Check out this article on Businessweek.
Both the analysts and the press are failing to realise the true value of Silverthorne. A SOC that's as powerful as today's computers at a fraction of the size and power consumption but most importantly ridiculously cheap to manufacture. At 2500 chips per wafer which is roughly 10x today's CPU's, Intel is basically able to manufacture Silverthorne at 1/10th the cost of say a Conroe CPU today. I know this is a gross over-simplification but I'm just approximating here to set the context. Let's assume a Conroe CPU costs anywhere between 50-80$ to manufacture. Hence, Silverthorne could end up costing as little as 5-8$ to manufacture. So even if Intel sells the part at 30-40$, that's a really sweet profit and some awesome gross margin.
Now think about all the places you could put a small, low power, high performance IA CPU:
- Ultra mobile PC's
- MIDs - the new devices Intel is touting.
- Consoles
- Handheld gaming devices
- Education devices for developing countries.
- Cheap PC's or laptops for emerging markets.
Man...the possibilities are phenomenal. Now consider the tremendous pricing advantage Intel will have. No matter how much AMD cuts prices, they can't win if Intel has a part where they turn a hefty profit even if they sell at AMD's cost price.
What Intel is creating here is the Toyota Corolla of the CPU market. Affordable, reliable, reasonable performance and fantastic value for money. They finally figured out that's it's not enough to trim headcount and operating expenses to bring down costs. They must have a true low cost part and not keep down-selling their best technology by reducing cache or removing features and selling it as a Celeron. The low end of the market is a reality and to win it, you must build the right products and brands for it.
If the analysts thought Otellini was only referring to re-structuring when he was talking about profits growing faster than revenues at the analysts meeting last week, they missed the point completely. Silverthorne if handled right will bring awesome profits even if it doesn't grow revenue at the same proportionate rate as the traditional CPU biz.
Hector and team instead of borrowing money and still shooting for 30% market share should be reducing costs and trimming headcount. They are too far behind Intel in terms of products for where the market is going. Barcelona will not save them this year as it will really begin to impact the financials next year...and Penryn is looking pretty good too. Intel is flanking them from all sides - product, manufacturing, cost and just plain old managing the business.
Expect AMD to lose another 500 million or so in the next two quarters.
Subscribe to:
Post Comments (Atom)
10 comments:
"No matter how much AMD cuts prices, they can't win"
You've completely lost me with this one. Why would AMD be trying to compete with this? This is under AMD's Geode NX line.
This is an embedded chip. AMD has completely ended its Geode GX line which would be this chip's competitor. I don't see how this would effect AMD to any extent. I suppose it could hurt VIA.
Toyota Corolla?
No try a Matrix or someting. Something dirt cheap, efficient and meets probably 70% of the buying publics need for a car; car around town errand, to work, pickup and dropp off the kids. Everyone should want a cheap, efficient car. It sips at the pump, easy to park, green, cheap to insure. THen you always need a do it all too; SUV-Minvan-Wagon-Truck for those other must do with the whole family. I'd say probably the same for your always in the pocket electronic device right? For the same reason a econ car ain't the best seller is the same reason a truck is the best seller and the same reason that Silverthorne will never sell a hundred million.
The reality of the business is that that the ultra miserly, ultra-heap, lower performing line has not caught too much interest. The capability has been around for years. If there was a need some clever CE company like Apple would already have a device there; Newton, tablest, PDAs are a tech toy chasing a business that doen't exist. It has NOTHING to do with lack of capability and everything to do with lack of application.
People either want a smart phone; make a phone call, take a impromptu picture/movie, look at some small screen clip etc. Apple and others will sell a couple hundred million smartphones in 2009, too bad it don't got a sliverthorn inside, but the real issue it don't need it!
Or people they want a 14" screen full featured keyboard In another two years laptops that go for 6 hours and weibh 4 lbs will rule with wimax will be the rule with instant access to the web. Once you have a full function PC why would you settle for 4x battery life for 1/8 usability?
Silverthone may find a market in niche places like medical electronics, educatoin PCs. It won't sell 100 million units unless Paul and INTEL can getinto a phone. Trust me the phone guys are aware and making sure that doesn't happen. The phone uC virtuous circle is as large as the x86. TO think Paul can break into phones is as absurb as thinking PowerPC, Transmeta or someother CPU wanna be will break into x86. Paul should know first hand with the tens of billions wasted on Itanium and tens of billions wasted on communication.
Speaking of that how did they allow 20 billion to get pissed away... Can you imagine where the x86 business would be if those tens of thousands of employees and billions were focused on x86. AMD would be nothing but a footnote already.
Lex - this product will end up inside some very cheap PCs and notebook like devices in emerging markets where the next hundreds of millions customers are waiting.
Guess where the highest cell phone growth rates are - in markets like India and China. When you have a PC that costs the same as a cell phone, the growth will be explosive.
Scientia - why is this a competitor for Geode when I'm talking about it ending up inside PC's and notebooks in developing markets.
Notebooks are not the primary market for this. You can build a similar notebook or PC with Geode LX with a 7" X 7" motherboard. The advantage for this chip is for something smaller than 7" X 7". This is a niche product and will have little effect on AMD.
tbh, i've always been skeptical about the UMPC strategy. . it's been revived so many times. I've seen the product in youtube. Personally i don't find any use for it.
this products sits in between the laptop/desktop and the mobile phone. Isn't that where the smartphone is heading? To me this is like the PDA all over again. I'm betting this will fail, sorry.
Guys - I said notebook like devices. Not notebooks. Think about Negreponte's device - what wouldn't he have done to have a 25$ platform.
There a few things to note here:
1. Outside of UMPC's and MID's as Intel envisions them, the opportunity is created for people to build many different types of compute and communication devices in different shapes and forms to meet different needs in different markets at with low cost structures.
2. This will drive the price of MIDs and UMPCs down below to the pricing of a high end mobile phone. These devices don't replace smartphones. That's not what Intel is trying to do. What they are trying to do is create a whole new category. Couple this with Wimax and you have a full fledged PC with always on Internet that's about the size of a PSP for probably under 500 bucks. And...like the cell phone market I can bet you the service provided will subsidize the cost of the device for you to sign on to Wimax.
3. The biggest opportunities for this lie in places like rural India. Where the penetration of cell phones is extremely high because there's no fixed line infrastructure. Now if you bring cost effective Wimax coupled with a very low cost (I'm talking maybe 200$) portable PC device, then you could potentially enjoy the same kind of explosive growth the cell phone guys enjoyed in India and China.
My point is - Silverthorne and it's successors provide the ability to move into several high volume markets where low cost for decent performance is the name of the game. Just like Nokia and Sony Ericsson are building low cost cell phones in high volume for these emerging markets.
As to why does this matter to AMD?
It matters because if Intel's growth is coming out of areas where AMD doesn't compete, it allows them the flexibility to make less money in their traditional business which means they can compete harder against AMD.
The same way AMD used their margins from their sever business over the last couple of years to fund their efforts to win share in clients.
Well, you can buy a real notebook today from HP for $500. The problem with a smaller unit is there is no keyboard and the display is smaller. If you can live with these limitations then I guess it could be useful.
ATI is more involved with consumer products than AMD so I'm guessing if there is a real market opportunity for this AMD will develop something.
Sorry for digressing, but I think Scientia has lost it or maybe Sharikou has hijacked his site... he is now doing made up interviews ala sharikook, much like in his other recent articles where he makes a bunch of assumptions and half baked conclusions and then states "we can now see..."
Scientia is now stating that "common sense" (which buy the way is code for HIS BIASED OPINION) is AMD is only 6 months behind on process technology right now (I assume he means 65nm?) and "This should make AMD and Intel about equal at 45nm"...
So I guess we will be seeing AMD 45nm parts FOR SALE this year? When again did Intel and AMD have their 65nm parts on the market? Adn exactly when did AMD finalize their process improvements, aren't they still implementing changes? (Intel was done at launch). Now he's going to compare what will be a dumb shrink from 65nm to 45nm with near 0 transistor improvements (these will come ~end 2009 or ~1-1.5 years later when they get around to implementing high K/metal gate) to an Intel process which has both the linewidth shrink AND the transistor improvements implemented THIS YEAR! Apparently Scientia thinks process technnology performance is solely goverened by the critical CD...
He also thinks an extra metal layer (or 2 in the case of K10, which now has 11 metal layers, but let's overlook that for now...), is offset by AMD's use of immersion on 45nm vs Intel's double patterning.
To put this in perspective he is comparing the extra litho patterning step to (for one metal layer):
- ILD deposition (CVD)
- Etch stop layer (CVD)
- 2 oxide etch steps
- 2 patterning steps (non-critial wavelength)
- barrier/seed deposition (PVD and /or CVD/ALD)
- electroplate
- CMP
- associated cleans
- any yield impact by having another metal layer and all of those extra steps
On top of this he fails to realize that immersion tools cost ~2X a dry 193nm litho tool (which would offset the double patterning cost by itself) and apparently he doesn't understand that immersion has a slower throughput than dry litho...but why let the facts get in the way of "common sense". And lets not even get into the cost of SOI wafers - I hear those are cheaper than Bare Si! :)
Funny he states all of this matter of factly as if he had some background in semiconductor manufacturing. NOTE: watching Conan tour Intel's fab does not give you the requisite manufacturing knowledge.
And finally...
"That's . . . just a coincidence. Intel has always had better process technology.
CS: You mean like back in 2000 when their PIII couldn't keep up with K7?"
Apparently Scientia has taken some dumb pills and doesn't realize the difference between PROCESS TECHNOLOGY and ARCHITECTURE - it was not AMD's "superior" process (or Intel's "inferior" process) that caused this - this was SOLELY an ARCHITECTURE difference (as anyone with half a brain would understand). The gap actually would have been much worse without Intel's process technology capability... if he took the time to read ANY scientific journal and look at key parametrics he would be able to compare the 2 process technologies as opposed to comparing 2 different products where the architecture difference completely swamp out any comparison of process technology.
It's kind of sad, he used to at least TRY to be objective and reasonable, now he apparently has become a Sharikou wannabe.
AMD is too far behind to even think of competing with this product right now. They will wait for Intel to create the category and then try and steal a piece of it.
Post a Comment