Want to know what makes the iphone so ridicu­lously prof­itable? It’s not just the apps, cam­era, or de­sign. It’s the multi­bil­lion­dol­lar, eightyear bet Steve Jobs made on this chip. A peek in­side the undis­closed lo­ca­tions where Ap­ple makes its brains


Bloomberg Businessweek (North America) - - Front Page - BY BRAD STONE, ADAM SATAR­I­ANO & GWEN ACK­ER­MAN

Alit­tle over a year ago, Ap­ple had a prob­lem: The ipad Pro was be­hind sched­ule. El­e­ments of the hard­ware, soft­ware, and ac­com­pa­ny­ing sty­lus weren’t go­ing to be ready for a re­lease in the spring. Chief Ex­ec­u­tive Of­fi­cer Tim Cook and his top lieu­tenants had to de­lay the un­veil­ing un­til the fall. That gave most of Ap­ple’s en­gi­neers more time. It gave a lit­tle-known ex­ec­u­tive named Johny Srouji much less.

Srouji is the se­nior vice pres­i­dent for hard­ware tech­nolo­gies at Ap­ple. He runs the divi­sion that makes pro­ces­sor chips, the sil­i­con brains in­side the iphone, ipad, Ap­ple Watch, and Ap­ple TV. The orig­i­nal plan was to in­tro­duce the ipad Pro with Ap­ple’s tablet chip, the A8X, the same pro­ces­sor that pow­ered the ipad Air 2, in­tro­duced in 2014. But de­lay­ing un­til fall meant that the Pro would make its de­but along­side the iphone 6s, which was go­ing to use a newer, faster phone chip called the A9.

This is the stuff that keeps tech­nol­ogy ex­ec­u­tives up at night. The ipad Pro was im­por­tant: It was Ap­ple’s at­tempt to sell tablets to busi­ness cus­tomers. And it would look fee­ble next to the iphone 6s. So Srouji put his en­gi­neers on a crash pro­gram to move up the roll­out of a new tablet pro­ces­sor, the A9X, by half a year. The en­gi­neers fin­ished in time, and the Pro hit the mar­ket with the faster chip and a 12.9- inch dis­play packed with 5.6 mil­lion pix­els.

Srouji was nicely re­warded for his ef­forts. In De­cem­ber he be­came the new­est mem­ber of Cook’s man­age­ment team and re­ceived about 90,000 ad­di­tional shares of Ap­ple stock, which vest over a four-year pe­riod.

He also stepped into the kind of spot­light he’s avoided since join­ing Ap­ple in 2008. Srouji runs what is prob­a­bly the most im­por­tant and least un­der­stood divi­sion in­side the world’s most prof­itable com­pany. Since 2010, when his team pro­duced the A4 chip for the orig­i­nal ipad, Ap­ple has im­mersed it­self in the costly and com­plex sci­ence of sil­i­con. It de­vel­ops spe­cial­ized mi­cro­pro­ces­sors as a way to dis­tin­guish its prod­ucts from the com­pe­ti­tion. The Ap­ple-de­signed cir­cuits al­low the com­pany to cus­tom­ize prod­ucts to per­fectly match the fea­tures of its soft­ware, while tightly con­trol­ling the crit­i­cal trade-off be­tween speed and bat­tery con­sump­tion. Among the com­po­nents on its chip (tech­ni­cally called a “sys­tem on a chip,” or SOC) are an im­age sig­nal pro­ces­sor and a stor­age con­troller, which let Ap­ple tai­lor use­ful func­tions for tak­ing and stor­ing pho­tos, such as the rapid­fire “burst mode” in­tro­duced with the iphone 5s. En­gi­neers and de­sign­ers can work on fea­tures like that years in ad­vance with­out pre­ma­turely no­ti­fy­ing ven­dors— es­pe­cially Sam­sung, which man­u­fac­tures many of Ap­ple’s chips.

At the cen­ter of all this is Srouji, 51, an Is­raeli who joined Ap­ple af­ter jobs at In­tel and IBM. He’s compact, he’s in­tense, and he speaks Ara­bic, He­brew, and French. His English is lightly ac­cented and, when the sub­ject has any­thing to do with Ap­ple, non­spe­cific bor­der­ing on koan like. “Hard is good. Easy is a waste of time,” he says when asked about in­creas­ingly thin iphone de­signs. “The chip ar­chi­tects at Ap­ple are artists, the en­gi­neers are wizards,” he an­swers an­other ques­tion. He’ll elab­o­rate a bit when the topic is gen­eral. “When de­sign­ers say, ‘This is hard,’ ” he says, “my rule of thumb is if it’s not gated by physics, that means it’s hard but doable.”

Srouji re­cently spent sev­eral hours with Bloomberg Busi­ness­week over sev­eral days and guided a tour of Ap­ple chip fa­cil­i­ties in Cu­per­tino, Calif., and Her­zliya, Is­rael. This was, no doubt, strate­gic. In­vestors have bat­tered Ap­ple stock over the past year, send­ing it down more than 25 per­cent. Most peo­ple are al­ready pretty sat­is­fied with their phones, the crit­i­cism goes, and aren’t com­pelled to spend an ad­di­tional few hun­dred bucks on an upgrade. ( In March, Ap­ple in­tends to an­nounce an up­dated ipad and smaller-screen iphone fea­tur­ing the lat­est A9x and A9 chips, ac­cord­ing to a per­son fa­mil­iar with the plans, who


wasn’t au­tho­rized to com­ment pub­licly.)

Ap­ple’s usual re­sponse is to point to Jony Ive and his team of fas­tid­i­ously cool, Wal­labee- shod in­dus­trial de­sign­ers, or to high­light el­e­gantly tooled alu­minum or an app or some new fea­ture or gad­get. There’s al­ways some­thing new to show off. But none of that has ever ex­plained any­thing about a cru­cial part of Ap­ple’s profit ma­chine: its chips.

“I think it’s too good of a story not to be told at this stage,” Srouji says. “Hope­fully, we won’t re­veal too much.”

When the orig­i­nal iphone came out in 2007, Steve Jobs was well aware of its flaws. It had no front cam­era, measly bat­tery life, and a slow 2G con­nec­tion from AT&T. It was also un­der­pow­ered. A for­mer Ap­ple en­gi­neer who worked on the de­vice said that while the hand­set was a break­through tech­nol­ogy, it was lim­ited be­cause it pieced to­gether com­po­nents from diff e rent ven­dors, in­clud­ing el­e­ments from a Sam­sung chip used in DVD play­ers. “Steve came to the con­clu­sion that the only way for Ap­ple to re­ally dif­fer­en­ti­ate and de­liver some­thing truly unique and truly great, you have to own your own sil­i­con,” Srouji says. “You have to con­trol and own it.”

One of Jobs’s trusted ad­vis­ers, Bob Mans­field, Ap­ple’s top hard­ware ex­ec­u­tive at the time, re­cruited Srouji to lead that ef­fort. Srouji, then at IBM, was a ris­ing star in the ar­cane world of semi­con­duc­tor en­gi­neer­ing. Mans­field promised him an op­por­tu­nity to build some­thing from scratch.

The de­ci­sion to de­sign semi­con­duc­tors was risky. About the size of a small postage stamp, the mi­cro­pro­ces­sor is the most im­por­tant com­po­nent of any com­put­ing de­vice. It does the work that makes play­ing games, post­ing to Face­book, send­ing texts, and tak­ing pic­tures seem easy. Small cur­rents of en­ergy move from the bat­tery through hun­dreds of mil­lions of tiny tran­sis­tors, trig­ger­ing com­mands and re­sponses in nanosec­onds. It’s like an in­tri­cate city de­sign that fits on the tip of your fin­ger. When the chip isn’t do­ing its job ef­fi­ciently, the de­vice feels slug­gish, crashes, or makes users want to throw it against a wall.

If there’s a bug in soft­ware, you sim­ply re­lease a cor­rected ver­sion. It’s dif­fer­ent with hard­ware. “You get one tran­sis- tor wrong, it’s done, game over,” Srouji says. “Each one of those tran­sis­tors has to work. Sil­i­con is very un­for­giv­ing.” Among com­puter and smart­phone mak­ers, in­dus­try prac­tice is to leave the pro­ces­sors to spe­cial­ists such as In­tel, Qual­comm, or Sam­sung, which sink bil­lions into get­ting the chips right and mak­ing them in­ex­pen­sively. ( Ap­ple used to co- de­sign pro­ces­sors for the Mac­in­tosh, but Jobs aban­doned the work in 2005 in fa­vor of more pow­er­ful mod­els from In­tel, whose chips still power all Macs.)

When Srouji joined Ap­ple, the com­pany had a group of about 40 en­gi­neers work­ing on in­te­grat­ing chips from var­i­ous ven­dors into the iphone. That grew by about 150 peo­ple in April 2008, af­ter Ap­ple ac­quired a Sil­i­con Val­ley chip startup called P. A. Semi, which had a power- ef­fi­cient semi­con­duc­tor de­sign. Srouji’s team found it­self in­ter­act­ing reg­u­larly with other de­part­ments, from soft­ware pro­gram­mers, who wanted chip sup­port for new fea­tures, to Ive’s in­dus­trial de­sign­ers, who wanted help mak­ing the phones flat­ter and sleeker. An en­gi­neer who sat in on Srouji’s meet­ings re­mem­bers se­nior man­agers pre­par­ing ex­ten­sively for pre­sen­ta­tions, be­cause his sup­port was crit­i­cal for get­ting new fea­tures ap­proved. He was known for pep­per­ing en­gi­neers with tech­ni­cally so­phis­ti­cated ques­tions, par­tic­u­larly about con­tin­gency op­tions if some­thing didn’t work out as planned. He’d ask, for ex­am­ple, if a dif­fer­ent form of plas­tic could be used that wouldn’t in­ter­fere with an­other com­po­nent.

The first pub­lic signs of Srouji’s work came in 2010 with the de­but of the ipad and iphone 4. The pro­ces­sor, the A4, was a mod­i­fied ver­sion of a de­sign from ARM Hold­ings, a Bri­tish com­pany that li­censes mo­bile tech­nol­ogy. The A4 was de­signed to power the hand­set’s new high-def­i­ni­tion “retina dis­play.” Srouji says it was a race to get that first sys­tem-on-a-chip pro­duced. “The air­plane was tak­ing off, and I was build­ing the run­way just in time,” he says.

Over the next few years, Ap­ple kept mak­ing im­prove­ments to its de­signs, in­tro­duc­ing chips to ac­com­mo­date fin­ger­print iden­ti­fi­ca­tion, video call­ing, and Siri, the iphone’s voice-ac­ti­vated as­sis­tant, among other en­hance­ments. When the com­pa­nies us­ing Google’s An­droid op­er­at­ing sys­tem started mak­ing tablets, they mostly used con­ven­tional mo­bile phone pro­ces­sors. Start­ing with the third- gen­er­a­tion ipad in 2012, Srouji’s team de­signed spe­cific chips ( the A5x and A6x) to give the tablet the same pixel-pack­ing

high-def­i­ni­tion screens as the iphone.

Th­ese mys­te­ri­ous semi­con­duc­tors com­ing from Ap­ple were the cu­rios­ity of the tech in­dus­try, but it wasn’t un­til the re­lease of the iphone 5s in 2013 that ri­vals re­ally started to pay at­ten­tion. The phone fea­tured the A7 pro­ces­sor, the first smart­phone chip with 64 bits—dou­ble the 32-bit stan­dard at the time. The new tech­nol­ogy al­lowed for en­tirely new fea­tures, such as Ap­ple Pay and the Touch ID fin­ger­print scan­ner. De­vel­op­ers had to re­write ap­pli­ca­tions to ac­count for the new stan­dard, but it gave way to smoother maps, cooler video games, and gen­er­ally more re­spon­sive apps that don’t hog as much mem­ory. (Ap­ple’s con­trol over hard­ware and soft­ware is also use­ful for en­crypt­ing ev­ery­thing on the de­vice, a ca­pa­bil­ity that has landed the com­pany in a con­tro­versy: On Feb. 16, a judge ruled that Ap­ple must help the FBI un­lock an iphone owned by one of the San Bernardino shoot­ers. Ap­ple is fight­ing the or­der, say­ing it would set a prece­dent that would un­der­mine the pri­vacy of all its cus­tomers.)

Qual­comm, then as now the big­gest de­signer of phone chips, made the ex­pen­sive de­ci­sion to scrap de­vel­op­ment of its 32-bit chips and put all its re­sources into catch­ing up. Hand­set com­pa­nies all “wanted the shiny new thing,” says Ryan Smith, the ed­i­torin- chief of Anand tech, a web­site that pub­lishes ex­haus­tive re­views of semi­con­duc­tor de­signs. “The A7 re­ally turned the world up­side down.”

Srouji can’t re­strain a smile when re­call­ing com­peti­tors’ re­ac­tions to Ap­ple’s 64-bit sur­prise. “When we pick some­thing,” he says, “it’s be­cause we think there’s a prob­lem that no­body can do, or there is some idea that’s so unique and dif­fer­en­ti­at­ing that the best way to do it is you have to do it your­self.”

Srouji was born in Haifa, a port city in north­ern Is­rael. He was the third child of four. His fam­ily was Chris­tian Arab, a mi­nor­ity within a mi­nor­ity in the Jewish state. “Haifa is one of the most in­te­grated cities in Is­rael,” he says. “You have Chris­tians, you have Mus­lims, Jews, Bahá’ís, you have any re­li­gion you want, and ev­ery­one lives to­gether in peace­ful har­mony. In­te­gra­tion worked for me.”

Srouji’s father owned a metal pat­tern­mak­ing busi­ness out­side the city, and from age 10, Srouji spent week­ends and sum­mers help­ing him pat­tern wooden mold­ings that were used to make en­gine parts, med­i­cal equip­ment, and other ma­chin­ery. His father had an un­usual phi­los­o­phy: He would un­der­charge cus­tomers for com­pli­cated work while over­charg­ing for eas­ier jobs. “If there was a very com­plex thing that he’d never done, he wanted to do it,” Srouji says.

His father, who died in 2000, con­stantly re­minded him not to get com­fort­able in the fam­ily busi­ness. Education was more im­por­tant. In high school, Srouji got per­fect grades in math, physics, chem­istry, and sci­ence. He was in­tro­duced to com­put­ers by an in­struc­tor who also taught at the nearby Tech­nion Is­rael In­sti­tute of Tech­nol­ogy, one of the world’s top en­gi­neer­ing schools. “I fell in love,” Srouji says.

He en­rolled at the Tech­nion, spent late nights in the com­puter lab draft­ing out code in pen­cil, and earned un­der­grad­u­ate and mas­ter’s de­grees in com­puter sci­ence. His mas­ter’s the­sis was on new tech­niques for test­ing soft­ware and hard­ware sys­tems. “At the time it was very pro­gres­sive,” says Orna Berry, gen­eral man­ager of the EMC Cen­ter of Ex­cel­lence in Is­rael and cor­po­rate vice pres­i­dent of in­no­va­tion, who met Srouji while he was at the Tech­nion. “I’m not sur­prised he is where he is.”

Af­ter grad­u­at­ing, Srouji got a job with IBM, which had placed its largest non-U. S. re­search fa­cil­ity in Haifa, the bet­ter to at­tract the big brains com­ing out of the Tech­nion and other Is­raeli univer­si­ties. He re­searched dis­trib­uted sys­tems, an emerg­ing field in which com­put­ers in dif­fer­ent lo­ca­tions are net­worked to­gether to com­plete com­pu­ta­tion­ally in­ten­sive as­sign­ments. En­sur­ing the ma­chines com­mu­ni­cated cor­rectly re­quired skill build­ing hard­ware and writ­ing soft­ware al­go­rithms.

“Some­times I won­dered— when he got an as­sign­ment and within a day it was com­plete and per­fect—if he was bril­liant or just didn’t sleep at night,” says Srouji’s first boss, Oded Cohn, vice pres­i­dent and lab di­rec­tor for IBM Haifa Re­search. “In some cases, the con­clu­sion was both.”

Al­though Is­rael grap­ples with Jewish-Arab ten­sions all the time, none of it mat­tered in Srouji’s world. Cohn, who re­mains friends with him, says their dif­fer­ent back­grounds never came up. “Tech­ni­cal peo­ple treat tech­ni­cal peo­ple based on per­son­al­ity and tech­ni­cal abil­ity,” he says. “You don’t think about it. You just work to­gether. The rest goes away.”

In 1993, Srouji left IBM for In­tel, where he cre­ated tech­niques for run­ning sim­u­la­tions that test the strength of semi­con­duc­tor de­signs. Dur­ing a visit to the U.S. in 1999, he used a 20-minute car ride with a man­ager, fel­low Is­raeli Uri Weiser, to lobby for a three-year stint at In­tel’s re­search hub in Austin. As­sum­ing Srouji was also Jewish, Weiser in­vited him to an Is­raeli Me­mo­rial Day


cel­e­bra­tion at a syn­a­gogue in Texas.

“He looked at me and said, ‘I’m a Chris­tian Arab,’ ” re­calls Weiser, who gave Srouji the Texas as­sign­ment. “I said, ‘Well, come and join and learn about your en­vi­ron­ment,’ and he said OK. He was there sit­ting with a kip­pah in the syn­a­gogue and fol­low­ing ev­ery­thing.”

Srouji lives a few miles from Ap­ple’s head­quar­ters at One In­fi­nite Loop, Cu­per­tino. He drives a black Mercedes- Benz and re­laxes by lift­ing weights and rid­ing his bike on week­ends. He smiles eas­ily, warmly touches a reporter’s shoul­der when shar­ing a laugh, blushes at com­pli­ments, and ab­so­lutely clams up when he’s asked about any­thing that could re­motely be con­sid­ered a cor­po­rate se­cret. “I don’t want to go into too much de­tail on that” is a com­mon re­frain.

Friends have no­ticed the height­ened dis­cre­tion. Srouji once in­vited his for­mer In­tel col­league, Weiser, to give a speech about chip de­vel­op­ment at Ap­ple head­quar­ters in Cu­per­tino. Af­ter the pre­sen­ta­tion, an as­sis­tant es­corted Weiser to Srouji’s empty of­fice, where he no­ticed that the pa­pers on the desk were all turned up­side down. Then Srouji en­tered the room and told Weiser he had to move. “He said, ‘We are at Ap­ple, you can’t sit here,’ ” Weiser re­calls. “He of­fered me to sit with his sec­re­tary and said, ‘If you want to go to the bath­room, she will es­cort you.’ ”

One morn­ing in Fe­bru­ary, Srouji con­ducts a brief tour of his do­main, which is scat­tered in un­marked lo­ca­tions around Sil­i­con Val­ley. A shut­tle bus leaves One In­fi­nite Loop and drives 10 min­utes through a se­ries of res­i­den­tial neigh­bor­hoods to a low-rise of­fice build­ing near the Santa Clara city lim­its.

One of his deputies greets Srouji at the bus and badges through sev­eral locked doors into a room where fu­ture chip de­signs are be­ing tested. The build­ing is eerily quiet and still, save for the hum of air con­di­tion­ers and the blink­ing red and green lights of large dark boxes that are stacked to­gether and re­sem­ble Zam­bo­nis. The room is Ap­ple-white and clean, but not tidy; thick wires and large plugs lie around. Old, un­used Macs are lined up on a shelf like books that have al­ready been read. All the equip­ment is op­er­ated re­motely. The boxes are run­ning soft­ware that scans for pos­si­ble flaws in the chip ar­chi­tec­ture. Test­ing pro­ceeds for sev­eral days on one el­e­ment of the chip, then moves on to the next, and then the next, un­til the process is done, which can take months. “We beat the sil­i­con as much as we can,” Srouji says. “If you’re lucky and rig­or­ous, you find the mis­takes be­fore you ship.”

In an ad­ja­cent room, cir­cuit boards are wired to­gether in milk car­ton- size stacks to sim­u­late the ca­pa­bil­i­ties of a fu­ture iphone or ipad. Ap­ple’s soft­ware pro­gram­mers, sit­ting any­where in the world, can re­motely test how their code holds up against a fu­ture chip de­sign.

Then the shut­tle takes Srouji a few more miles away, to an­other un­marked build­ing, where rows of cus­tom­ized Mac Mi­nis are test­ing pro­to­type chips un­der var­i­ous tem­per­a­ture and pres­sure con­di­tions. Stand­ing in an aisle, sur­rounded by ex­posed cir­cuit boards and dig­i­tal in­nards, is like be­ing in­side the Matrix. “No one has seen this be­fore,” Srouji says.

Ev­ery­thing looks ex­ceed­ingly com­pli­cated. Srouji won’t dis­cuss costs, but Ap­ple’s re­search and de­vel­op­ment ex­penses hit $8.1 bil­lion last year, up from $6 bil­lion in 2014 and $4.5 bil­lion in 2013. Many an­a­lysts at­tribute the rise in large part to chip de­vel­op­ment. All Srouji will say about his bud­get is that Cook doesn’t scru­ti­nize it. “I run it very tight,” he says. “I truly be­lieve that en­gi­neers will do their best when they are con­strained by ei­ther money, tools, or re­sources. If you be­come sloppy be­cause you have too much money, that’s the wrong mind­set.”

Ap­ple isn’t com­pletely in charge of its own des­tiny. It re­mains in many ways a pris­oner of its sup­ply chain. Dis­plays come from Sam­sung, and cel­lu­lar modems from Qual­comm. Sam­sung and TSMC, based in Tai­wan, still man­u­fac­ture the pro­ces­sors. Ap­ple’s abil­ity to keep up with de­mand is in part de­pen­dent on the pro­duc­tion ca­pac­ity of those com­pa­nies. It also lags be­hind Sam­sung in some ar­eas of chip de­vel­op­ment, such as adding a mo­dem to the cen­tral pro­ces­sor to con­serve space and power and tran­si­tion­ing from a 20-nanome­ter chip de­sign to a more compact 16-nanome­ter for­mat, which means even more tran­sis­tors can be crammed into a smaller space. “If I was just ar­gu­ing hard­ware and not Ap­ple’s mar­ket­ing, I would say Sam­sung has the best pro­ces­sor,” says Mike Dem­ler, a se­nior mo­bile chips an­a­lyst at the Lin­ley Group, a tech­nol­ogy con­sult­ing firm in Sil­i­con Val­ley.

Or Ap­ple could just be get­ting started. It re­lies on sup­pli­ers for Wi-fi modems now, but will it for­ever? “I don’t want to go into Wi-fi specif­i­cally,” Srouji says.

Ap­ple could also take a page from Tesla’s play­book and start de­vel­op­ing its own bat­ter­ies. “I don’t want to get into bat­ter­ies too deeply,” he says.

And since Ap­ple is do­ing a fine job with mo­bile pro­ces­sors, it could con­ceiv­ably de­cide to get into con­ven­tional chips and bump In­tel out of its Mac lap­tops and desk­tops. Srouji, of course, won’t go there, though he does al­low that his team’s mis­sion is fi­nite. “If we at­tempt to do ev­ery­thing on the planet,” he says, “I don’t think that would be very smart.”

Chip- dura­bil­ity test­ing at an un­marked Ap­ple lab in Cu­per­tino

In Is­rael, Srouji ( se­cond from left) and Cook (right) with Ap­ple em­ploy­ees

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.