SteevoApple TV was the Windows ME of DVR/Tivo. Not so good.

Why Did Apple Buy PA Semi?



Daniel Eran Dilger
Just ahead of its recession defying, record setting Q2 2008 earnings reports, Apple revealed plans to buy PA Semi, a chip designer specializing in processors based on IBM’s Power architecture. This news sparked a flurry of confusion from observers: why is Apple getting into the semiconductor business after partnering with Intel in its Mac systems, aligning with ARM licensees for its mobile WiFi iPhone platform, and particularly after decisively migrating away from PowerPC in 2006?


What is PA Semi?
Apple’s acquisition target isn’t a chip manufacturer. As a “fabless” chip designer, PA Semi (short for Palo Alto Semiconductor) only develops chip designs that are actually built by other companies. One of PA Semi’s investors is Texas Instruments, a chip fabricator that both designs its own processors, such as the ARM-based OMAP series that dominate the smartphone market, and builds chips designed by others, including Sun’s SPARC. This suggests that TI is the primary fab for PA Semi’s chip designs, although this has not been publicly announced.

PA Semi licensed IBM’s Power architecture technology to design its PWRficient series of 64-bit processors suitable for use in applications from desktop computers to server storage controller backplanes. Last spring, the ghost of Amiga, Inc. announced plans to build a new desktop system based on PA Semi’s PWRficient 2GHz PA6T-1682M dual core 64 bit CPU.

The new PWRficient processor has also gained high profile customers from real companies, from storage magnate NEC to server maker Mercury to aerospace and defense titans Lockheed Martin and Raytheon. A report by Rick Merritt in EETimes stated that the processors have uncharacteristically been rapidly adopted for use in more than ten defense systems designed into US Department of Defense programs currently in use “in every major branch of the armed services.” Earlier in the year, PA Semi bragged that a hundred companies had expressed interest in its new processors.

Why Would Apple Want PA Semi?
Given that Apple migrated away from the Power architecture in its desktop Macs in favor of Intel’s x68 Core processor, and has used ARM processors in its mobile devices from the 2001 iPod to the iPhone, how does it make any sense that the company would buy up a chip designer that develops high end, yet highly efficient processors Apple is unlikely to ever use?

Apple just exited the server storage market with the termination of its Xserve RAID. Its server line sells to a relatively small population of primarily education institutions and video production companies. Apple has no products that really correlate with the flagship PWRficient PA6T-1682M.

When asked about the acquisition in its earnings conference call, Apple declined to answer the question. Instead, Apple CFO Peter Oppenheimer noted that the company acquires lots of smaller technology companies but that it does not “comment on our purpose or plans” for those acquisitions.

No PWRficient Macs.
While Apple’s Universal Binary architecture would theoretically allow the company to introduce Macs, MacBooks and Xserves based on the 64-bit dual core PWRficient processors, it is extremely unlikely that the company has any interest in doing so.

One of the main reasons for moving to Intel’s Core x86 processor family was to share economies of scale with the PC industry at large. Part of that relates to software compatibility, and in particular the ability of Macs to run Windows, both in a virtual environment such as Fusion or Parallels and natively via Apple’s Boot Camp configuration tool. This flexibility, or at least the advertising merits of the Mac’s potential to run Windows, has unquestionably boosted Mac sales and lowered barriers to adoption.

Apple is even seeing additional interest in its server line due to the development of server versions of VMWare and Parallels, which offer the ability to host multiple Windows, Linux, and Mac OS X Server environments on a single Xserve. A migration to any other processor architecture, even one with significant performance or power efficiency advantages, would only hamstring existing strategies to migrate from PCs to Macs. Apple is committed to Intel Core processors in its Mac computers and servers and is well served by Intel’s current lineup and future roadmap.

No PWRficient Mobiles.
Back in the late 80s, after growing dissatisfied with the scant options available in mobile processors (principally AT&T’s Hobbit chip), Apple partnered with its UK equivalent Acorn Computer to codevelop a new generation of mobile savvy processors known as the ARM architecture. Acorn used them in its desktops, Apple used them in the Newton, and today roughly 80% or more of mobile devices are based on ARM chips.

Apple discontinued its use of ARM processors with the Newton in 1998, but resumed in 2001 with the iPod. Today, all of its iPods, the iPhone, and Apple’s current AirPort base station products use ARM processors. The ARM architecture has developed into a vibrant marketplace, where multiple vendors compete to design and build processors based on technology licensed from ARM, from Texas Instruments’ OMAP line to the “system on a chip” products from Marvell and Samsung that Apple uses.

Apple currently has no need to look elsewhere for mobile processors; even if it did, it would soon have options from Intel’s new Atom line, which plans to offer a low power, x86 compatible chips that could be suitable for use in mobile devices like the iPhone as early as next year. Clearly, Apple didn’t buy PA Semi to go into business developing its own mobile alternative processor given the already cutthroat competition between ARM licensees and Intel’s Atom, both of whom would kill for Apple’s business.

On top of all that, PA Semi’s PWRficient chips, while sipping only a fifth of the power of a comparable Intel x86 or Freescale PowerPC processor in server applications, are not comparable to the low power mobile ARM processors Apple now uses nor the Atom mobile processors Intel promises for the future. Reconfiguring PA Semi to be a mobile chip designer would take several years; its first product took nearly a half decade to complete.

No PWRficient?
The lack of any obvious application of PWRficient in Apple’s products has already spooked PA Semi’s existing clients. EETimes reported that just two days prior to the announcement of its purchase by Apple, “PA Semi informed its customers it was being acquired and it could no longer guarantee supplies of its chips. The startup did not identify the acquiring company but said that company may be willing to supply the chip on an end-of-life basis, if it could successfully transfer a third-party license to the technology.”

The report then flatly reiterated that “PA Semi customers were told the acquiring company was not interested in the startup’s products or road map, but is buying the company for its intellectual property and engineering talent.”

Among PA Semi’s talent is founder Dan Dobberpuhl, who at DEC developed the pioneering T11 in 1981, the blazing fast Alpha in 1992, and the highly efficient StrongARM processors in 1995. After Compaq bought DEC in 1998, the StrongARM business was acquired by Intel and rebranded as XScale.

After pouring around $5 billion into Xscale to develop it into a mobile competitor to TI, Intel gave up and sold off the business to Marvell for $600 million in 2006. After failing miserably as an ARM licensee, Intel has attempted to build an efficient new mobile processor business around its low power x86 chips known as Silverthorne and Moorestown, an effort recently rebranded under the trademark Atom.

The problem for Intel so far has been that no customers have shown any interest in Atom outside of Microsoft’s UMPC partners, who haven’t managed to sell units in any serious quantity. Intel has targeted Apple as a possible suitor and even portrayed the iPhone as a candidate in its Atom roadmaps. Getting its Atom processors inside Apple’s iPhone, iPod Touch, and AirPort base stations would be a major coup for Intel, but current and future ARM processors already deliver stiff competition in cost, size, and efficiency.

In 1998, Dobberpuhl left Digital and founded SiByte, which developed the first integrated multicore systems on chip. His company sold to Broadcom for $2 billion in 2000. In 2003, Dobberpuhl left Broadcom to start PA Semi, recruiting top talent from Intel, Broadcom, and AMD.

How Does PA Semi Serve Apple?
Given that Apple currently has lots of competitive, commodity processors to choose from, both for its conventional computers and its emerging mobile devices, why would the company invest in a chip designer? One reason hinges on the intersection of differentiated diversity and shared similarity.

While Apple benefits greatly from the open markets for x86 and ARM chips and from relying on Intel to design much of its Mac logic boards, it no longer has any edge in components that it once held as a PowerPC user. Between 1994 and 2005, Apple experienced waves of efficiency, performance, and cost advantages over PC vendors tied to Intel’s x86 architecture, particularly as Intel delivered a string of disappointments from the Pentium Pro to Itanium to the blazing hot Pentium 4.

Those PowerPC advantages were difficult to maintain in competition with Intel’s vast economies of scale, particularly as Apple’s PowerPC partners lost interest in investing in the desktop market. AMD also contributed significantly to the x86 market by offering Intel credible competition and forcing Intel to abandon its unwieldy 64-bit processor plans and the megahertz myth of the Pentium 4 in favor of copying AMD’s own 64-bit implementation of x86. Once Intel laid out plans to design an AMD-64 compatible, efficient chip based on its earlier Pentium III architecture, Apple had no reason to keep waiting for new blood to squeeze from the PowerPC turnip.

Apple’s migration to Intel was particularly traumatic for PA Semi, which had been working to develop an efficient PowerPC processor that Apple could use in its laptops in place of the aging G4 chips from Freescale (Motorola) and IBM. The problem was that PA Semi wouldn’t have its new chip ready until well into 2007, while Intel’s new Core processor would be available in early 2006. Intel could also provide Apple with a full array of chips for everything from desktops to workstations to servers. Apple was able to complete a full transition of its entire product line within a year, well before PA Semi’s potential laptop processor would even become available. Apple also adopted a new low power Intel x86 processor for use in Apple TV.

PA Semi was simply gunned down by the juggernaut of Intel, and at this point, backtracking to salvage the work done to develop a workable PowerPC processor makes no sense. While Apple has no apparent use for PWRficient, there are other ways it could use PA Semi’s assets and engineers.

Differentiating Through Mac Hardware Acceleration.
While the transition to Intel has afforded Apple tremendous new opportunities, the downside to using commodity chips is that Apple’s roadmap is now closely tied to Intel’s. That means there are fewer surprises Apple can pull off and less differentiation between Macs and generic PCs. While it makes no sense for Apple to jump back into the PowerPC business for its own products, the company could use PA Semi to deliver specialized processors and chipsets to differentiate its product line.

Apple has invested heavily in building software tools that spin processor intensive tasks out to specialized hardware. Core Video and Core Graphics allow developers to harness the raw and often idle capacity of graphics processors without needing to specialize in GPU programming or dealing with the specifics of any particular graphics hardware. Mac OS X also offers an Acceleration Framework that intelligently targets other features in the hardware to speed things up, whether its the AltiVec unit on PowerPC or the SSE features of Intel’s processors.

That gives Apple the ability to add specialized hardware of its own design to its Mac products and fully exploit that potential with tight software integration, much of which is already in place. Because it would own the designs, Apple could establish significant performance advantages over commodity PCs in general tasks, and also deliver acceleration for specific tasks such as video and audio transcoding. Because Apple now sells millions of Macs per quarter, adding highly efficient hardware acceleration chips across the board would add only minimal cost while delivering a significant and tangible advantage that would be expensive and complicated for other PC vendors to match.

This would not only fuel the embarrassing shootouts that Steve Jobs loves to showcase, but also highlight the difficulty Microsoft has in herding its stray cat PC hardware partners when it comes to tight hardware and software integration. Just as the highly integrated hardware and software of the iPod and iPhone demonstrated how clumsy PlaysForSure and Windows Mobile were, additional Mac hardware acceleration, transparently available to developers, would further prove the Windows PC to be a third rate product.

Differentiating Through Specialized Embedded Chipsets.
Just as with the Mac, the other half of Apple’s business, its mobile device offerings including the iPod and iPhone, similarly lacks any need for an outright CPU replacement. However, Apple could use PA Semi’s assets to develop its own mobile chipsets, from wireless components to signal processors. Apple currently relies upon a series of vendors to develop these parts, and regularly moves between vendors to find the best prices and components possible.

However, this exposes Apple to easy copying. Anyone can take apart an iPod and duplicate Apple’s product by simply buying up the same off the shelf components and developing their own software. That exactly what Microsoft did, as noted by Mark Kaelin in a News.com feature titled “Cracking open the Microsoft Zune,” forwarded in by reader Kevin Marchand.

Kaelin pointed out, “On the inside, the Zune is remarkably similar to the iPod Nano. Many of the parts are exactly the same. The difference is that the interface chips and software in an iPod are made by Apple. I know some die-hard fans will protest, but the insides don’t lie. The Microsoft Zune is a few Apple chips from being an iPod.”

By developing more its own integrated components, Apple could potentially save money, support new proprietary features, and throw copycats off its trail and force them to develop their own devices from scratch. As Apple blazes into uncharted territories by accelerating its iPod line into a new series of WiFi mobile devices, cost savings, differentiated features, and difficult to copy designs will all become increasingly important. PA Semi’s hardware expertise can help in that regard.

Love the Players, Not the Game.
Of course, Apple could also be interested in PA Semi primarily for its 150 engineering employees, its patents, and its relationship with Texas Instruments, which apparently builds the processors PA Semi designs. By acquiring PA Semi, Apple could custom design its own ARM ’system on a chip’ designs for TI to build, forging an interactive partnership with TI on the mobile side that reflects its cooperative partnership with Intel on the laptop, desktop, and server side.

Additionally, those 150 engineers have experience in developing low level software that likely complements Apple’s efforts in both the mobile and desktop arenas. Apple is migrating toward LLVM as its development compiler, developing new applications for multitouch, and managing who knows what other engineering projects as it doubles it campus size and expands into new markets. Hiring a productive team of experts, described in the Microprocessor Report as “an all star group of designers,” is easier than recruiting 150 individuals and teaching them to work together.

Over the last five years, PA Semi has received $100 million of venture capital investment, making Apple’s $278 million purchase seem like a steal. Prior to Apple’s move to Intel, announced mid 2005, Apple worked closely with PA Semi to determine the viability of its chip design due a year or so later than Intel’s Core processors. The tech press saw so much potential between Apple and the advertised specs for PA Semi’s new PowerPC chip that many couldn’t get over the idea that Apple chose to work with Intel instead.

In late 2005, Jon Stokes of Ars Technica wrote that in view of IBM’s promises of a faster desktop CPU and PA Semi’s planned PWRficient chip, Apple “jumped ship because they no longer care about making leading-edge computer hardware” and that subsequently, “the Mac line is no longer the foundation for Apple’s future growth, and it could very well go bye-bye.”

Instead, powered by Intel processors, Apple has seen phenomenal growth in Mac unit sales. That has helped the company amass a $19 billion cash pile that allowed Apple to snatch up the team at PA Semi at a bargain basement price in the midst of a recession. Apple is having its cake and eating it too.