SAN JOSE, Calif. - At present, Samsung Electronics Co. Ltd. is making the Apple-designed A4 and A5 processors on a foundry basis for Apple Inc.
That could soon change. As reported, Apple and Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) have entered into a foundry relationship for the A5 and follow-on chips, sources said.
Now, according to an analyst, another player is pursuing Apple's foundry business: Intel Corp. Intel is already supplying x86-based processors for Apple's PC line. Intel is also dabbling in the foundry business and hasrecently struck a dealwith Achronix Semiconductor Corp.
''Based on a number of inputs, we believe Intel is also vying for Apple's foundry business,'' said Gus Richard, an analyst with Piper Jaffray & Co., in a new report.
''It makes strategic sense for both companies. The combination of Apple's growing demand and market share in smart phones and tablets gives Intel a position in these markets and drives the logic volume Intel needs to stay ahead in manufacturing,'' Richard said.
''Intel's manufacturing lead gives Apple an additional competitive advantage in these markets and distances it from Asian competitors that are knocking off its products,'' he said. ''Furthermore, it would also serve to weaken Samsung who is a significant competitive threat to both companies.''
Samsung will remain Apple's main foundry-at least for now. ''While it will take a few years for Apple to shift foundry suppliers, we believe Apple is shifting away from Samsung,'' he said. ''We believe TSMC will start getting revenue from Apple in Q4 of this year. We believe the recent patent lawsuit between the two companies is further evidence to support our belief that Apple is moving its silicon needs elsewhere.''
Samsung and TSMC each have the fab capacity to support Apple. The question is clear: Does Intel have the capacity?
''Samsung has just completed a 30K-40K wafer start per month logic plant in Austin Texas to support its foundry business of which Apple is its largest customer,'' he said. ''Based on the die size of Apple's A5 processor, Apple needs roughly 23K wafers a month for the A5. We
believe that Apple moving its foundry business away from Samsung is what has recently driven Samsung to reduce equipment orders, as it will likely repurpose this capacity for memory.''
This is very bad news for Samsung. Moreover Samsung's profit has fallen by 30% during this Q1. Needs to be seen how Samsung will compensate for this. On the other hand I am sure Intel will do everything it can to get business from Apple.
It could be the turn of the events that will hit Samsung to oblivion. Yet, we have to ask if Intel should be everywhere. This is very scary to have Intel hold the hottest Apple in their hands. It consolidates that power - badly.
I can see Intel fabs supporting their own relatively high margin products. Being able to compete for more "commodity" offerings is a different ball game. However, Apple did move to Intel CPU's on their computers. Time will tell...
Samsung is the biggest competitive threat to Apple so it makes sense to move away from Samsung. TSMC is a good cost competitor, but it’s always a question how much technology goes out the back door in Taiwan and China and Apple is coming under increased pressure from no-names. Intel has shown itself to be grossly incompetent at making system level products, so Apple has nothing to fear there. Maybe Intel wins by being incompetent, assuming it really wants to be in the foundry business. 25k wafers a month is a lot of wafers when good times returns to CPU. Intel has no history of saying no to spare resources when CPU calls.
For our product (40nm and 45nm), wafer from TSMC is still much higher than from Samsung.
Looking at the RE A5 layout, you'll noticed lots of similarity with A4. I wonder how much of the A5 work is done by Apple internally vs. by Samsung design services; I would guess the standard cell, soft IP and layout is done by Samsung, that makes it more difficult to switch Foundry. Also because Apple/Samsung deal is "Good Chip" based contract, with the current A5 yield, I doubt if Samsung is making any money out of A5. I would consider Global Foundry (Fab 8) to be a more reasonable foundry replacement since the 45nm process characteristics is much similar to Samsung compare to TSMC. Intel will not be the foundry for Apple on 45nm products; maybe for 32nm or 22nm node; But Intel's process is optimized for regularity design and design rule restrictions gonna be a huge problem for low power products. By comparison, the cost from Intel will be around 20-40% more than tsmc or Samsung makes it even less likely to manufacture for Apple.
Well given that Apple designed the A4 and A5 making use of an architectural license from ARM, it would be a radical change if they turned to x86 chips for future generations of i-hardware.
The implication if this story seems to be that Intel COULD become one of the leading manufacturers of ARM-based processors, both in terms of technology and volume, through a foundry engagement.
It sounds quite interesting and probable in future Ax SoCs. Both Apple and Intel are already partners for Mac (also thunderbolt). There is no apparent rivalary with intel compared to Samsung. Going with intel can also provide Apple process node differentiation over rivals in future 22nm ... and for intel it is a significant volume product to fill Fab.
Good to see that Intel is pursuing Apple - something I had written about here a couple of months back.
Compeared to newcomer foundries like either TSMC or Samsung, stolid old Intel still has better physics and processes as well as the bank balance to turn them into leading edge Fabs at least a node ahead of the pretenders ( incl. IBM ).
Intel can offer Apple an unique capacity for shrinking ( at 22 nm ) their massive A5 die and provide lower power dissipation / long battery life due to high k metal gate process combined with lower interconnect delays from ELK dielectrics( k lower than 2.5 ) / Cu metalliztion. All this translates into competitive advantage for Apples i-whatevers.
But they had better move fast because TSMC is very aggressively pursuing 20 nm and FinFETs.
Regarding Intel, they would have to get off the Stalinist Gulag mode of running Fabs and get used to the somewhat flakey ex PA Semi designers / prima donnas at Apple who are doing the A5 etc.
Apples challenge is to prevent its i-whatevers from turning into over-priced boutique products for its cultists, the same fate as the Macintosh after the Wintel juggernaut replicated the GUI of the Mac.
The ARM - Foundry ( TSMC / Samsung / Global F ) - Android OS combo is generic like WinTel, lower cost and already has a larger market share than iPhone. Many new / second tier companies in growth markets are using this platform.
If Apple thinks Intel can deliver better ARM products than Samsung, and give Apple an edge over the crowded market then it will do it. Apple needs something like this. Either Samsung or Intel.
I don't think Intel would ever make low margin SoC/CPUs on the state of art process node if they didn't have a choice. I however, did hear CEO Otelleni say in 2006, that Intel's future is high volume, and process tech, and lower margins, not what they have been doing with HVM, but expensive CPUs. They could hope to make up the profits with more traditional Server/HPC CPUs for businesses and governments. Not really surprising, this idea is floated.
Intel would have to deliver a stunning x86 low voltage, higher performance SoC device to even be looked at as an ARM replacement. That seems unlikely with ATOM, even at 22nm.
Once again, EE Times readers have offered a lot of valuable opinion and reasonable speculation. Let me just add a couple points briefly even if they have been alluded to previously. Moving to Intel as a foundry does not mean moving from ARM to X86. It could be further down the road and be something proprietary developed (or acquired) by Apple. Also, Intel has lots of motivation to take on Apple business, not least of which might be staving off a surging Samsung who have long set their sights on being the world #1 semiconductor company.
This gives too much to Intel, in return for what exactly? I am not sure Apple should go down this road. They should stick with the well-known and trusted pure foundries. Granted the tie-up with Samsung is problematic but a tie-up with Intel would open up more and bigger problems in the future.
Apple is moving in the direction of rivalary with an old friend. I think it's the inner fear of Android and also the fact that Smamsung keeps it's profit margins lower and thus can outperform Apple any time!!!
@hm: it may be too premature to declare the "success" of Thunderbolt high speed interconnect. Apple has had a history of going it alone throughout its corporate history; the only differentiators this time are the much higher volume consumer devices (as opposed to desktops of the past where Apple received a drubbing!). Apple also makes it easier for others to fill in the gaps it doesn't serve (read: flash-capable devices, for one!).
I don't believe Intel is going after Apple's foundry business entirely for revenue reason (there are TSMC's, UMC's of the world for lower costs for that). I think Intel sees an opportunity for technology collaborations for its next generation of Silicon.
Dr. MP Divakar
Food for thought: Intel once made one of the fastest ARM chips ever, the StrongArm (later called XScale), before selling it off to Marvell.
One of the architects of that chip? Dan Dobberpuhl, who later started PA Semiconductor, which was afterwards bought by Apple.
So many of Apple's ARM team were once designing ARM chips for Intel.
January 2016 Cartoon Caption ContestBob's punishment for missing his deadline was to be tied to his chair tantalizingly close to a disconnected cable, with one hand superglued to his desk and another to his chin, while the pages from his wall calendar were slowly torn away.122 comments