Slashdot Mirror
AMD Gives Up Its Share In GlobalFoundries
MrSeb writes "Three years ago today, AMD spun off its fab division, in a move the company claimed would allow it to more effectively leverage its assets, inject new capital into the foundry side of the business, and make it more competitive vis-à-vis Chipzilla. Today, that dream is dead. AMD announced today that it would give up its 8.8% equity stake in the company. When AMD created GlobalFoundries in 2009, the company held a 34.2% share in the foundry. The main thing that AMD gains from this deal is manufacturing flexibility. Previously, Sunnyvale had agreed to manufacture 28nm APUs solely with GlobalFoundries. This new agreement voids that arrangement, freeing AMD to work with TSMC and other foundries.. It's not an agreement that came cheap, though — not only is AMD giving up its 8.8% equity share of GF, it's agreed to pay the manufacturer some $425 million by the end of Q1 2013. AMD will take a $703M charge against the transaction. It's unclear how this move will pan out. We know AMD killed Krishna/Wichita due to manufacturing problems, Llano limped along for most of 2011, and GF's problems at 32nm impacted AMD's ability to sell 45nm chips into the channel. From a macroeconomic perspective, AMD is simply transferring its business to a foundry partner that's more able to meet its needs. One could argue that AMD's decision to get out of the foundry business is a logical extension of new-CEO Rory Read's plan to de-emphasize cutting-edge silicon in favor of SoCs. Time will tell."
AMD is dead.
Long live AMD!
It was a smart (but expensive) strategic move for chasing the cutting edge, but if their business plan is to leave the cutting edge behind, then I fear we lost one of the biggest drivers of progress. Intel might have the technology, but AMD gives them the incentive to keep running with it.
while(1) attack(People.Sandy);
After some eyebrow knitting, my best guess is "Systems on a Chip"? Eschew obfuscation, expand jargon abbreviations.
[
It's like... Ford decides to spin off its auto business so it isn't tied down to one manufacturer, and can then produce Ford's at Chevy and Dodge and even Honda plants. Why does it seem like someone decided the AMD brand was more valuable than its product? Does this help the consumer in any way to separate brand from product?
The Admin and the Engineer
thats expensive, but we must aplause for amd
From a macroeconomic perspective? In an article about one company's decisions about inputs and savings? Oh, really?
It's like... Ford decides to spin off its auto business so it isn't tied down to one manufacturer, and can then produce Ford's at Chevy and Dodge and even Honda plants.
Chevy, Dodge and Honda are competitors of Ford, AMD isn't doing this to be able to manufacture chips at a foundry owned by Intel.
Why does it seem like someone decided the AMD brand was more valuable than its product?
Because you don't understand what's happening. This move enables AMD to build chips at any foundry, in fact it means they can use the best foundry rather than being tied to an underperforming one thus resulting in a better product.
I think some of you are forgetting about smaller form factors including mobile. This might be a genius move in the end.
Pack it up. AMD is dead.
Tell that do my 4x12 6100s.
Intel is faster per thread on the top end. But, bu the time you have 32/48/64 you're obviously running a parallelizable workload. AMD gives way more bang for the buck.
Speakingof which, they just bought SeaMicro. Apparently they can "cram" 512 atoms or 1024 cores into 10u drawing about 5.5kW.
Well, 5.5Kw will get you 6 1U 4x64 6200 servers, which has 1536 rather faster cores and fits in only 6U, with the same power draw.
I'll bet the SeaMicro box is much, much more expensive too.
Not sure how the SeaMicro numbers add up, TBH.
SJW n. One who posts facts.
I hope they know what they're doing because I for one do not look forward to a PC marketplace dominated by only Intel and Nvidia.
Intel is faster per thread on the mid range too. What AMD has is they're cheap.
I'm starting to think GNU is the problem with "GNU/Linux" these days.
Not so much dead, but during times of allocation, they'll find it much tougher to service their regular customers. W/ their own fabs, they at least had a committed capacity that would enable them to service their priority customers. Now that they're totally fabless, fabs like TSMC would be @ liberty to support them only when they have reliable forecasts, but drop them whenever the market is overheated.
After all, AMD's processors are far lower margin than a lot of the other processors that get fabbed from various places. So when there is a market shortage, fabs would prioritize on their high margin chips, and I'll bet that AMD's is not amongst the top there. And if AMD has trouble supporting its top customers, not only will it lose them to Intel, but price would be their only leverage in getting them back.
As for Global Foundries, maybe they could start fabbing FPGAs and other chips from other vendors.
Not anymore. Once you have no dedicated manufacturing capacity of your own, then during times of general market chip shortage, the fabs will go for the higher margin chips, which AMD can't be since they're selling them @ near break-even prices. The only advantage for AMD here is that that during times of inventory glut, they are under no pressure to keep ordering to keep the lines running, but downside even to that is that fabs would give them a lower priority when the market returns.
AMD has been here before. More than once. And somehow has managed to escape oblivion each time.
AMD is still stuck in the manufacturing mentality of the 1980's where "do whatever it takes no matter what" was the mantra. Their former SDC, aka Fab 23, was full of these people, mostly from MMI, who tended to reject new ways of doing things and using better equipment and practices. They wasted millions on equipment that would be installed, qualified and then promptly destroyed by bad, primitive, caveman maintenance practices which also had the effect of ruining processes, resulting in drastically low yields and contaminated product that failed early and (not too) often on their customers. They fell short of production goals time and time again. They stopped telling their own employees what the cumulative yield was so that no one would know how bad things really were.
Intel improved their practices and processes from the start and retrained all their "cavemen" and those who could/would not be trained and continued to use clubs to maintain equipment were sacked. Intel, unlike AMD is focused more on quality rather than quantity, which improves yields which, by some magic that AMD has yet to grasp, increases quantity.
There is nothing here that isn't known within the industry. It's just not general public knowledge.
AMD has some excellent products but their manufacturing hasn't been so excellent. If they aren't willing to get rid of the cavemen and strictly enforce quality, then this may be their final act.
I don't know what it makes you if you don't even own SHARES in a fab....
6 servers with 4 cpus in each one, with 16 cores per server does not equal 1536 cores, its 384.
You're out by a factor of 4 there. AMD don't produce a 64 core CPU.
You can buy a 64 core system that contains 4 16 core 6200 opterons.
The benefit is that if they can't sell the volume required to justify the cost of upgrading GlobalFoundries to a smaller process they can use someone else who can already do it.
How the hell does one end up PAYING MONEY to give up a share? Usually one GETS PAID for such an action.
Back when they first announced their foundry spin off, I posted The Following:
1) AMD Spins off Fabs. ...
2) Intel/VIA/TMSC/IBM buys AMD Fabs.
3) Intel/VIA/TMSC/IBM Fabs charges huge price to manufacture AMD CPU's.
4) AMD CPU Prices skyrocket. Unable to find a cheap reliable FAB, AMD loses price competitive edge.
5) AMD Stock tanks.
6)
7) LOSS.
We are now currently at Step 2. Although I never would have known three years ago that Step 2 would turn out to be "Globalfoundries Buys AMD out of Fabs" but either way, here we are today.
Now, time to move on to Step 3...
In Soviet Russia, Trojan exploits YOU!
There are many advantages for AMD, they should have done this some time ago. Dedicated founderies with multiple customers can afford to front the billions required to keep on the leading edge of the feature size curve, and can't be mugged by Intel.
Have you got your LWN subscription yet?
AMD has seen the writing on the wall: there is very little incentive to spend the money required to further the state of the art in x86. Intel is slowing down its development pace on x86 and AMD is as well; there simply isn't much money in making faster x86 processors because they have already achieved sufficient speed for 95% of what 95% of consumers do with x86 CPUs 95% of the time.
What would be the point of sinking huge funds into becoming more competitive in a market that is going to become increasingly irrelevant going forward? Mobile devices are the trend and x86 does not compete there. Aside from Intel, which has momentum built up that will take a little while to wind down, x86 development is in the process of stagnating. It's quite clear when major x86 CPU announcements are now years apart instead of less than a year like they used to be. This trend will continue.
Hope you are satisfied with the current crop of i7 processors because x86 is not going to get significantly faster, at least not at the consumer level.
AMD will instead focus on trying to compete in a segment of the x86 market that may remain relevant over the long term: SoCs for embedded applications. I think it's a smart move because it's the market that AMD has the best chance of being competive in.
I predict that the fastest x86 CPU will ever be made will be no more than 50% faster than the current fastest Core i7. Intel's development dollar momentum will carry us through to that but nobody, including Intel, is going to be willing to invest significantly more in x86.
Divesting the company of its core competency is never a good idea. Making killer fabs was one of AMD's core competencies. I don't see a bright future for AMD with this move. And that's sad, because the CPU space is lacking competition as is. The last thing we need is to see AMD wither and die.
They are a node behind. TSMC just got their 28nm stuff out the door (they decided to skip the 32nm node and do only the 28nm half node). Products using it are on retail shelves, but only as of like a month ago. So where? Intel? Just about to launch 22nm for full retail availability. They are a node ahead, they are almost always a node ahead.
That would, by definition, make TSMC not cutting edge. If someone else is on newer technology than you, you aren't cutting edge. Not saying that is horrible or anything, just that it isn't cutting edge.
However I think it also perhaps demonstrates one of the things that makes Intel great is they invest a shitload of R&D in fabs. It isn't magic they are where they are in terms of being ahead, it is billions of dollars a year poured in to development.
As an example their 22nm is up and in full swing, however they've already been working on their new 14nm fab in Chandler. I don't mean working as in "playing around with the idea" I mean they are moving full steam ahead with the construction (extremely, extremely exacting construction), hiring people, etc.
The ARM heads love to go on about that but as of yet there are no ARM chips that compete in the desktop space. ARM chips start going out around the level Intel chips start coming in. What's more, they'd face a real uphill battle due to binary compatibility. It is just easier to run a chip that'll run all your old shit unmodified. ARM would have to offer some serious benefits to win people over.
All the "shopping around" in the world doesn't change the fact Intel has better fabs than anyone else. As I pointed out in another post TSMC just got 28nm fully online, retail parts on the market (AMD videocards mostly). That's the 32nm half node, they skipped over the 32nm node for some reason, which lead to delays for them. Ok fair enough but Intel has 22nm up and running full swing, retail availability coming shortly. They are, as usual, a node ahead.
Now this doesn't happen because the magic faeries like Intel or something, it happens because Intel pours billions in to fab R&D, and the own the fabs. That means that they get to reap the benefits and nobody else does.
I could totally buy the "Fabs? We don't need no stinking fabs!" argument as legit in terms of getting better tech for less money except it isn't the case. Intel has the most advanced fabs out there and they have them because they spend massive money building them.
This story strikes me as more gloomy for Global Foundries. AMD is effectively paying to get out of their stake in the company. Last I knew Global had ST and AMD for major customers only. Now with AMD obviously unhappy with the line yields and slow execution on advanced processing nodes we can only assume that they will at least in the short term be looking to TSMC. If Global is not able to quickly back fill with orders from somewhere else their cost situation is only going to get worse. The only bright point for Global I can see is the 2012 contract not being pay for good die only, something I have never heard of in any other supply agreement in the industry. (I have seen price breaks for yield dips or non-acceptance of yields below a certain point, but nothing like pay for good die only).
The e-350 was a very nice chip. Low power, but I could run vm ware under windows so I could run linux on a vm under win7 and still have 6- 8 hours of battery life in a 3.x lb laptop. AFAIK, atom cpus don't have hardware vm support or 64 bit support. While this summer's intel cpus will probably do better, I needed that last year, not this year. Sure, it'll play old games, but more importantly, it supported vmware, could be light, and had good battery life.
True, but problem w/ multiple foundries is that during times of high demands, they're not going to prioritize low margin CPUs from AMD. They'd also be less inclined to do process tweaks that may improve the yields w/ one customer, but impact their production w/ another.
A lot of people on this thread seem to take it for granted that the spin-off was a bad move. The fact is that only time will tell. Just because independent foundries in the past could not manufacture cutting-edge CPU's does not mean that it cannot be done. One thing we know from history is that this industry is never static (eg. at one time no one thought a fabless company would become so big, but now look at ARM).
GloFo has had its share of problems - as an independent foundry they are very young. But a lot of other fabless companies are rooting for them to succeed just so there is a cutting-edge or latest tech-node foundry option in the market. Intel may not be a direct competitor for the foundries but through AMD they are still a big competitor. If the foundries cannot compete with Intel then AMD falls. If that happens, the collective foundry industry loses a huge client. And one that can give them the high margin leading-edge business. Worldwide foundry capacity has increased rapidly (and is still increasing), and the capacity calculations obviously include AMD as a client. Everybody foundry company knows what is at stake. If they fail in competing with Intel, Intel will lap up the 'AMD vaccuum'. Trailing-edge mobile chips or console CPUs alone won't fill up worldwide foundry capacity.
GloFo's Germany fab was AMD just a few years ago. They were manufacturing leading-edge then, and will look to do so in the future too. ARM's processors are catching up, and in a few years they will need latest node technology. So this whole argument of a business 'not being vertically integrated' cannot succeed is a bunch of crap. There is no one single good business model. Business models change all the time (e.g Netflix or Dell when they started). Who knows, in a few years TSMC or GloFo may be able to release latest technology at around the same time as Intel, or even earlier. Might not happen for the next 3 or 4 technology nodes, but it can happen. Good leadership at AMD and the foundries can still make this business model succeed. This is NOT the business model of the old (with foundries making mostly older generation nodes). When AMD spun out GloFo, everybody in the industry understood that.So all the people claiming that AMD going fabless was a bad move or that foundry model cannot work are honestly clueless. AMD did what it did to survive. Rest of the industry knows it has to adapt to account for this, and they are obviously trying. Let's just hope they succeed.
I'm not sure how you can be a chip manufacturer without any foundries or production facilities whatsoever, but I think I speak for everyone when I say, "Good luck, AMD!!"
I know Wiki ain't the last word, but according to them, GF also fabs for Qualcomm and Broadcomm as well.
You're out by a factor of 4 there.
Not quite sure how I did that. Brain on idle I guess.
SJW n. One who posts facts.
they're not going to prioritize low margin CPUs from AMD.
As opposed to what? Which are the high margin ICs fabbed on the open market?
SJW n. One who posts facts.
For TSMC, I'd imagine that Apple's processors - the A5 and A6 - would have much higher margins than AMD's chips, despite being ARMs. Given that Apple gets premium prices for its iToys, I can easily see them signing service level agreements w/ TSMC that would trump just about any of TSMCs other customers - AMD included. TSMC's other customers include Qualcomm, Altera and Broadcom, and I'd imagine that they too would have more clout than AMD
Is my understanding of this correct? I think that's a really bad move. When you have to rely on someone else to manufacture your product, only bad things can happen. When it's something as complex as a CPU, the risk shoots up several orders of magnitude.
The desktop market hasn't shrunk a bit, it just isn't growing as fast as it was, nor nearly as fast as personal devices like smartphones.
New computer markets don't tend to kill off old ones. Like mainframes. Not only are they still around and sold, but there are more of them now than when they were the only computers you could get. Desktops out number them by many orders of magnitude but they didn't kill them. Nor did laptops kill desktops nor will smartphones kill laptops (and desktops).
Turns out that a smartphone and tablet are nice toys for surfing the net but when you need to get shit done, a regular old desktop and/or laptop are what you want.
You still got modded +5 Interesting with incorrect figures though