:2)
AMD isn't stealing market share, hand over fist, from Intel anymore... That, however, is an absolute world away from bankruptcy, which you claim. AMD is a closer second than they've ever been before, and doing extremely well. What's more, over the past few years of dominance they've made the same inroads with systems manufacturers that kept Intel artificially propped-up throughout the years of the P4 fiasco.
Re: (Score:2)
Are you serious or are you just trying to ignore the fact that I used the word "may" in bold? Nowhere am I claiming that AMD is about to go bankrupt. I am just saying that it is more possible now than it has been for a very long time.
AMD has a more significant share of the laptop market now than they did for basically the rest of their existance.
I never said anything else. You are missing the point, however. Yes, they have a
Re: (Score:2)
When Intel is using illegal anti-trust tactics, that certainly hurts you.
The sad thing is that even if AMD wins their case, the damage both in lost revenue and market share will be almost possible to replace. Everyone in the world knows the name Intel. By illegally leveraging themselves, they've guaranteed that most people will demand their products regardless.
The lesson here is that illegal tactics work. I
Re: (Score:2)
Re: (Score:2)
out sourcnig to cut costs never works. (Score:3, Insightful)
add to this many outsourcing companys don't have a very good understanding on your business and it's a recipe for failure. I work in an industry where out sourcing is common, and most of the time the contractors are hopeless.
Re: (Score:2)
"Remember, grunts -- your weapon was made by the lowest bidder!"
Maybe a little of topic... (Score:3, Funny)
Re: (Score:2)
The sad state of affaird (Score:5, Informative)
Very revealing (Score:2, Interesting)
This makes me question whether Goldman Sachs has any business "analyzing" tech companies like AMD in the first place. Out sourcing low-end fabs that can be done on larger equipment just makes sense, there are other companies that can do it cheaper than they can. Outsourcing high end, cutting edge, small scale fabrication on the other hand doesn't really
Re: (Score:3, Informative)
So while it's okay to doubt, I wouldn't bet too much against the top investment banks right now, because they fund much of the world's ind
Performance (Score:2)
In making a public case for AMD divesting its fab business, Goldman Sachs is speaking to two audiences: the stock buying public, and AMD executives. And Goldman is hop
But who started the rumors? (Score:5, Interesting)
Speculators speculate on money moving, so it's rather unsurprising that they'd suggest that the response {large company} would have to lackluster performance would be to spin off the cost centers and reorganize to maximize the synergy of the core competencies.
Now, it is beneficial to make sure you're only worried about the business you're in. A lot of companies in the 90s for instance had huge in-house IT departments despite IT not being the thing that makes them money. They'd have a lot less headaches if they'd subscribed to an IT service to take care of their needs there, freeing them up to worry about the thing they really sell. You wouldn't worry about that any more than you'd worry about a company purchasing paper instead of milling it themselves.
To my untrained eye, AMD appears to be in the business of selling microprocessors. The manufacture of those isn't an incidental part of the business (the manufacture of the tools to manufacture the chips would be such an outside activity), but a key layer in their vertical integration. Unless their numbers are really small, I can't see why it'd be cost effective to drop that.
Bad idea... (Score:3, Interesting)
However, if chips are their core business, then they should probably maintain at least some manufacturing capacity of their own just to be able to maintain control of their own destiny. It might be cheaper for a third party to make some of their chips now, but in just a short while I bet it becomes more expensive when the third parties realize that AMD can't make their own chips any more. What are they going to do? They will have AMD between a rock and a hard place. Besides, AMD has already had problems in the past with ramping up to production fast enough to satisfy demand, and more than one person mentioned potential availability concerns as one of the reasons Apple went to Intel instead of AMD.
If AMD does this, I hope they look to copy how Apple does things. As far as I understand it, Apple doesn't manufacture much of anything themselves any longer. Apple's core business is not "making computers" or "making ipods." No, Apple's core business revolves a lot around design, usability, etc. With that clearly understood, then it makes sense for Apple not to be a "manufacturer" (building computers and circuit boards from scratch like they used to).
I certainly hope this isn't a short pier that AMD will be taking a long walk on. Time will tell.
Re: (Score:2)
Just another rumour by just another analyst. (Score:5, Informative)
Will AMD go completely fabless? I highly doubt it. IMHO, top-end chips pretty much require in-house fabs. That extra 10% of control and 10% of benefit to tweaking a fab to your own specific needs and 10% benefit to setting your own time lines can make the difference between being competitive in the high-end and not. (Yeah, I'm making those numbers up, but you get the idea).
Sure, AMD is having a tough year, but hopefully things will get on track. When they do, having at least one in-house fab is pretty much crucial to being competitive in the top-end... and the top-end counts because the margins are incredible there.
The mid-range chips and lower end stuff can probably be pushed off to a 3rd party... and I think we might see something like that from AMD.
Debt carrying costs (Score:5, Insightful)
AMD's "tough" years are in part because as a company with its own fabs, it has massive fixed costs (and the interest expenses associated with it), which means that when cyclical demand trends downward, their numbers get destroyed by the high fixed costs. High fixed costs are irrelevant to huge market leaders, but the nimble competitor gets eaten up when things get painful.
OTOH, if one can move capital intensive projects off balance sheet, the company's financial reports improve, which can improve their bond rating and lower their interest costs on other areas.
Right now, AMD must focus on chip design, chip manufacturing, chip marketing, and financial maneuvering. Going fabless would let them focus on designing and selling chips, instead of manufacturing them and managing complicated financial operations to fund everything.
Whether they gain a competitive edge by owning the fabs is another question, and the only people that know that are inside of AMD. Whether the CEO and Board will ask them is another question, but AMD's internal guys know whether they are really good at manufacturing or not.
Intel carries "virtual" debt (Score:2)
If Intel were to sell $5 billion in bonds secured by the fabs, and then return that $5b to the shareholders, their economic position would be similar. The decision to carry debt or not versus give the cash to the shareholders is theoretically neutral. In evaluating the companies performance, one needs to compensate for the leverage difference. Basically, Intel chooses to
Anal-ysts (Score:5, Interesting)
These analysts don't know anything. They just want every business to cut costs and debt while still producing the most revenue, for the most short-term profits, even if trying to do so is a stupid strategy that wrecks the company. When was the last time any published equities analyst was right about some surprising transformation of an industry leader? If they understood business strategy, they'd be running one, or privately advising one on equity development. These are people who can't even hold a job speculating in the market, so they try to make it speculating on the market.
AMD would never be this stupid- core business (Score:2, Insightful)
Re: (Score:2)
Nowadays I'd agree that there partnerships with IBM and other fabrication companies makes their role redundant. So I suppose you could say that today their core business should be
Worked so well for dog food (Score:2)
Probably good for them... (Score:2)
I've been using AMD from Duron 700 -> Athlon 1200 -> Athlon 2000+ -> Athlon64 3500+
Outsourcing: it makes sense (Score:5, Interesting)
The semiconductor market is one of the most brutal markets in existence. Capital costs are through the roof, demand is unstable and hard to predict, and the margins are razor thin. AMD is doing itself a favor to extract as much of itself out of the market as possible and focus on design. Design and production are as different as night and day. Competency in one speaks little about competency in the other.
What AMD is gaining is mass market production above and beyond what they currently have. Do they have to pay a middleman a cut? Sure, but in return they are getting access to massive foundries that can produce on an industrial scale. The foundry doesn't care what runs through its lines, so long as something is running. The more they run, the cheaper it is. It isn't like they will just run AMD chips. They will run a whole pile of other chips that run on the same equipment. The result is that they can sink the massive capital costs that a modern day semiconductor factory costs and run enough volume to make it profitable. Short of becoming diving into the foundry business and running lines for other companies, AMD has no way of running the massive volume it takes to make justify the horrific capital costs that a cutting edge semiconductor foundry demands.
The semiconductor foundry business is a cut throat world to be in. Massive capital costs, low profit margins, and over capacity makes keeping a foundry running a full time struggle. AMD is doing itself a favor by doing what AMD does best. AMD designs good chips. AMD isn't a semiconductor foundry. The slightly higher costs in paying 'middlemen' is pittance compared to the horrific cost of dropping a multi-billion dollar foundry down every couple of years while at the same time selling and junking your old multi-billion dollar foundries.
Re: (Score:3, Insightful)
For one, it requires both design and production competency. AMD is very competent at both, as they have to be in order to be in the same ball park as Intel. Without AMD-owned equipment with AMD engineers tweaking the knobs, the already existant gap between AMD's and Intel's manufacturing would widen. As you say, the foundry companies don't care who runs through its lines, they have many other designs,
Re: (Score:2)
Re: (Score:2)
AMD cpu's are not really that far behind INTEL, and yet they cannot make a dime because Intel pushes the ASP down below costs because THEY (intel) has the high end market, and that is the ONLY place money is being made and the ONLY place AMD cannot compete.
The high volume high end parts make money and give bragging rights. (yes, boutique parts like 4x4 and v8 will loose money)
Re: (Score:2)
Re: (Score:2)
Besides there are plenty of strong chip fabs outside of China. IBM, TMSC and well Intel to name a few. Sony has a fab as well. There are probably dozen others.
Fabing wafers is not an unskilled labour job. So even if they did it in China it wouldn't be done by your traditional rice farmer in the field. It would have to be done by people with an education in electrical engineering,
Re: (Score:2)
Tom
Re:When it looks like things couldn't get any wors (Score:2)
The perfect time to start buying up stock.
Or a long term short on it.
As it reminds me of like when NorTel divested it's manufacturing, it was not long after they got into big trouble.
If AMD was smart, they would get a "true" quad core out there, then price it at $60. Go out in the true competitive spirit and give Intel a licking.
Re: (Score:2)
So AMD lost the limelight to Intel with the release of the Core and Core-2 architectures. That doesn't mean that AMD is going out of business though. In another year when everybody is buying AMD Barcelona and Phenom chip-sets, people