Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ The Truth About IT Consultants ]   |  Apple to the Core  |   [ Iron Man ] >>

Weekly Column

Apple to the Core: Why Steve Jobs is buying PA Semiconductor.

Status: [CLOSED] comments (119)
By Robert X. Cringely
bob@cringely.com

Apple this week bought a fabless chip company called PA Semiconductor and pundits far and wide are trying to explain the deal with broadly varying ideas, some of which are close but none seem to really understand what the deal is about. In the short term this acquisition means precisely nothing to Apple users. In the long term it could be quite significant, however, and gives a number of tantalizing hints about Apple's hardware strategy.

Why would Apple, having already jumped from PowerPC to Intel, spend $278 million to buy a company that is best known for designing PowerPC chips? Are they preparing to dump Intel? No. Does it have anything to do with Intel? Yes.

This deal has Steve Jobs' fingerprints all over it. His first formal position at Apple was as head of purchasing and Jobs was known for pushing suppliers to ever lower component prices. He still brings to his work a purchasing manager's perspective and a desire to beat up suppliers whenever possible. So in that sense this acquisition is all about Intel. And the purchase price, which probably appeared to have been pulled right out of the air by Jobs, who then wouldn't budge from the figure, is really based on Apple's target savings over the next two years after forcing Intel to cut prices based on fear of a possible Apple switch back to PowerPC.

"We got it for free!" I'm sure someone at Apple said after factoring in the expected Intel price cuts. There is nothing that makes Jobs happier than forcing one supplier to effectively finance its own demise. It's that Devil thing again.

Whether PA Semi cost $1 or $278 million, Apple still has to DO something with the company and its technology and there has been some speculation that we're looking at the next iPhone chip, or perhaps the one after. That is very unlikely. PA Semi has aimed at workstation and server and high-power embedded chips that use far too much power for any iPhone. And while many pundits argue (and PA Semi even told some of its customers) that Apple was mainly acquiring intellectual property (IP), companies aren't typically bought that way these days. They are purchased for what they have already completed, not for what they might do in the future. There's a PA chip that Apple wants very badly but it won't go in an iPhone.

The fact that PA Semi told its customer that the purchase was about IP is meaningless. How could they know the motivation of this purchaser, especially THIS purchaser?

Of course that doesn't mean Apple couldn't switch to PA silicon for the iPhone over time. Once Apple has in-house this considerable design capability there is no limit to the good it could do on all platforms. But it is very unlikely -- almost impossible -- that this acquisition was based with the iPhone in mind.

Then why did Apple do it?

In the short term, it was to scare Intel into lowering prices by at least $278 million over two years. And in the long term it was to create a replacement for Intel as the prime CPU for Macintosh computers.

Here's the important part: CPU architectural advances are achieving minimal returns these days. It is all about multiple cores and clock rates, nothing else. This means that Intel chips are a commodity, and have become effectively interchangeable with those of AMD. I'm sure partisans on both sides would disagree, but my statement is still true. Except at the very top end of each product line, you can always find an equivalent Intel or AMD chip through balancing numbers of cores and clock rates. All that really varies is the price and that does darned little.

This commoditization means that what keeps Apple buying from Intel and Intel alone is only one thing -- money. Through whatever combination of discounts, non-recurring engineering, co-marketing funds, whatever, Intel is cheaper for Apple than would be AMD at this time. And as the only big all-Intel PC shop any more, now that Dell has succumbed, Apple has special value to Intel -- value worth $238+ million to retain.

But Apple having value for Intel doesn't at all mean that Intel has value for Apple. When it came to jumping from the PowerPC a few years ago, Apple had incentives on all sides. IBM was incapable of shipping PowerPC chips that would run reliably at higher clock speeds while Intel was already approaching 3 GHz. Intel was offering to throw money at Apple while IBM was distracted by manufacturing the PS3's Cell processor, based on PowerPC hardware and built in the same fab. Intel was offering higher performance at lower prices, so Apple made the jump. It was price and clock and nothing else and that's key, because Intel would like us to believe the X86 architecture played a role, too, which it didn't.

Apple is not in the least tied to the Industry Standard Architecture (ISA). Oh it worked out well and has helped Apple sell computers into Windows shops because of Boot Camp and the ability to run Windows as well as any desktop from Dell or HP, but this advantage is fading fast with the increasing popularity of virtualization.

Where Windows now runs better on an ISA (X86) chip, because that's what Windows was designed for, we are rapidly approaching the point where desktop virtualization will allow us to throw multiple cores at the problem, removing forever the X86 advantage, even for Windows.

The scenery is even more compelling looking from the other direction. Where X86 offers no true advantages for running OS X, it is easy to see that it could offer DISADVANTAGES, simply because OS X, as a Unix variant, was never designed specifically for X86, making a lot of Intel hardware simply unnecessary. If there are instructions that will never be used, why spend the silicon real estate to hard code them? CPUs optimized for OS X would be smaller, cheaper, and use less power than any Intel or AMD alternative simply because they could be simpler overall.

Which brings us back to PA Semi. In a world of multiple cores and high clock rates, PA Semi can produce a family of processors to drive everything from the Apple TV to the most powerful Xserve. Because they are not X86 compatible, these chips will run OS X faster on a per-watt basis, which is key for a computing world going largely mobile. Because they'll be Apple designs built on a competitive basis by almost any of the world's fabs, these processors will be cheaper MIPS-for-MIPS than anything Apple could buy from Intel or AMD, which Apple will convert into a profit-margin advantage, not lower system prices. If Windows performance suffers, that can be handled by adding more cores, but most users won't even notice.

Heck, by that time Windows will probably be virtualized anyway (what Microsoft should have spent five years and $5 billion on instead of Vista).

Apple is not tied to Intel or to X86. Jobs said they had OS X running on Intel for two years before announcing the shift, so it is logical to assume they have recompiled the OS to run on almost every competitive processor available today. OS X on the PS3? I'm sure it is running in an Apple lab.

Apple has changed processor families twice before in the Macintosh era so it is more likely, not less, that they will change again. It's even possible we'll see a jump to AMD for some machines before the final days of Apple/Intel. But just as the Intel changeover took a year and was predicted to take two, we're 3-4 years out on this transition. Your next Mac will probably have an Intel CPU, but the one after will be all Apple, through and through.

Comments from the Tribe

Status: [CLOSED] read all comments (119)

Nope, I'm not buying your line.

Some time in the not too distant future, when Apple is done taking over the consumer market, the enterprise market will become Apple's next growth engine. As instrumental as the switch to Intel has been in convincing consumers to take the Mac plunge, it's even more important in the enterprise market.

Furthermore, you just mentioned that AMD is right there ready and willing to supply chips to Apple. Why does Jobs need to fire a 278 million dollar shot across Intel's bow when AMD can furnish the cannonball gratis?

al | May 02, 2008 | 12:47PM

What a lot of people don't realize is the fact that Jobs has always promoted and favored "garage shop" operations segregated from the traditional corporate environment, allowed to think outside the box, free to invent the "NEXT" successful gadget, and not tied down by corporate policies stifling innovation. This was exemplified by the development of the original Apple I, Macintosh, Pixar, Next, etc. Current Apple corporate campus is set up in a more traditional way, focused on supporting and improving existing products as opposed to bringing up something new and revolutionary. This acquisition puts a ready made organization in Steve's hands fully under his control and away from the main thrust of Apple Inc. It will also be easier to keep a smaller organization under wraps until they are ready to reveal their new product which can be a challenge at the main campus given its current size, as I'm sure it was when the iPhone was developed.

Joe | May 04, 2008 | 9:52AM

"making a lot of Intel hardware simply unnecessary" please state one piece of the x86 architecture that is unnecessary for OS X.

Incog | May 08, 2008 | 1:08PM