Apple in China, page 11
The Pyramid
The fact that it was Cook, in Operations, who developed the relationship with Foxconn is important in understanding Apple’s product development cycle. The shape is a pyramid, with Jony Ive’s Industrial Design team at the top. ID would conceive how a product would look and feel, with an exactitude that drove fear into the rest of the organization. Their creation goes to Product Design, whose role is to figure out the mechanical engineering and make the product functional. Then it moves to manufacturing engineers—a small team back then, and later a distinct unit called Manufacturing Design, or MD—who work with component makers and final assembly partners to actually build the product. Finally, the Operations team comes in. They work with vendors that their colleagues have already selected, and then optimize everything to help the product scale. But crucially, Ops also has a prerogative to find other component makers, other final assembly partners, to build resiliency, cut prices, and to instill competition.
This model evolved over the course of more than a decade following the return of Jobs. When it worked well, it created discipline and pride. Each domain knew their role, their place in the hierarchy. Jony Ive’s group was put into an elevated position often described as godlike. Unless what ID asked for defied the laws of physics, whatever ID said was the way it would be. Saying no to ID was possible, but only if you had reams of experimentation behind you and the data to make your point. A common saying was: “You’d have to drive the car off the cliff to prove the brakes don’t work.”
But it’s not as though other divisions were simply taking marching orders. The respect went both ways. ID’s job was pushing the limits while conceiving only of products that were feasible to build. How they would be built was the role of PD and MD, where the ideal employees loved the steepness of the learning curve on every product. “It’s loads of hard work,” says a veteran of Ive’s ID studio. “But the engineer that you want to talk to is just like, ‘I have no fucking idea how to make that, but I’ll figure it out.’ ”
By empowering ID, Jobs had created a design-first organization that had zero tolerance for imperfection from the earliest stages of product creation. That approach cascaded down the pyramid, so the Ops team couldn’t accept imperfection, either. Mike Bell, a vice president in the 2000s, recalls that for the Titanium PowerBook, an especially difficult-to-engineer laptop released in 2001, Product Design engineers “had to come up with methods of production nobody had ever done before.” He characterizes this whole time period as “designing the impossible, then manufacturing the impossible.” Engineers would joke that Jobs had asked for “anti-gravity” or wanted them to make the next product out of “unobtanium.”
If something was too easy, ID would push the envelope even further. In the making of the Mac Mini desktop around 2004, one engineer recalls Ive asking if he could make the computer such and such a size. The engineer said he could. Ive narrowed the dimensions and asked if he could build that. The engineer said he could. So Ive minimized the dimensions again. This time the engineer said, “No, no, that would be really difficult.” And Ive said, “Great, those are the dimensions.”
The tricky thing about understanding the pyramid structure is that there were not firm barriers between the four divisions. It’s emphatically not the case that ID conceived of a product, threw it over the fence for PD to fit in the electronics, who then threw it over the fence for others to build the machinery and deliver the product at scale. Something like that was happening, but it was concurrent. The teams were in constant communication, so operational breakthroughs would feed back into design excellence.
But the pyramid structure could cause problems due to the inherent tension between teams, or their differing incentive structures. Operations was rewarded for low price and high volume, whereas the other teams were focused on quality. In its drive to push scale, the Ops team risked working with second-rate vendors. So when Cook chose to partner with Foxconn, a less experienced company just beginning to prove itself in the late nineties, he was going out on a limb. The project was initially kept secret from the ID and PD Apple engineers working with LG, as Ops first needed to bring Foxconn up to snuff. One engineer from PD recalls being aghast at the decision. “Foxconn had a bad reputation as a second-source copycat who stole other manufacturers’ designs and undercut them in price after production started to take over a chunk of production volume,” this person says. Engineers in PD complained that Foxconn didn’t understand Apple’s long development design cycle, but Ops loved Foxconn’s low pricing.
And this is where Foxconn made a big impression. Without any interaction with ID, Terry Gou’s team deconstructed an LG-made iMac into its constituent parts, worked out how it was assembled, then created or found parts and built their own version, including the CRT monitor. “Every part in that product was reverse-engineered and reverse-sourced by Hon Hai, to replace LG,” says an engineer on the iMac project. It was far from perfect, but it was good enough to demonstrate Foxconn’s competence. “When you build a tool, the moving parts have to come together, and ID had crafted those to be beautiful,” this engineer says. “Foxconn didn’t know about that, so they made something where the outside shape was right, but none of it matched what ID wanted.” At this point ID and PD were brought in to help, with teams sent to Longhua to give extensive training. “PD practically lived in Foxconn for a year,” says an Operations executive.
An Apple engineer recalls telling Foxconn: “If you guys want this business, you have to earn the business. You need to do all the same hard work that LG did to ensure that the tools meet the part quality, interchangeability, and everything, such that the final cosmetic fit and finish is consistent across each and every product, no matter how you mix them or where you ship them across the world.” It was a giant effort.
The Hub Model
Once Foxconn started to work on building the iMac at scale, Apple engineers were astonished at what became known as “China speed”—an ability to get things done at a rapid pace beyond the comprehension of Western visitors. It started with Terry Gou pledging that he would build the iMac tooling in just twenty-five days. “It was unprecedented. We’re used to twelve weeks for building tooling,” says Wayne Miller. “And sure enough, twenty-five days later, everything was there. The quality was amazing. They really impressed the hell out of us.”
At the time, Dell was the industry leader in manufacturing and efficiency, particularly with custom build-to-order models. When a Dell customer purchased a computer with a bunch of idiosyncratic options, Dell would take a standard unit off its production line, then place it on a separate line for rework. Foxconn quickly proved itself as an innovative force by managing to work out how to handle both types of orders on the same line—a significant improvement in efficiency. This innovation was born out of necessity. Apple’s design of the iMac in five colors was something no computer maker had ever done before. It would have been a mess to produce defined batches of, say, 100 Blueberry computers followed by 100 Strawberry variants. So Foxconn built an automated system. As orders for a particular color iMac came in, the information would be sent to a next-door molding factory—where Foxconn had huge machines to do the entire casing. Once completed, those shells were sent to the production line for final assembly, in the order they were needed.
Foxconn’s system could handle other tweaks to the configuration as well, such as different processor speeds and memory upgrades, so regardless of what the customer ordered, Foxconn could put it together without delay. The production lines Foxconn built were just for Apple, and around each assembly line, it formed a hub of sites to build everything else that was needed: the molding factory, plus production lines for the logic boards, keyboards, mice, and language kits. “The whole supply chain was built around the assembly line,” says an Apple employee who worked on setting it up. “Only when the order is triggered [would Foxconn] start to pull all the different components from the hub out and into the assembly line.” And once an iMac was assembled, it went immediately into Apple’s sales division. The system was designed so that Apple’s inventory was virtually zero—as soon as Foxconn delivered a finished iMac to Apple, it was placed en route to a customer. If there was no order, there was no inventory. Of course, Foxconn had to buy the necessary components in advance, but that was its problem, not Apple’s.
Foxconn Expansion
Once the Shenzhen line for iMacs was up and running, Foxconn established sites on two other continents. In Europe, Foxconn executive Jim Chang found a Soviet-era electronics site in Pardubice, a city of 100,000 people sixty miles east of Prague. The site had previously been run by a state-owned company called Tesla, whose specialty was radar systems and whose biggest client had been the government of Iran. The site had an eerie feel to it, like it had been hit by a neutron bomb. Forklifts stood motionless on the floor and cups of tea, their contents long gone cold, had been left on the tables. In May 2000, Foxconn was able to buy the plant for just 102 million CSK (€2.9 million), a fire-sale price because it was bringing in jobs. Foxconn also won from the government a ten-year tax holiday.
The company once again established its hub model for iMac production, bearing the up-front costs. “Apple didn’t own any of the inventory parts that were there. Everything was borne by Foxconn,” says a Czech-based Apple worker. “Apple had zero capital expense or investment.” More than 300 people were hired to run multiple assembly lines, and Foxconn established a giant molding site for the plastic enclosures. Apple teams from Singapore were sent in to train the workers and ensure quality. Apparently smitten with the country, Terry Gou purchased a twelve-bedroom castle near the Czech factory in 2002, for a reported $30 million, and began spending his summers there. The fabulously rich CEO no longer had to worry about tin roofs.
In California, Apple found Foxconn an old Compaq service center in Fullerton, twenty-five miles southeast of Los Angeles. One of Terry Gou’s two younger brothers, Tony, ran the site. It was less ambitious in scale than the other two factories, but it was an important location for build-to-order Power Macs, and Apple engineers could go there and work intimately with Foxconn to build prototypes. Later, in 2004, Apple laid off 235 workers in Sacramento, closed its plant there, and shifted all production to the new Foxconn site.
Strike Emergency
The experience in the Czech Republic was an important proving ground for Foxconn and its hub model, but what it really demonstrated was that producing hardware in China was cheaper, more efficient, and less subject to media scrutiny. In China, assembly got done at incredible speed and with few complaints. Workers did twelve-hour shifts and lived nearby in dorms. At the Czech site, workers put in fewer hours and were represented by a trade union; they protested conditions and spoke to the press. Plans to build dormitories met local criticism and were abandoned. Over the course of a decade, Foxconn expanded its work in the Czech Republic, continuing to build for Apple, adding another location, and taking on production for Hewlett-Packard, Sony, and Cisco. For Compaq alone, it was making 10,000 PCs a day in 2001. But Apple’s more complex designs caused problems. One Foxconn employee from the late 2000s calls the whole experience “a disaster,” even as the plant was given simpler products to assemble, including the Mac Mini.
At one point, according to an ex-worker named Andrea, workers making Apple products didn’t receive an annual bonus as they were promised, so they threatened a “strike emergency” just before the ramp-up ahead of Christmas. “Afraid,” the Foxconn managers deposited the bonuses within a week. The incident triggered an audit by Apple, which interviewed workers about their experience. Apple, Andrea said, advocated for better conditions, but “instead Foxconn closed the division within half a year and 330 people were dismissed.” Around the same time, in August 2009, Foxconn shut its Fullerton site, too.
How Foxconn laid the Czech workers off is worth highlighting. Mass dismissals—defined as laying off more than thirty people—need to be reported to the Labor Office, but it was important for Foxconn to avoid scrutiny. “What Foxconn did is they dismissed twenty-nine workers every month,” Andrea said. “Each month, regularly, they fired twenty-nine people.” The threat of a single strike ended all large-scale Apple assembly in Europe. But that was later. In the summer of 2000, Apple was still in a stage of experimentation. For the first time, Apple was overseeing major manufacturing operations in China. But it was one country of several. Apple products were also being assembled in America, Mexico, Wales, Ireland, the Czech Republic, Singapore, South Korea, and Taiwan. Within a decade, virtually all production would shift to China, with much of the final assembly done by a single partner, Foxconn. But the rapid consolidation was just beginning, and it didn’t yet feel inevitable.
CHAPTER 12 A FAREWELL TO MACTORIES
Steve Jobs held on to a hope that Apple could play a bigger role in manufacturing. In 2000, more than eighteen months after hiring Tim Cook to run operations, Jobs distributed T-shirts to staff emblazoned with the word Mactories. Apple had already significantly downsized its assembly operations in Singapore, Ireland, and Sacramento, but Jobs contemplated expanding them.
Tim Cook gradually persuaded Jobs on the merits of outsourcing—that it was cheaper, faster, nimbler, and most critically of all, could meet Apple’s quality expectations. But this wasn’t some overnight decision; the global outsourcing strategy dated back to 1996, preceding Cook’s arrival by nearly two years. The nail in the coffin was in late 2000, though Apple continued playing a role in assembly through 2003. Had the first three years of Steve Jobs’s comeback been smooth and lucrative, it’s possible he would have clamored more for Apple to retake direct control of its manufacturing. Instead, the late-1990s speculative mania in tech stocks abruptly reversed itself in March 2000, threatening the whole industry and causing the entire market to lose trillions of dollars in value.
Apple’s comeback story was dramatically rewritten when, on a single day in late September 2000, the company’s stock fell by more than half after it issued a profit warning. It was the second such warning in two years. In September 1999, the problem was too few computers: Apple didn’t have enough G4 chips from Motorola to meet demand. Now the problem was too few customers, since nobody was buying the G4 Cube, a brilliantly designed but expensive desktop computer ill-suited for the mass market. “We’ve clearly hit a speed bump,” Jobs told the public. The stock’s one-day collapse wiped out nearly all the gains made since Gil Amelio’s departure. A story called “Apple R.I.P.” in Forbes demonstrated how tenuous Apple’s future felt. “Even after all the successes of the last two years, Apple is still where it was during the Spindler era,” Forbes wrote. “It is a niche player, with an anomalous operating system, trying to survive in a market dominated by giant corporations like Compaq, Dell and IBM.” Gateway, another PC clone, even explored acquiring Apple. “The shit hit the fan, and our sales dropped off a cliff,” according to Rubinstein. “Before Gateway had gone off the cliff—but we had—Gateway came in and tried to acquire us.”
In that year’s holiday quarter—the fiscal first quarter of 2001—Apple posted a loss of $195 million. For all of 2001, total sales dropped by a third to $5.4 billion, the lowest revenues for Apple since 1989. With a US computer market share of 4 percent and a global share of 3 percent, Apple, declared Fortune, risked becoming “as inconsequential as Liechtenstein.” Whatever hopes Jobs may have had to bulk up Apple’s manufacturing capabilities, the company had no cash for such endeavors. The Singapore and Sacramento factories were soon shuttered. (Cork was about to be closed, too, until Apple learned it would receive major tax advantages for maintaining the facility.)
Hollowing Out
The dotcom crisis reshaped tech manufacturing for years to come. In the prior decade, leading contract manufacturers had begun to purchase factories from major brands including IBM, Texas Instruments, Ericsson, Siemens, and Lucent. The deals were often seen as a win-win, with the big brands saving on costs. Negative media headlines could be avoided, since factories were not being shuttered so much as being put under new management. When Apple sold its Fountain, Colorado, factory in 1996, most of its 1,100 employees simply got a new uniform. But the result was a huge transfer in practical knowledge from the computer brands to the contract manufacturers. In the 1990s, when these deals were happening within America or within Europe, they caused few alarms. But as the internet went mainstream, the idea was promulgated that borders didn’t matter. As companies competed on cost and scale, their search expanded internationally.
The 1990s had seen an incredible stock-market boom among US-headquartered contract manufacturers that had followed in the wake of SCI. There was massive consolidation as companies few people had ever heard of merged in a battle for scale. Solectron alone acquired some twenty companies in the decade, and in the five years leading up to 1999 its market value grew tenfold. Revenue at Flextronics, a rival, experienced average annual growth of 78 percent in the late nineties, aided by mergers.
Then the dotcom crisis triggered the tech industry’s biggest-ever recession. It caused bankruptcies, forced people to question the sustainability of the tech sector, and compelled organizations to rethink the value of in-house manufacturing. Producers of components were left holding huge inventories amid a sudden dip in demand. SCI became an attractive target amid the downturn and was acquired for $6 billion by Sanmina, a smaller San Jose–based rival.
As more manufacturing work was off-loaded overseas, the transfer of knowledge intrinsic to outsourcing underwent an important shift. It was no longer just company to company, it was country to country. At the start of the 2000s, the world’s four largest contract manufacturers were headquartered in the United States; the fifth was Canadian. At the time, it looked as though these companies’ rapid growth would continue to soar. The share of the circuit board and consumer electronics manufacturing sectors penetrated by contract manufacturers had grown only to 13 percent, “leaving a great deal of room for continued growth,” as one widely cited study forecast.
