I'm Feeling Lucky: The Confessions of Google Employee Number 59

by Douglas Edwards

  Things could have been worse. In fact, they could have been far worse.

  I rarely heard profanity in the halls of the Googleplex, where raised voices raised eyebrows. People could be infuriating, but the approved response was to bludgeon them with facts until they succumbed to superior logic. Not that Larry and Sergey couldn't be caustic, even to each other. David Krane, who traveled extensively with the founders, sums up their relationship this way: "Those guys had a communications channel that was very direct, very open. When there was tension, it was when they were fighting over data. They would be downright rude to each other, confidently dismissing ideas as stupid or naive or calling each other bastards. But no one would pout."

  It didn't take long to figure out that Larry and Sergey wanted to optimize the organization until it was as efficient as its technology. Anything that increased costs or reduced productivity added friction and slowed our progress. Ideally, we would churn out perfect products while incurring no expense. That was impossible, of course, but it didn't stop us from trying.

  Free Is Too Much

  "I negotiated a really great deal for this data," I bragged to Sergey as I presented a spreadsheet showing all the countries for which we would get traffic rankings once our research contract was signed. "They offered twenty percent off at first, but I hammered them down to half."

  Sergey poked and sniffed at the spreadsheet as if looking for chunks of meat in a vegetable stew.

  "I don't think this is going to be very useful data," he told me. "They don't have much experience in these markets, and they're trying to expand their customer base. Having Google as a client would give them some great credibility." I knew what came next. It was as inevitable as rain after a carwash. "They should really pay us," Sergey said.

  "They should pay us" was the starting position for any negotiation with a vendor. As Google grew in prominence, the benefit of being tied to our rising star increased exponentially—at least in Larry's and Sergey's minds. As our brand value approached infinity, our costs should go to zero. To the founders, winning a deep discount wasn't a victory; it was an admission of failure to get something for free. Their high expectations for low prices became a self-fulfilling prophecy, because I dreaded bringing them any proposal that shaved only ten or fifteen percent off the rack rate. "Never pay retail" was a Google core value. I was lucky, though. Marketing didn't buy services every day, so I didn't spend all my time beating suppliers into submission. It was different for facilities.

  By late spring of 2000, Google's rate of growth was accelerating on all fronts. Traffic was building, our advertising had caught on, and so many dot-com refugees claimed space in the Plex that the walls bulged every time someone exhaled. Lines began forming outside the men's room, leading an engineer to install a "take a number" machine next to the stalls. In oversaturated offices, the person closest to the door had to stand up when an office mate wanted to leave or enter. Corridors were so choked with desks and power cords and boxed computer components that a stroll down the hall felt like a tour of Akihabara.

  "We took a building that supposedly had a capacity of 220 and packed it with 298 people," George Salah confided when no fire marshal could hear him.

  Engineers Howard Gobioff and Ben Smith complained to eStaff (the executive staff) that lack of room was poor for morale and adversely affected productivity. In the midst of a hiring frenzy, they were conducting phone interviews in the stairwells and out on the grass because all the conference rooms and private offices now housed multiple full-time occupants. Cubes for two people held four workstations, and whenever employees were out of the building for more than a day, newly hired engineers would appropriate their spaces. Worst of all, work teams were split up and located in different parts of the building, hindering communication.

  Larry and Sergey gave George the okay to acquire more space. Sergey helpfully advised him that SGI's office six blocks away looked "nice." It was a joke. SGI's world headquarters complex had four massive buildings, each of which dwarfed ours.

  "In April 2000, an office building across from us came available," George recalls. "They wanted $8.00 per square foot and we were paying just over $2.50. I said to Larry and Sergey, 'This is a ridiculous price. I don't think we should even bother with it.' They said, 'No, let's put in an offer.'"

  George saw trouble brewing. The Silicon Valley vacancy rate was parked at one percent, and landlords saw no need to negotiate. "We're going to have to deal with this guy again," he warned them. "He owns all of Mountain View. We don't want to get on his bad side. Let's not place an offer."

  That only made Larry and Sergey more eager. "No, no. Put in an offer," they insisted. "A lowball offer. Let's offer $6.45 per square foot."

  George called a broker. The broker called back. "The landlord screamed at the broker," George remembers. "'How dare you send me this offer? You know better than this. Don't bring me this kind of stupid offer. And tell them it's now $8.25!' And he slams down the phone. So we didn't get the building."

  Two weeks later the dot-com fever broke and the real estate frenzy passed like a kidney stone. Suddenly everything was fluid. George sublet space in that same building for $3.50 per square foot, and a year later he leased a completely furnished building nearby for forty-five cents a square foot.

  It became a pattern. Just at the brink of disastrous overcrowding—or perhaps a few weeks after—George acquired more space at fire-sale prices and the Pacific Islanders reappeared with hammers and rails. Google swallowed every adjacent building on the Pac-Man path leading straight to the SGI complex. Three years later, in 2003, we gobbled that too.

  Whenever Google expanded, new cubicle configurations appeared. There were half-height cubes, three-quarter-height cubes, cubes with cut-out pass-through windows, and cube suites with courtyards for couches or dogs. Larry and Sergey wanted to optimize the tradeoff between productivity and rent. What would happen if you put three people in a space designed for two? How about five? How many interns could you line up along rickety tables in a busy hallway? What if you made a fishbowl with glass walls all around and put four engineers inside it? Would they get more done? Less? Would ideas breed in captivity? They conducted the servers-and-corkboard experiment with bodies and fabric walls—rearranging the elements to yield the greatest computational output while generating the least amount of heat. Things could always be more efficient and cost less, in either time or money.

  And that's where Gerald Aigner came in.

  Gerald was the flaming sword of frugality wielded by the lords of Google—a cost-cutting obsidian blade with an Austrian accent and close-cropped blond hair. Blue-white sparks glinted from his angular glasses and lit his Jack Nicholson smile, and the air crackled when it squeezed into the furrows of his brow. Before meeting with Gerald, I checked my shoes for rubber soles and whispered thanks that I wasn't across a negotiating table from him. There are still parts manufacturers who shudder and blink uncontrollably at the invocation of his name. After doing a deal with Gerald, one vendor refused to talk to Google again unless we agreed to their price up front, without any discussion. "You're too hard to deal with and we won't make any money," they complained.

  "Gerald is probably the most difficult person on the planet to please," George confirmed. "He wants everything for nothing, has to have it faster than it can be delivered, and it has to be to his exact specifications, which can be unrealistic and just bizarre."

  Jeff Dean heard Gerald put competing vendors on the phone together and make them beat each other's bids, fight-club style, until only one was left standing—offering a price Gerald was willing to pay.

  When Google needed data-center capacity in Ireland, Gerald went to all three major facilities in Dublin and pitted them against one another to win the business. Who was most willing to undercut the competition to become Google's premier partner? Who was willing to sacrifice the most to win a foothold for future business? Only after he had browbeaten them all into deep concessions did he sign a contract—w
ith each of them. He needed all the capacity he could buy, but there was no advantage in revealing that in advance.

  "The man had a penchant for cheap," recalls hardware designer Will Whitted, who worked closely with Gerald. "I loved him dearly for that. Anytime you told him the cost of anything, it was too much. The guy made me absolutely nuts, but he was almost always right. Brilliant guy."

  "This is ridiculous," Gerald said about the cost of the boxes encasing our servers. "We shouldn't pay three hundred dollars for a piece of sheet metal." He convinced Urs to get rid of the boxes and go with an open design based on the bread racks he had seen in Charlie's kitchen. The lack of a case made everything far easier to service.

  "There wasn't any part that I couldn't change in under thirty seconds," Whitted remarked about his implementation of Gerald's idea. "That allowed us to have far fewer data-center technicians. The average in the U.S. is one tech per thirty machines. Google was running three thousand machines per tech because there was no box."

  Gerald's expertise extended from the lowest level of the hardware to the software that ran on it, enabling him to identify savings others couldn't see. Despite that according to Whitted, "Gerald was hugely afraid of electricity. He was sure if he touched a thing with five volts on it he was going to be electrocuted."

  "That never happened," Gerald insisted. He disputes most of the stories his former colleagues tell about him. He denies that a vendor dumped boxes of computer parts in our parking lot and just drove off when he realized he had sold them to us below his own cost. Bogdan Cocosel, who worked with Gerald on hardware, assured me that it did happen and that when they were tested, only half of the machines worked. Fortunately, Gerald had such a deep understanding of hardware that he knew how to salvage even problem parts.

  There were other bargains to be had. When Google needed data-center space in Atlanta, Gerald approached the recently bankrupted company Cable and Wireless. He bought eighty million dollars' worth of equipment from them for a million and a half. He admitted to me that a ninety-five percent discount was a reasonably good deal.

  Gerald dealt with numerous bankrupt data centers, though not all of them had reached that point before taking on Google's business. Exodus and GlobalCenter both hosted Google servers before sinking together under a sea of red ink. MCI WorldCom went out of business before they could sign their pending contract with Google. Gerald denies that he single-handedly drove data centers out of business. "The dot-com bubble burst. They had built all these data centers and couldn't fill them. They didn't realize they were in the real estate business." Besides, he explained, "I don't think I was a great negotiator. I only got eighty to ninety percent off list."

  Urs gives him more credit than that: "Gerald understood the cost of ownership. Often, people didn't know what they were signing on the other side and were guaranteed to lose money. They should have known that power costs money and you can't just give it away. Or you do it once, but you don't do it twice—and they let us do it multiple times."

  Gerald made a point of knowing everything about the companies with which he was negotiating, from their cost structures and breakeven points to the tariffs on their power bills. He knew what data centers had to pay for electricity even if they didn't, and he used that knowledge to Google's benefit. Other Googlers took advantage of Gerald's abilities to smooth over their own mistakes.

  "I'm in front, cabling some of the machines to the switches," Jim Reese told me about a visit to Google's data center, "and all of a sudden I hear 'Zzzzt! Ahhhh!' I walk back and Harry [another early engineer] is just standing there, kind of dazed. We had fried the entire rack—all the RAM. RAM was difficult to get and very expensive, and we didn't have credit—I would drive down to Exodus with a half-million-dollar check in my hand so they'd offload the servers from the truck. We had Gerald convince the vendor the RAM was defective—because how could all three hundred and twenty sticks of RAM not work—right? So they replaced it all and sent it back to their manufacturer."

  When I asked, Gerald disputed this account as well. There's something about him, though, that spawns such tales, perhaps because, as Urs described him, Gerald was "a very intense guy": "Sometimes people felt that he was angry, and he was just so energized about the problem that it appeared he was physically struggling with things." Because Urs felt Gerald was "an incredibly good match to what we needed," he protected him.

  "Urs shielded me from Larry and Sergey," Gerald claimed. "I had a pretty good life because I could do what I wanted to do." Eventually he decided he wanted to live that good life in Europe, giving rise to one last—possibly true—story to add to the canon of Gerald legendry.

  "When Gerald wanted to work in the Swiss office," Jeff Dean heard, "he somehow managed to get a Canton outside Zurich to change their tax laws for options. It's unclear if they did it for him, but it certainly would have applied to him."

  To me, Gerald represented the unattainable ideal of bargaining savvy. When I brought in contracts, still wet with red ink bleeding from the slaughtered profit margins of vanquished vendors, Sergey sniffed, "These terms don't look very favorable. Maybe we should have Gerald talk to them." In his suggestion I heard, "Summon the Prince of Deepest Discounts! Let him pick the carcass clean and suck the bones of their marrow!" Perhaps my fears were exaggerated, but I knew Gerald, if set loose on my deal, would dash any hope of future favors from my supplier. The specter spurred me to put the screws in, which often was enough to sufficiently improve the terms. The sales reps I dealt with will never know the kindness I did them by squeezing them as hard as I could.

  Think big. Stay flexible. Embrace data. Be efficient and economical in the extreme. I was mapping Google's unique terrain and starting to feel less like a toddler lost in the wilderness and more like an intrepid explorer boldly exploring terra incognita. I kept my eyes open and took one step after another in a direction I hoped was forward.

  Examining My Options

  Serious men wearing jackets (but no ties) were gathering in the I'm Feeling Lucky Lounge—our largest conference room. I watched the shadow army of laptop-toting factotums come and go, speaking in dry cellular whispers. Inside, I imagined our directors laying out visions, making revisions, moving, acting, arguing, and trying to keep their egos in check. Hours later they would emerge as the factotums checked their Palm Pilots and scuttled off to fetch their cars. At that point, I would start scanning my email very carefully.

  I never dealt with our board face-to-face, but I always read the notes distributed after their meetings. Why? Well, of course I wanted to know about major shifts in our strategy, and announcements about significant executive hires, and... Ah, hell. Mostly because the board set the price of our stock.

  Maybe this would be a good time to talk about money.

  Almost no one talked about money at Google. It was an easy topic to avoid because nobody at the company had money to speak of. There may have been one or two people who left other startups with spending cash, but if they were rich enough to retire, they hid it well. What we did have was wealth potential.

  I mentioned that in coming to Google I took a twenty-five-thousand-dollar pay cut. That did not help my cash flow, but at dot-coms stock options carried more weight than salary. Each option locked in the purchase price for a single share of stock, set at the stock's value the day the option was granted (the "strike price"). Even if the shares' price had skyrocketed by the time I actually exercised my option and bought the stock, I would still pay only the strike price, guaranteeing me a hefty profit. Companies usually offered their first hires significant numbers of shares but got stingier as the company grew. The earlier one joined, the greater the risk. The more risk, the greater the potential reward. Buy low, sell extremely high, retire young.

  Stock options could compress the distance between a life of daily toil and one of affluent ease to the width of a sidewalk crack. I had seen friends and neighbors step across, buying mansions and taking charter-jet vacations, while my kids shared a room
and were lucky to get day passes to Raging Waters. Why them? Why not me?

  The truth is, most people in Silicon Valley worked hard, made smart decisions, and yet never passed that threshold. Randomness played too big a role. Or call it luck. I wasn't an entrepreneur or a brilliant scientist. I needed to jump on someone else's bandwagon and pray it didn't wind up in a ditch.

  "We're prepared to offer you a salary only slightly lower than what you're making now," the hiring manager across the table from me at Yahoo had said, "and"—he paused dramatically—"one thousand stock options." I wasn't blown away. Yahoo's stock was already at fifty dollars a share. If it went to a hundred, I would make fifty thousand dollars. But if the stock dropped and never reached fifty again, my options would be "underwater" and hold no value.

  For tax reasons, it's advisable to buy your options at the time they're issued, which would have meant putting fifty thousand dollars of my own money on the table: fifty thousand dollars I would never see again if the company took a dive. Plenty of dot-coms did go under in the big bust the following year, leaving little in their wake but suburban bathrooms wallpapered with worthless stock certificates. And even though I would be putting out the money up front, I wouldn't actually take ownership of the options immediately. I would take possession or "vest" on a schedule that began with a "cliff" of one year, meaning that not one share would actually belong to me until I had completed twelve months at the company. Then I would earn the rest of my shares month by month over the next three years, at which point I would be fully vested. Until the company went public, however, I couldn't sell the stock to anyone but the issuer.*

  When I read my offer letter from Google, I thought there must be a typo. I was prepared for the pay cut, but the comma was in the wrong place on the number of options granted. It was an order of magnitude larger than I expected. Google was offering me a large stake, but I would have to bet it all on the company moving into the black. If Google did well, so would I. Salary? That was just a free drink and a complimentary crab cake while I waited for the roulette wheel to stop spinning and my chips to come in.