Apple should think about building a wafer fab. I'm not joking or merely trying to be provocative. I'd even bet that Steve Jobs has flirted with the idea.
Apple should consider—and I mean just consider—building its own fab to make the A4 (the engine for the iPad and iPhone) and follow-on processors. And at some point, an Apple-owned fab might be a necessary evil.
Fabs, of course, are expensive, and for years Apple has done fine without one. As a traditional OEM, Apple bought components, integrated them into systems and sold the finished products, content to let its chip suppliers shoulder the fab burden. Why mess with the formula?
Because Apple, with the iPhone and iPad, has become more systems house than traditional OEM. Like Cisco and IBM, Apple now designs its own complex ASICs. It even acquired IC design houses P.A. Semi and Intrinsity to bolster its ASIC design expertise.
What's more, Apple and its key foundry partner, Samsung Electronics, are on a collision course in several end-user markets. But as a foundry, Samsung gives Apple a sweet deal for ARM-based processors and NAND flash supply. That makes it hard to walk away. There's no guarantee another foundry vendor would offer preferential treatment or pricing.
So the Apple-Samsung foundry arrangement isn't likely to dissolve anytime soon. But ego could be the third wheel in this relationship.
I believe Jobs wants more control of the supply (and perhaps manufacturing) chain to buffer his company from the chip industry's boom and bust cycles. In boom times, for example, foundries struggle to keep up with demand, and OEM customers are at a disadvantage—a position from which strong egos don't like to negotiate.
Meanwhile, Samsung and Apple are competitors in the smartphone, PC and tablet PC markets. Today that doesn't worry Apple much, because the companies are worlds apart in branding. Apple’s brands are known the world over and are synonymous with cool. Samsung, too, builds solid, interesting products. But can you rattle off the names of its phone models and TV lines? Didn't think so.
Apple and Samsung's existing arrangement suggests a relationship in flux. Apple buys Samsung’s ARM-based MP3 chip set for the iPod. For the iPad and iPhone, however, Apple designs the ARM-based ASICs, and Samsung makes them on a foundry basis.
Apple today seems to call some of the shots at Samsung. The South Korean company recently announced a $3.5 billion fab expansion plan in Austin, Texas. Most of the spending will be funneled into Samsung’s logic business, and much of that will go toward serving Apple requirements, according to my sources.
Samsung’s logic business is currently centered in its home country, a geopolitical hot spot. I'd venture a guess that North Korea's recent saber rattling in the direction of its southern rival made Jobs uncomfortable, and he nudged Samsung to make the logic-capacity investment in Austin.
Still, push might one day come to shove. As Samsung accelerates its smartphone activity, enters the tablet market and reenters the PC market, its identity as a competitor to Apple could weigh more heavily on the pair's foundry relationship, and Jobs would want to exert more control over the A4's supply.
I don’t expect to see Apple build a megafab for NAND flash; that kind of investment, given the dictates of Moore's Law, would be unsustainable. But I could see Apple building a traditional, smaller-scale logic fab to gain more control over the A4 and its follow-ons.
Logic fabs can cost $4 billion to maybe $10 billion to build out, depending on the tool sets, but Apple could swing the funding. I am sure Texas would pick up some of the tab for a new fab on its soil. The same goes for New York.
In an age when real men go fabless, I concede it's an unconventional idea. You might think it's absurd. But an Apple A4 fab today could keep the iProduct franchise in hay—and Samsung at bay.
"Logic fabs can cost $4 billion to maybe $10 billion to build out, depending on the tool sets, but Apple could swing the funding."
Perhaps so, given how much silicon volume Apple generates. But consider this:
* A fab would only handle one type of fabrication, for example the ARM SoC chips in CMOS. Apple has significant costs in FLASH and the displays as well, so all that capital would go to solving only one part of the total BoM.
* And all of that capital would be sunk into a single node, likely 28 or 22nm for a fab targeted to open in a year or two. It would take an even larger investment to convert to the next node. By using foundry partners, Apple can purchase latest node devices at market rates, without the capital outlay.
Apple has made some adept moves in this area, like switching from to Intel processors for Macintosh. This allows them to ride Intel's cost and performance curves without capital investment. It is unlikely they will suddenly decide their best use of capital is to build a fab!
Apple should design their own IC's but, not necessarily own the fab house. I am most familiar with power converters bought from the far east which where huge, bulky, noisy modules no better than hobby junk from the 80's. How many millions of dollars due to lost sales on the mac7 because of these cheap things blasting noise? Apple could have it's own dedicated IC's designed into compact circuits with the technology of 2010.
Apple would benefit more from staying away from the Foundry business and negotiate with the current ones. There are some foundries to talk too and Samsung is not the only one ....
But Apple has already looked at second sources or else they would not be Apple ...... The negotiations are probably already at the end phase and pretty soon you would see the announcement.
Do you know how much it costs to have a fab? Try 1 Billion. Plus the headache of filling it to capacity. Real men go fabless because they are good in math...
If you are taping out a digital chip, there is no need to own a fab. TSMC, UMC or any other fab will do. Unless they have analog chips that need secret recipes to make it work...
Apple doesn't even have its own manufacturing. It all done in China by Foxconn. That's why it's so profitable.
Adam Smith in his 'Wealth of Nations' described precisely why this 'division of labor' can create wealth.
Asking Apple to have a fab will definitely cause both their profit, gross margin & stock to tank..
Hi Steve (Jobs)! I'm sure you've noticed this: Your chip/foundry partner, Samsung, has rolled out its tablet PC. It's an Android-based system. Samsung plans to sell 1 million units this year and 10 million next year, according to HSBC. Vodaphone will sell the product, which can make calls. Did you also see this? Samsung has increased its smartphone shipments to 25 million units, up 39 percent from its previous forecast. Last but not least, Samsung rolled out a notebook PC line. Steve, time to call TSMC and GlobalFoundries and look into your own fab.
If Apple wants control of its destiny, it needs control over the technology used in its products. This means they should invest more in their R and D efforts to develop technologies required, which may be cheaper than that offered by a third party. Ofcourse this isn't easy if there are third pary patents/licenses involved, but it will help reduce the cost of produc development. Running a fab is expensive and for it to be economical, Apple will need to sell a lot of their products to keep the fabs busy. Maybe in a few years when Apple have greater domination in the market, but not now I think.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.