CONNECT WITH US

The pros and cons of Apple building its own chips

Jessie Shen, DIGITIMES, Taipei 0

If Apple builds its own silicon at its own chip fab, the risk it will face will be greater than the benefits for the company, according to Nobunaga Chai, analyst for semiconductors at Digitimes Research. But if Apple has to own a fab, either Globalfoundries' Fab 8 in New York or IBM's 300mm (12-inch) chip plant - B323, also located in the same state - should be more favorable choice for Apple than any other fabs, said Chai.

Chai was commenting on recent rumors that Apple may be looking to take full or partial ownership of a fab.

But Chai noted it is unlikely that Apple is targeting TSMC's 12-inch facilities. For TSMC, having strategic investment from Apple will likely hurt the foundry's existing partnerships with Qualcomm and many other mobile processor developers.

Globalfoundries will strive for a big customer like Apple at any cost, Chai believes. Globalfoundries is a relatively newcomer to the foundry industry and still needs to build up its customer portfolio.

As for IBM's 12-inch fab, Chai pointed out that IBM has in recent years moved towards a fab-lite strategy. Besides, acquiring IBM's facility could allow Apple to gain access to IBM's advanced chip-making technologies, Chai said.

Owning a fab will definitely give Apple greater control over its supply chain and help shorten time-to-market as a result of an integrated design and manufacturing approach, in addition to ensuring a steady supply of mobile chips during periods of high demand, Chai indicated. Nevertheless, these beneifts will only exist when the fab is able to produce advanced chips in commercial volumes.

Apple will also need to brace itself for higher operating costs if the company decides to produce its mobile processors in-house, Chai pointed out. As the industry is about to enter the sub-20nm era, chipmakers are encountering technical and financial challenges. If Apple moves to operate a fab, the company will face considerable risks and uncertainty especially in the migration to 1Xnm process manufacturing, Chai said.

In fact, many IDMs such as Renesas and STMicroelectronics have pursued a fab-lite strategy, Chai noted. Qualcomm, at least for now, would not want to run its own chip plant, Chai said.

Chai also does not think Apple is going to buy a UMC fab. The Taiwan foundry's 12-inch manufacturing capability is not even ready to fabricate Apple's current A5 processors, Chai said.

UMC is expected to move its 28nm process technology to commercial production as early as fourth-quarter 2013. In terms of process technology, UMC is already lagging behind TSMC, and even Globalfoundries and Samsung, Chai said.

Rather than running a wafer plant, Apple should stay focused on its core mission and work with its foundry partner/s, Chai suggested. Companies with advanced manufacturing capabilities are the ideal partners for Apple, especially those which are now engaged in the development of EUV and 450mm manufacturing and will certainly be capable of making sub-10nm devices, Chai said.

Chai commented that if Apple is returning to Samsung as the primary producer of A-series chips, it would make no sense for Apple to place 20nm chip orders with TSMC.

Korea Economic Daily cited unnamed sources as saying in its July 15 report that despite the loss of 20nm chip orders from Apple, Samsung already signed a new deal with Apple to supply A9 chips that would be built on its 14nm FinFET process.

A Digitimes report in June 24 cited industry sources as claiming that TSMC has struck an agreement with Apple to supply 20nm, 16nm and 10nm A-series chips, followed by more rumors suggesting Apple could be exploring possible deals with other foundry chipmakers.