Intel Gets Start of Antitrust Backlash from OEMs

[enigma067]

Intel Gets Start of Antitrust Backlash from OEMs

By Erik Sherman | Jan 4, 2010

A recent announcement that Lenovo would use CPUs from AMD (AMD) in a couple of its ThinkPads rather than chips from Intel (INTC) is the beginning of the price the chip giant could end up paying for its alleged anticompetitive activities: OEM customers shifting their orders.

In two separate statements, Lenovo said that it would use AMD chips in the ThinkPad X1003e ultraportable as well as the 13-inch ThinkPad Edge series, which is aimed at small- to medium-sized businesses. This is the first time that the ThinkPad brand, originally owned by IBM, will have used non-Intel chips:

An ultraportable PC positioned between a notebook and a netbook, the ThinkPad X100e can be equipped with AMD’s Athlon Neo single-core and dual-core, as well as the Turion dual-core processors. The ThinkPad Edge model, the smallest of three offerings in this product family and targeted at small and midsize businesses, may be paired with dual-core AMD Turion and Athlon Neo processors. The 14-inch and 15-inch ThinkPad Edge versions will still be powered by Intel’s Core 2 Duo chips.

Before you say, “But those are the small systems,” remember that the smallest systems, like netbooks, are the ones whose sales are really growing. To put it differently, AMD may not be in the prestige machines, but they’re going into the ones that may get the greater volume sales.

Starting in mid-November, I began noting that the upshot of all the antitrust activity focused on Intel would be customer defections:

PC vendors get completely wary of being sucked into the investigatory void and start shifting a significant portion of their purchasing to AMD. Forget fines and forget legal fees. That’s going to be the real price tag for years of allegedly using money and influence to keep a competitor constrained, and it will be a number with a whole lot of zeros.

I think the Lenovo switch is the first sign of that real price tag. Who knows how large a card it will need to be to record all the potential long-term loss for short-term gain?

Image via stock.xchng user MeHere, site standard license.

http://industry.bnet.com/technology/10004584/intel-gets-start-of-antitrust-backlash-from-oems/

 

[JAYDEEJOHN]

When it comes to mobile, Ive already said enough. Time will tell.

 

[fazers_on_stun]

If you read the interview with Dirk I posted http://www.itp.net/578576-exclusiv [...] ceo-of-amd you'll see that AMD have plenty of inroads to make in the classic arenas, ie notebooks. Why spend $hundreds of millions developing an atom-like chip when it is easier to make money by actually competing in the classic arenas?

AMD have their own set of technology to build around, ie graphics tech. What possible benefits could they offer over intel in netbooks?

Isn't Bobcat version 2.0 going to be for the netbook market?

What I see is notebook vendors pairing up Arrandales with AMD mobility 5000 GPUs, and whatever nVidia comes up with for competition, for the higher-end.

 

[Guest]

Yes Bobcat is designed for around the netbook size market, but it will also probably find itself into low end notebooks.

It is only with Bobcat + HKMG that AMD will be ready for that market. Any attempts on the current architecture would be pointless, that's why there haven't been any attempts.

 

[verndewd]

Well,
This whole game has now changed.
And I don't know what it means.
The picture is now very different.


Jan 13, 2010

GlobalFoundries Integrates Chartered


GF now owns 11 fabs. They bought Chartered; and now that deal is a done deal. It's all different now. And I don't know how it works because this is GF doing it's own thing.

New York will now be known as "Fab 8".


http://www.semiconductor.net/article/444020-GlobalFoundries_Integrates_Chartered.php

wow so much for production constraints

i had condemned amd for this fab light move , but adding chartered is genius.

 

[Guest]

Understand that most of Chartered's fabs aren't right at the cutting edge. You aren't going to get 45nm stuff even, at least not SHP I think. In terms of AMD, Chartered's fabs are a bit obsolete.

However, older chipsets etc can still be made on them, and plenty of other business too. The most important thing I think is, globalfoundries has a huge bunch of real estate that can be upgraded if required. Upgrading fabs isnt cheap, but if AMD start making real cash, and start showing a real need for extra capacity, global can deliver it.

 

[verndewd]

i realize that but I think doug grose will make great choices.

 

[JAYDEEJOHN]

Theres still billions commited to upgrades, and thats for past GF fabs. This leaves them close to Intels total now.
And tho I realize Intel is pushing 450mm wafers, itll cost more, and Im sure AMD is thankful Intel is doing it instead of them, as by the time GF goes there, itll be bug free and cheaper.
Anyone have any dollar amount commitments for the new Chartered acquisition improvements?

 

[Guest]

450mm wafers look like the wrong bet now.

Let's say intel have 8 450mm fabs, all Global needs is 12 300mm fabs. I know it's not quite that simple, but you can see the point. Global already has the fabs to create the same wafer capacity on current tech whereas intel is looking at doing it without building new fabs.

Intel don't get it wrong much, but tbh they are at the stage now where any mistake could be fatal.

 

[Guest]

edit - before anybody lol's at that, I know the difference between 450mm and 300mm has nothing to do with fabs, and that moving to 450mm is to get an improvement on yields per wafer due to less/smaller 'corner' dies.

 

[B-Unit]

Heheh, love the new quote, but fatal?

 

[fazers_on_stun]

450mm wafers look like the wrong bet now.

Let's say intel have 8 450mm fabs, all Global needs is 12 300mm fabs. I know it's not quite that simple, but you can see the point. Global already has the fabs to create the same wafer capacity on current tech whereas intel is looking at doing it without building new fabs.

Intel don't get it wrong much, but tbh they are at the stage now where any mistake could be fatal.

I'd bet a lot of industry folks were saying the same thing when 200mm wafers were the norm 😛.

There are some advantages to 450mm (smaller percentage of wasted edge chips, higher throughput, less variation and better QC due to fewer batches, and ultimately lower cost per die) that make a good case for switching. Of course there are disadvantages such as a bad batch making for more waste, plus the capital expenditures needed, but when upgrading a fab they'd have to spend lots of $$ on new equipment anyway.

IIRC Intel bought a substantial stake in one of the fab equipment manufacturers, so if Intel succeeds in pushing the industry to 450mm, they'll make $$ off the deal. Including GF if they want to compete 😛.

I'd bet Intel has analyzed the pros & cons in great detail and can make an excellent business case for the switch. And given their track record, I'd bet on Intel's foresight long before agreeing with the - let's face it - relatively uninformed opinion here on THG...

 

[Guest]

Pretty much.

Intel have really difficult choices to make in order to stay at the top. They have always had a huge capacity lead over AMD. Everybody talks about better architecture etc, hardly anyone talks about the fact that intel can make up to 10x more cpu's than AMD can at any one time.

GF are seriously changing that by buying everybody elses fabs. Although those fabs aren't useful for AMD right now (for cpu's anyway), they could be useful in future.

Let's say intel spend $10bn upgrading their current fabs to 22nm. If GF spend $20bn on the same upgrades, they'll have twice the capacity. We are looking at 2014 probably, and I have no doubt intel will react in a big way long before then, but as it stands GF basically have more fabs than intel and better engineers than TSMC.

 

[Guest]

I'd bet Intel has analyzed the pros & cons in great detail and can make an excellent business case for the switch. And given their track record, I'd bet on Intel's foresight long before agreeing with the - let's face it - relatively uninformed opinion here on THG...

I said so in my followup.

Intel don't make serious mistakes at the process level, they've led on it for years now. They need to keep leading though or it will all come tumbling down. Intel is basically a company on the edge in every regard. They cannot afford to lose any advantage they have, so one mistake can really be 'fatal', especially now that others are out to bring them down.

 

[fazers_on_stun]

Pretty much.

Intel have really difficult choices to make in order to stay at the top. They have always had a huge capacity lead over AMD. Everybody talks about better architecture etc, hardly anyone talks about the fact that intel can make up to 10x more cpu's than AMD can at any one time.

GF are seriously changing that by buying everybody elses fabs. Although those fabs aren't useful for AMD right now (for cpu's anyway), they could be useful in future.

Let's say intel spend $10bn upgrading their current fabs to 22nm. If GF spend $20bn on the same upgrades, they'll have twice the capacity. We are looking at 2014 probably, and I have no doubt intel will react in a big way long before then, but as it stands GF basically have more fabs than intel and better engineers than TSMC.

IIRC when upgrading to a new node or process, much of the fab equipment is rendered useless and either gets put out to pasture making non-leading-edge products like chipsets, or sold to some company that makes non-leading-edge products like chipsets. In fact I have heard that often the entire infrastructure gets upgraded - new wiring, plumbing, and air filtration. So there is quite a lot of capital equipment that cannot get reused. I'd imagine GF bought Chartered mainly for its customer base rather than fab capabilities.

As for the unlimited upgrade funds, I saw a report on MSNMoney that Abu Dhabi lost a whopping $125B in 2009:

Slightly further afield, the mighty Abu Dhabi Investment Authority is reported to have lost $125 billion last year, according to economists at the Council on Foreign Relations. That would push Abu Dhabi as the second largest sovereign fund in the world, behind Saudi Arabia.

Though the size of Abu Dhabi’s fund has been pegged at around $627 billion by the Sovereign Wealth Fund Institute, CFR states that the actual number is closer to $328 billion, down from $453 billion a year earlier. In December, Abu Dhabi stepped in with a $10 billion loan to rescue its neighboring emirate Dubai from default.

Abu Dhabi has historically invested heavily in US Treasuries and holds positions in Hyatt (H), Citigroup (C), Ziopharm Oncology (ZIOP), Toll Brothers (TOL), and Advanced Micro Devices (AMD). In 2008, it started tending toward more exotic investments, such as private equity, hedge funds, emerging markets, as well as some less than obvious corporate stakes, such as Chicago Parking Meters, LLC. Abu Dhabi owns more than 25% of that company, which privatized Chicago’s 36,000 parking meters last year.

 

[someguy7]

If AMD could and did produce as much chips as Intel did a day jenny then they would go bankrupt. They have to be able to sell all those chips.

Is AMD struggling to meet cpu orders now? There is no demand for 10x the amount of cpus AMD currently produces. Just AMD having the ability to be able to ramp up the amount of chips it makes is going to mean nothing on its own.

 

[Reynod]

Yes there is no point in have surplus CPU / GPU stock you can't sell.

That is what panicked Intel into "adjusting" OEM's to get rid of their horrible Pentium 4 stockpile ... a few years back when the A64 hit the shelves.

 

[amdfangirl]

Intel should make it's own computers.

 

[jimmysmitty]

Intel should make it's own computers.

Everyone would scream anti-trust on them just like if Microsoft started their own in house PCs.

As for laptops, AMD hasn't had a decent notebook CPU for competition in a while. They look to have something but once Intels 32nm hits its going to be hard to overlook them for AMD since the battery life will go up on them.

The 450mm waffers are probably the way to go in the end. And Intel won't have to build new FABs. The last one they built in Chandler, AZ is advanced enough to swap out the equipment. I am pretty sure you will see AMD and GF follow suit.

When it comes to the actual process, Intel has always been a leader in which way to go. Back when Intel put out HKMG, a lot of AMD fans said it was useless but AMD and GF along with every other chip maker is going to use it in the near future (IBM and AMD have it set for 32nm but there are 9 other companies with them too).

If Intel is going 450mm in waffers then everyone will go. They have a lot of hard decisions but when it comes to the process node and such they never seem to have problems leading the industry.

 

[verndewd]

Intel should make it's own computers.

id rather build, thanks.

afaik 450mm isnt quite up to snuff and as far as chartered they also pal'ed up with ARM which is a win win for the entire group. GF clients with AMD having 30ish % stake is just a beautiful deal all around. ARM pretty much owns the cell phone world right now so this developement is interesting to say the least.

 

[pausert20]

IIRC when upgrading to a new node or process, much of the fab equipment is rendered useless and either gets put out to pasture making non-leading-edge products like chipsets, or sold to some company that makes non-leading-edge products like chipsets. In fact I have heard that often the entire infrastructure gets upgraded - new wiring, plumbing, and air filtration. So there is quite a lot of capital equipment that cannot get reused. I'd imagine GF bought Chartered mainly for its customer base rather than fab capabilities.

Faser, I do have some insight on the cost of outfitting a new fab from my friend that works at Intel.

The building shell costs peanuts compared to the actual Fab Tools that get installed. The part of the new Wiring, Plumbing and Air Filtration does cost a lot to any of us but in the big scheme of things when either building a new fab or upgrading one those cost are minuscule.

D1C just finished being updated to 32nm. They are now the current leading manufacturer of 32nm chips at Intel. D1D is also building 32nm chips but they are now working hard at bringing up 22nm for the next round and also doing experimental work at the next node.

I understand New Mexico is next to get changed over to 32nm at Intel's megafab 11x. Not sure who else is going to be tranformed next. Might be Israel, Ireland or Arizona.

 

[jimmysmitty]

id rather build, thanks.

afaik 450mm isnt quite up to snuff and as far as chartered they also pal'ed up with ARM which is a win win for the entire group. GF clients with AMD having 30ish % stake is just a beautiful deal all around. ARM pretty much owns the cell phone world right now so this developement is interesting to say the least.

ARM might have a bit of competition if moorestown can deliver Atom performance and power figures in a cell phone envelope. But I think 22nm will be better for that.

I understand New Mexico is next to get changed over to 32nm at Intel's megafab 11x. Not sure who else is going to be tranformed next. Might be Israel, Ireland or Arizona.

My guess would be Chandler. They tend to get a lot of the newer stuff after Oregon who gets first dibs. Plus the last FAB built there was pretty high tech, they would want to take advantage of that and save a bit of costs in upgrading.

 

[pausert20]

When it comes to the actual process, Intel has always been a leader in which way to go. Back when Intel put out HKMG, a lot of AMD fans said it was useless but AMD and GF along with every other chip maker is going to use it in the near future (IBM and AMD have it set for 32nm but there are 9 other companies with them too).

If Intel is going 450mm in waffers then everyone will go. They have a lot of hard decisions but when it comes to the process node and such they never seem to have problems leading the industry.

Intel was not first go immersion. They made the decision that they could get great results by staying dry immersion at the 45nm node. It turned out they were right.

They are using immersion lithography for the just released Westmere cores (32nm)

It is actually Samsung and Intel driving to go 450mm wafer technology. Just imagine building Atom processors at 45nm on 450mm wafer technology. The theoretical maximum for good Atom die is ~2500 on a 300mm wafer. It will costs down on Atom on 450mm wafers to below $5 I would bet. (note: pure speculation on die cost there.)

 

[verndewd]

http://www.theregister.co.uk/2010/01/07/globalfoundries_gets_qualcomm/

AMD's former chip arm to bake Qualcomm wafers
Alert Print Post commentGlobalFoundries lands third customer

By Timothy Prickett Morgan • Get more from this author

Posted in PCs & Chips, 7th January 2010 22:38 GMT

Register: Cisco®, NetApp®, VMware® Web Event on 1/26

AMD's former chip manufacturing arm GlobalFoundries has inked its third customer for its wafer baking operations: wireless chip-maker Qualcomm.

Qualcomm said it would be submitting masks to GlobalFoundries sometime this year that make use of the wafer baker's 45 nanometer low-power processes - current Fab 1 - with an eye toward the future 28 nanometer low-power tech. Presumably this will be Fab 2, but there were no more details.

Qualcomm said that it would be creating chips to run through GlobalFoundries that support the CDMA2000, WCDMA, and 4G/LTE cellular standards, including the hot smartbook segment that really gets underway this week at the Consumer Electronics Show (CES) in Las Vegas, Nevada.

 

[verndewd]

ARM might have a bit of competition if moorestown can deliver Atom performance and power figures in a cell phone envelope. But I think 22nm will be better for that.



My guess would be Chandler. They tend to get a lot of the newer stuff after Oregon who gets first dibs. Plus the last FAB built there was pretty high tech, they would want to take advantage of that and save a bit of costs in upgrading.

If Arm can use amd tech under license, issue solved.

 

[jimmysmitty]

If Arm can use amd tech under license, issue solved.

They can use AMD tech but not Intel tech. So they in the end would be stuck without x86. The agreement right now gives AMD access to Intels stuff and Intel access to AMDs stuff. But it by no means says ARM can use x86.