Editorial: Apple's Ax SoC move from Samsung to TSMC can't happen fast enough
Justice delayed is competition demanded
Any efforts to make changes to its complex, delicate supply chain, particularly in the realm of sourcing its customized Ax chips, would take years to plan and require billions of dollars to execute. While lawyers from Apple and Samsung argued their positions in court over many months, Apple's executive team wasn't willing to bet that the courts would resolve matters satisfactorily.
Apple's plan B for its Ax chips was further complicated by the fact that there are a very limited number of fabs on earth with the capacity and technology to build the parts the company needs next year and the year after that. Chip fabs are so expensive to maintain that they simply can't afford to sit around with idle capacity, waiting for customers. Finding a new Samsung was not going to be easy.
The global economic slowdown Apple was coping with as it rolled out successive generations of the iPhone was having an even stronger impact on his competitors. While Apple's own sales resisted the predicted effects of "global macroeconomic conditions," which many analysts had insisted would throw it off course, slack projections of sales for other electronics producers had slowed the global expansion of chip production facilities.
While Apple continued to coordinate work with Samsung in 2011 to produce the sophisticated new A6 at the fab's newer, 32nm logic process for last year's iPhone 5, along with an even more sophisticated A6X version for the iPad 4, it was also busy working to line up the production and supply of a successive generation of new 28nm chips. But not with Samsung.
Instead, Apple turned to TSMC, a fab with the technology to pull off the next Ax chip but lacking Samsung's competitive arms in smartphones and other electronics.
TSMC had already produced iPhone and iPad chips as the foundry for Broadcom, CSR, Cirrus Logic and Qualcomm, but speculative rumors that the fab might take over production of Apple's Ax SoCs first started in early 2011.
Apple was said to be booking rush orders to test TSMC's ability to produce reliable yields and quantities of its existing chip designs throughout the year, and by September 2011, it was reported that Apple had actually signed TSMC on to produce its future chips.
How fast can you build a fab?
"How long does it take to go from a muddy field to full 28nm capacity?" asked Paul McLellan in an article on SemiWiki.
McLellan noted that TMSC's vice president of operations outlined that company's ramp to production of 28nm chips "was the fastest TSMC has ever done," with the firm's build out being accelerated from one new fab construction phase per year to three phases being orchestrated in parallel each year.
Source: TSMC, via SemiWiki
The company's first 28nm fab "was a muddy field in Taichung" in the summer of 2010. One year later, the building and its clean rooms were completed in phase 1. Another ten months were spent installing and qualifying equipment in phase 2. By the winter quarter of 2012, the fab had ramped up to a production speed of 50,000 wafers per month.
Next month, the fab will begin a new phase that will double its production to 100,000 wafers per month. Apple reportedly wanted to buy all of TMSC's production capacity, but will have to settle for as much as it can afford.
TSMC's capacity to ship 28nm parts was reportedly strained because its existing customers, spooked by the recession, initially reported a major drop in their component needs. However, demand has actually turned out to be much greater than they forecast, resulting in subsequent insufficient manufacturing capacity and a constrained supply for everyone.
Fortunately for Apple, it's making three quarters of the mobile industry's profits, enabling it to plan and build for its capacity needs. Going forward, TMSC plans to move to even smaller 20 and 16nm fab processes even faster, and is also working toward production of larger yielding, 450mm wafers.
In addition to its fabs in Taiwan, TSMC is also believed to be the company behind what is known as "Project Azalea," a mysterious American development thought to be a chip manufacturing facility connected to Apple.
Again, the efforts to build, equip and ramp up production at new fabs takes years to complete. This indicates that once Apple completes its transition to TSMC, it will enjoy a relief period where it can redirect the energy currently being lost in its tense relationship with Samsung and its efforts to transition its business elsewhere, and refocus those efforts on rapidly scaling the sophistication and efficiency of new Ax chips.
Once its investments in new chip production begin to bear fruit, Apple appears poised to drive a huge leap in ARM-based chips, which will begin to occur just as the conventional x86 PC market reaches its lowest, weakest point ever. No wonder the company is spending billions to proceed with its fab plans at a blistering pace.