Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Reportedly Buys Into a Chip Fab, Looking to Produce Its Own Chips
- Thread starter MacRumors
- Start date
- Sort by reaction score
This claim only makes sense under two hypotheses:
(a) Apple doesn't know how to run a fab AND
(b) What fabs produce is a perfect match to Apple's needs (or at least good enough)
Neither of these are obviously true.
(a) Few thought, the day before the A6 was announced, that Apple would have the skills to create not just an adequate ARM CPU, but a leading edge ARM CPU.
(b) Even though Apple can now control the design of their SOC, they cannot control the power/performance tradeoffs of the fabs to whom they subcontract. It's not at all obvious that these fabs are INEVITABLY going to do what Apple wants. The historical pattern is that Apple sees further into the future than everyone else, and has to struggle to bring their partners along. Apple had to work damn hard to get Intel to take power as the most serious metric of CPUs (and Intel's failure so far in the phone space tells us something about how seriously Intel took this five years ago WITHOUT Apple forcing the issue).
To my eye (yeah, yeah hate as much as you like) the transition to quad-core CPUs (which are simultaneously starved of memory bandwidth) by some ARM vendors shows that already we have designers who don't know how to move forward, what really needs to be done. Apple (and apparently only Apple) have the foresight to do things like
- work on a REALLY low power (and slow) CPU which, however, has fast bluetooth and maybe WiFi --- the right CPU for an iWatch, which needs some interesting display/GPU capabilities, some wireless capabilities, but practically no compute capabilities (because all that will happen on the phone) OR
- pay the money to do something everyone talks about and no-one does, to develop 3D interconnects to replace PoP OR
- license from IBM, or invent their own process for creating eDRAM, allowing their ARM SOC to both run that much faster and at lower power.
All three of these require process innovations which don't appear to have happened in the shared fabs, and which Apple has no interest in researching for a partner, who will then use to for competitor's chips.
Anybody checked out how much the report costs? $1000. How's that?
Still, that buys you a year's subscription to SemiAccurate so for a site the size of Macrumors ...
That's what I say.
I'm also a chip guy and worked running a design center for a FAB house for a while.
Buying and running a FAB and keeping it competitive is not for the faint of heart.
Qualcomm sells more radio chips and processors than anyone and still does not have a FAB, nor is it a good idea.
Not only would Apple need the fab, they would need to hire process engineers and buy or develop the next shrink technology. There is an army of people that get the next node viable and to market.
They would also need a company/companies like Dolphin or ARM develop cell libraries, memories, analog phy's, pll's, dll's, etc and all the pieces that go into making a chip.
Apple does not currently develop all the pieces that go into a chip.
Companies that develop all the other IP, exist for a reason.
Dolphin, ARM, and other IP companies will need to port their stuff to Apple's FAB process. Unless Apple is willing to throw ton's of money at them, and remember this is supposed to reduce cost. The companies not only want cash up front, but often a royalty on parts moving forward. Try getting something in 28HPL (not the mot popular process) at TSMC and see if all the IP you want is there? The IP vendors go where they can leverage the IP in to as many design wins as possible while they amortize the cost over the widest audience. This drives their profit up and overall cost to deploy down.
As a person that works in the semi-conductor industry, I can't see how this rumor makes any sense or leverages Apple in any positive way.
Yea, they shipped 600 million processors since 2007 and Intel has only shipped 50% of that. The difference is that Intel has one competitor for x86 and that competitor is fabless, and even AMD spun off their FAB arm as Global Foundries. They spun it off because trying to keep pace in FAB technology was killing their bottom line.
I'll believe it when Apple starts hiring ton's of PHD's in thin film and chemical vapor deposition (CVD) technology. When they start a research arm in advanced semiconductor physics and process technology, I'll believe the FAB story.
How will I know? Since I work in the industry, I'll see people I know getting crazy offers to leave their current jobs and jump ship to Apple.
You don't start up a FAB without it affecting the salary and hiring practices of others in that industry.
It's like this. You can't make 600 million phones without buying 600 million, LED's, screens, buttons. screws and everything else that went into making the device.
So I haven't seen the spike, so I'm a skeptic.
The rumor is NOT that Apple are creating a fab from scratch, it is that they have bought a (presumably substantial) share in an existing fab.
The point is to gain CONTROL over the direction of the fab, not to start from zero.
Interesting to the extent that Apple is probably shoring up their logistics and supply chain for future anticipated product demand.
I doubt we will ever see a "power by Apple" campaign.
I doubt we will ever see a "power by Apple" campaign.
You miss the boat.
The laws of physics are what limit what can be done and contrary to what you may think. The same low power, low leakage, high performance process that Apple wants is *EXACTLY* what Qualcomm and everybody else making a chip these days wants.
Everybody walks into a FAB and tells them the *EXACT* same thing.
I want zero leakage.
I want almost no power for dynamic switching.
I want to run at 4GHz while doing it.
I want to do all that and not have a heatsink.
Now, what can you do??
You can hire the right people to run a FAB so that is not an issue.
Apple developing it's own ARM variant wasn't a shock.
They enjoy a relationship with ARM that dates back to the inception of the company. They get licensing that nobody else gets. Designing a processor is a lot different than designing the next node in process technology. There is a reason why various FABs have created joint ventures to share technology to get to the next node faster. COST. Anyway they bought several companies that designed processors and they were pat of AIM, Apple/Intel/Motorola to develop PowerPC architecture processors.
You can't honestly believe that Apple will be able to, excuse me, out FAB all the current FAB companies like TSMC, Global Foundries, IBM, Samsung, SMIC, etc???? These companies drive the technology so that companies like Apple and Qualcomm, etc. can do what they do best.
Mind you the rumor of Apple building a FAB is not new:
http://www.cultofmac.com/206791/pro...10-billion-chip-fab-might-be-built-in-oregon/
Also read this article:
http://semimd.com/blog/tag/global-450-consortium/
But to address your supposition that Apple can't control the power/performance of the FABs they contract you are wrong.
They can, within the process technology ask for skew lots and specific transistor doping, They can get custom libraries that trade leakage and channel length. You presume that Apple sees further than anyone else but Intel, Samsung and others are working on 450mm wafers and looking into the next decade. Apple can look out as far as it wants but lithography techiques and the machines that do wafer processing are going to limit whatever they do. You can't get to 14nm if there are no companies making machines that do it. Intel and others work closely with the companies that make wafer processing equipment to get them what they need.
Apple is not poise to leapfrog what everyone in the semi space is already doing. They goals aren't any different that any other company that has a FAB or is fabless.
Low static power, low dynamic power, which means lower heat and super fast performance are what everyone in the industry is going for.
There is not a company around saying "we don't care about power consumption and heat".
You can call me wrong later but unless they buy the FAB that Samsung has dedicated to Ax processors in Austin or some other one, they are easily 2-3 years away from bringing anything on line if they have to build it from the ground up.
Last edited:
Seems unlikely that any current company like TSMC, UMC, etc. would be interested since they already walked away from truckloads of cash more than a year ago. They balked at both Apple and Qualcomm about buying in to secure capacity.
There is no motivation for any FAB company that is already full to the gills slice out capacity when they can already charge what the market will accept.
The only logical choice is that Apple may have bought the Samsung Austin FAB that makes Ax processors already. If that is not the case, then wait for two weeks and listen to the quarterly report.
Other than that they could buy a FAB that is being retired, but that would mean a capital investment of $3 Billion or so to upgrade to 28nm or 22nm and a wait of two years to come online.
It makes perfect sense to secure their supply chain. It does not make sense to buy a FAB and run it themselves. That is why the Samsung Austin fab was ideal.
They don't.
They do.
Yes, they are.
Writing Verilog is a world removed from running a leading-edge process fab. Design and manufacturing are, like, different galaxies.
Anybody holding Apple stock better hope this report is incorrect, unless what they are actually trying to say is Apple is purchasing a guaranteed production capacity of somebody else's fab. Who that somebody else might be is a mystery, though, as the major players have already rejected such a proposal, and not just from Apple.
Except that it was announced yesterday that Apple is pushing for improvements in Sanyo's IZGO screen technology and they're looking to move retina displays to that tech. And they already bought an Israeli flash memory designer and fab last year, IIRC. Lastly, Apple is moving Ax chip production to TSMC - and now, possibly, they're gonna fab their own.
It's a brave new world at Apple - and that new world seemingly has Samsung as persona non grata... We'll see what happens but uh...
Actually Intel doesn't own every single patent on x86/64 processors, AMD owns their fair share too.
Really? You have insider information from Intel on that? Intel has licensed many of their patents to AMD as has AMD to Intel.
----------
Not really something to be proud of.
----------
eeeh.... even if he/she grows own food...god/nature already designed it unless its a gm food
What if it is open sourced food?
Actually Intel doesn't own every single patent on x86/64 processors, AMD owns their fair share too.
Really? You have insider information from Intel on that? Intel has licensed many of their patents to AMD as has AMD to Intel.
----------
Not really something to be proud of.
----------
What if it is open sourced food?
amd got its licensing dating back to when they together were both making chips and amd is lucky in how it all went down to keep its x86 license
amd is no comp to intel and sold all its fabs any way and is only worth a few billion if that.
Apple with its massive amount of money could pay people enough to become a major problem to intel and they would never give them a chance to even compete by licensing x86 to them
intel tried hard to even terminate there license when amd sold its fabs and started using glofo
Last edited:
Good move by Apple. Although this move will not benefit them right away because it will take many years before they can wean themselves from 3rd party chip suppliers.
Still a good move for the industry, because Apple is now a new player, and the chip manufacturing industry will become more competitive, maybe prices will get cheaper for the entire industry. More competition, long term beneficial for all tech customers.
Still a good move for the industry, because Apple is now a new player, and the chip manufacturing industry will become more competitive, maybe prices will get cheaper for the entire industry. More competition, long term beneficial for all tech customers.
eeeh.... even if he/she grows own food...god/nature already designed it unless its a gm food
God/Nature 'designed' vanishingly little of our modern food even without 'GM'.
Fake. Apple had its factories. Jobswasted billions to get his personal–styled factory. At least now they are making something pragmatic.
So much for understanding what you are commenting about....
Troglodytes.... honestly.
Surely this is far more about apples legendary production control and cost minimising rather than them suddenly becoming the next intel (which I doubt is feasible)? They'll just carrying on licensing arm designs, minor tweaking and even if the fab is defunct in X years, I'm sure someone's done the maths (possible it's not producing CPUs either?).
Am I remembering wrongly, or did Apple not already buy a chip fab a few years ago? P.A. Semi was it?
Pretty sure pa semi were fabless designers - exactly what apple needed to tweak arm's IP
Until recently I would perhaps have agreed. I work for a company that designs and manufactures a wide range of semiconductors, from memory and microcontroller devices through to discrete devices. We have 2 of our own fabs and also use contract fabs.
There are many fabless semiconductor companies, and they use contract fabs, such as TSMC, UMC etc, to manufacture for them.
When the financial crash occurred in 2009/10, many companies cut back on their requirements, leaving companies like TSMC with excess capacity. Fab manufacturing costs are very sensitive to volume - less devices being made means fixed costs have to be spread across fewer devices and so profits can fall sharply, even become significant losses...
With the ongoing economic challenges world wide, the variability in demand for semiconductors has continued.
What it means in practice is that the contract fabs like TSMC etc have taken a more active and cautious approach to 'wafer starts' - beginning to manufacture more wafers containing a specific customer's chips. This has extended lead times out from say, 12 weeks, to double or more as they manage volumes and when to start. This has been done so that the contract fabs aren't left 'holding the bag' should a customer then alter or even cancel an order.
Apple has always taken a very active role in managing its supply chain, so taking a stake in a fab company in order to secure a more preferred status as a customer for chips makes a lot of sense.
Buying its own fab outright and then running it has an easy bit, and a hard bit. The easy bit for Apple is stumping up the $2-3Bn or so that a good (up-to-date) second-hand fab would likely cost today - and with excess capacity issues in the fab market, there's probably a few potential sellers of such a fab around.
The difficult bit is actually running the fab efficiently over a period of time.
So the key deciding question over whether to outright own a fab or simply buy a significant stake to secure preferential treatment comes down to how much value Apple places on being able to control delivery/lead times for what it needs. I suspect it attaches considerable value to this, but not sure if it's enough to warrant going 100% in to the chip manufacturing business...
Either way, we live in interesting times!
----------
The chip industry prices don't quite work like this. Apple becoming it's own supplier means that their current purchases simply move from one supplier to a different one - it doesn't grow the $300Bn a year semiconductor business 1 iota. Competition could go DOWN if anything because Apple, one of the top 5 semiconductor purchasers in the world, is moving its purchase of SOME (not all) of its chips effectively OUT of the competitive chip market. There is currently excess capacity so removing some capacity from the market will rebalance to some extent any supply-demand tensions. At the very least, any changes in price will likely be hidden by the annual 5-8% reduction in prices the industry has seen for many decades - and notice how the cost to end users stopped dropping some time ago - you don't pay less each year, but you do get more for your money - usually in terms of performance, power use or memory.
They should have started doing this a long time ago. You're always vulnerable when you rely on components from other companies.