Amazon Unbound: Jeff Bezos and the Invention of a Global Empire

by Brad Stone

The first company-wide motto was “Get Big Fast.”

The Amazon that I had written about was worth nearly $120 billion at the end of 2012. The company’s market capitalization touched a trillion dollars for the first time in the fall of 2018—eight times more valuable in less than six years—and returned to surpass that threshold, apparently for good, in early 2020. My Amazon had under 150,000 employees. By the end of 2020, it had an astounding 1.3 million employees. I was writing about the Kindle company, but this was now the Alexa company.

Bezos had peculiar ideas about how customers might interact with these devices. The engineers working on the third version of the Kindle discovered this when they tried to kill a microphone that was planned for the device, since no features were slated to actually use it. But the CEO insisted that the microphone remain. “The answer I got is that Jeff thinks in the future we’ll talk to our devices,” said Sam Bowen, then a Kindle hardware director. “It felt a bit more like Star Trek than reality.”

Inspired by the conversations with Hart and others about voice computing, he emailed Hart, device vice president Ian Freed, and senior vice president Steve Kessel on January 4, 2011, linking the two topics: “We should build a $20 device with its brains in the cloud that’s completely controlled by your voice.” It was another idea from the boss who seemed to have a limitless wellspring of them.

Bezos would remain intimately involved in the project, meeting with the team as frequently as every other day, making detailed product decisions, and authorizing the investment of hundreds of millions of dollars in the project before the first Echo was ever released.

first company Amazon bought, Yap, a twenty-person startup based in Charlotte, North Carolina, automatically translated human speech such as voicemails into text, without relying on a secret workforce of human transcribers in low-wage countries. Though

this is a real thing for audio & image processing

After the meeting, Adams delicately told Hart and Lindsay that their goals were unrealistic. Most experts believed that true “far-field speech recognition”—comprehending speech from up to thirty-two feet away, often amid crosstalk and background noise—was beyond the realm of established computer science, since sound bounces off surfaces like walls and ceilings, producing echoes that confuse computers. The Amazon executives responded by channeling Bezos’s resolve. “They basically told me, ‘We don’t care. Hire more people. Take as long as it takes. Solve the problem,’ ” recalled Adams. “They were unflappable.”

Characteristically secretive, Amazon has never revealed the name of the voice artist behind Alexa. I learned her identity after canvasing the professional voice-over community: Boulder-based singer and voice actress Nina Rolle. Her professional website contained links to old radio ads for products such as Mott’s Apple Juice and the Volkswagen Passat—and the warm timbre of Alexa’s voice was unmistakable. Rolle said she wasn’t allowed to talk to me when I reached her on the phone in February of 2021. And when I asked Amazon to speak with her, they declined.

Evi employed a programming technique called knowledge graphs, or large databases of ontologies, which connect concepts and categories in related domains. If, for example, a user asked Evi, “What is the population of Cleveland?” the software interpreted the question and knew to turn to an accompanying source of demographic data. Wired described the technique as a “giant treelike structure” of logical connections to useful facts.

Bezos met the Tyto team every few days for three years, at the same time he was meeting the Alexa team as frequently. He was infatuated with new technologies and business lines and loved spitballing ideas and reviewing the team’s progress. And while he was inordinately focused on customer feedback in other parts of Amazon’s business, Bezos did not believe that listening to them could result in dramatic product inventions, evangelizing instead for creative “wandering,” which he believed was the path to dramatic breakthroughs.

Bezos also had some worrisome blind spots about smartphones. “Does anyone actually use the calendar on their phone?” he asked in one meeting. “We do use the calendar, yes,” someone who did not have several personal assistants replied.

For the next few quarters, Amazon avoided buying Google ads in Mexico and tried to compensate with billboards, radio, and TV ads, and shipping discounts. As Garcia had feared, it hobbled the site. The offline ads were more expensive and less effective. Google brought in $70 billion in annual advertising revenues because search ads worked and were a relatively inexpensive way for websites to attract visitors. “I wanted to see if we could get traction in a country launch without using Google,” Wilke later said, “and it turned out, the answer was no…. We weren’t reaching enough customers.”

course, Ballmer had little grasp of how Amazon’s eventual engine of profitability, Amazon Web Services, was performing—and that was how Jeff Bezos wanted it. Over its first decade, AWS’s revenues and profits were a closely guarded secret. The division generated $4.6 billion in sales in 2014 and was growing at a 50 percent annual clip. But Amazon disguised those numbers, along with nascent advertising revenues, in a sundry “other” category on its income statement, so that potential competitors like Microsoft and Google would not recognize how attractive a business cloud computing actually was.

The name “Redshift” was suggested by Charlie Bell, a former engineer for Boeing on NASA’s space shuttle and the senior vice president in charge of AWS’s operations; it’s the term for the change astronomers see in light, the fastest thing in existence, that’s emitted from a celestial object like a star as it moves away from the observer. Nonetheless, Larry Ellison, then the CEO of Oracle—whose logo happens to be red—saw it as corporate trash talk and began to see red himself. That “never even occurred to us,” Jassy said. “When we were told later that Oracle believed that, we thought it was kind of funny.” A bitter rivalry between Oracle and Amazon, already simmering with Amazon’s entry into databases, intensified.

flagging. The highlights of the AWS week were two Wednesday morning meetings. Jassy ran the ninety-minute midday business review, where the top two hundred managers discussed the minute details of customers, competitive developments, and the financial health of each product unit. But the real centerpiece of the week was the forum that preceded that meeting: the two-hour operations review to assess the technical performance of each web service. Held in the large conference room on the third floor, it was run by the intimidating and direct Charlie Bell, the former space shuttle engineer.

Getting selected could be a career-defining moment at AWS. Managers could boost their prospects with a comprehensive and confident presentation. But if they employed ambiguous language, erred with their data, or conveyed even the whiff of bullshit, then Charlie Bell swooped in, sometimes with awesomely patronizing flair. For managers, a failure to deeply understand and communicate the operational posture of their service could amount to career death.

members of the Pancake Group had fanciful notions that Bezos was going to rescue the Post by spending uncontrollably, he quickly dispelled them. In early 2015, they trekked back to Seattle and presented him with a multiyear operating plan that called for the paper to lose more than $100 million over the next four years. Bezos shot it down immediately. “Yeah, I’m not interested in that” is how one participant recalled his understated reaction. After the meeting, Bezos and Fred Ryan sat down and hashed out a plan to run the paper as a disciplined, stand-alone business, not as the hobby of someone with limitless resources. Over the next few years, there would be a quiet, targeted series of layoffs in the company’s print advertising division, which were partially offset by a smaller but louder number of hires of digital media specialists.

When Bezos acquired the paper, Prakash had been developing a content system for the Post, called Arc Publishing, to manage functions like online publishing, blogging, podcasting, and advertising. Naturally, Bezos loved the idea of supplying that technology to other papers and encouraged Prakash to license it to broadcasters and any company that needed publishing software. By 2021, Arc powered fourteen hundred websites and was on a path to generating $100 million in annual revenue.

The company spent an estimated $3.2 billion on Prime Video in 2016 and nearly $4.5 billion in 2017. Even its usually agreeable board of directors was apprehensive over the growing expenditure and asked pointed questions about it. “Jeff was ahead of us in thinking about the relationship between content and Prime” is how former board member and venture capitalist Bing Gordon put it.

They tracked each show and analyzed how many people watched and then converted from free Prime trial periods or extended their current membership. But there was little evidence of a connection between viewing and purchasing behavior—especially one that justified the enormous outlay on video. Any correlation was also obfuscated by the fact that Prime was growing rapidly on its own. The truth was this: Bezos wanted Amazon to make TV shows and films. He could see that the decades-old way that TV shows and movies were produced and distributed was changing and sought a principal role for Amazon in that future.

Many female employees at Amazon Studios in 2017 continued to be unhappy with their boss or their work environment. One described a conference room at the Amazon Studios office with walls that were covered in portraits of Jeffrey Tambor, Woody Allen, and Kevin Spacey (star of an Amazon film, Elvis & Nixon). All three would fall in the gathering backlash against sexual misconduct, known as #MeToo. The movement was also about to ensnare Roy Price and entangle Amazon in a scandal that its executives thought they had put behind them.

The movie performed reasonably well at the box office, but unfortunately for Amazon, the toys didn’t sell for whatever reason, and ended up sitting on the shelves of Amazon’s FCs, gathering dust. “It was just a thud of a license,” recalls Jason Wilkie, a former in-stock manager in the toys group. “You couldn’t even discount it. Nobody wanted it.” Analyzing the mistake, Amazon’s retail execs concluded that capricious human emotion had interfered with the cool-headed evaluation of the available data, which might have led to a more conservative purchase order. The project that emerged was called “Hands Off the Wheel.” Over the next few years, in-stock managers across the retail group were moved to different jobs or booted from the company and replaced by automated systems. Software, not people, would crunch the numbers and place purchase orders. Algorithms might not be able to perfectly gauge demand for cinematic toy licenses, but they could predict the surge of interest in, for example, canine anxiety jackets to soothe dogs before the July 4th fireworks display, or snow shovels before a predicted winter storm in the Midwest, and so on.

Other “Jeffisms,” recorded by the FBA team in their annual OP1 planning session with Bezos, further shaped their perspective. Among his greatest hits, recorded in a memo that was later passed to me: “Focus on lowering cost structure. It is better to have low costs and then charge to maximize your value versus charging to cover costs.” “Having a stupid rate card equals stupid things happen. Rate cards must be equal to the value.” “We do not charge more because we can’t figure out how to make it cost less. We invent to make it cost less.” “We should be able to fulfill 100% of the 3P business. I do not know what the debate is, yes, we must fulfill low priced selection, it is crucial.” “Averages are bad measures. I want to see actuals, highs, lows and why—not an average. An average is just lazy.”

Chinese startups, some quite formidable, arose almost out of thin air to sell on Amazon.com. In 2011, a software engineer named Steven Yang had quit his plum Silicon Valley job at Google and moved to Shenzhen to start an electronics company called Anker to sell accessories like replacement laptop batteries. Over the next few years, his product line expanded to include just about every kind of cable, charger, and battery imaginable, many of which soared to the coveted top spot of Amazon’s bestseller lists.

The executives on the trip were amazed by what they saw. The group on the fashion leg visited an apparel factory that was making $9 sport coats for the retailer Abercrombie & Fitch, which then sold them at retail for $500. The same factory was also selling the coats with a different button pattern directly online for $90—and still making a fat profit.

What execs still didn’t like to admit, particularly at a politically sensitive time in bilateral trade relations, was that 49 percent of the top ten thousand largest sellers on Amazon were based in China, according to Marketplace Pulse, a research firm that monitors the site. In April 2020, the

Herrington’s memo pointed out that Walmart, Carrefour, Tesco, Metro AG, and Kroger were the world’s five largest retailers at the time. “All of them anchor their customer relationship in groceries,” he wrote. If Amazon’s retail business was going to grow to $400 billion in gross merchandise sales, it needed to transform a model based on infrequent shopping for relatively high-priced goods to more regular shopping for low-priced essentials. In other words, if the company was going to join the ranks of the biggest retailers, the S-team had to figure out a way to profitably sell supermarket items. If they didn’t, Amazon was going to be vulnerable to rivals who already enjoyed the shopping frequency and cost advantages of the grocery model.

Logistics employees who worked on the California service said this hub-and-spoke model ended up being inefficient and unreliable; one said that Amazon was “basically stapling a $10 or $20 bill to every order.” The Fresh team also tracked a metric called “perfect deliveries”—when an order was promptly delivered and included every item. They found they were hitting that target less than 70 percent of the time. Grocery industry veterans belittled the effort from afar. “Amazon Fresh is their Waterloo,” John Mackey told me during our chat in 2014. “What’s the one thing people want? Convenience. You can’t do that with distribution centers and trucks.”

its own vendors and sellers? One of Amazon’s databases, Heartbeat, had all customer reviews across the website; accessing it allowed employees to look for revealing patterns that might indicate how they could improve well-established products. For example, customer reviews in the all-important category of dog poop bags indicated regular confusion about which end of the bag opened. So the Amazon Basics version included a blue arrow and the words “open this end.” Another valuable tool was Amazon’s Vine program, in which influential product reviewers received free samples in return for their written appraisals—and were thus more likely to produce more exuberant critiques. Speaking on the condition of anonymity, several

Then the night before, after the press had been alerted to the unveiling, internal testing revealed a software bug that might erroneously inform customers who purchased the product that it was sold out. The launch was delayed, and beleaguered project members had to come in on a Saturday to analyze the issue. Amazon also sent in the “principal engineers”: an elite squad of about a dozen technical wizards at the company who parachute into troubled projects to diagnose problems.

Together, the trio navigated daunting challenges. In 2000, Amazon bolstered its fragile balance sheet by agreeing to handle online sales for the retail chain Toys “R” Us, with all merchandise sent to Campbellsville. The following year, it struck the same arrangement with Target. The excess inventory overwhelmed the 770,000-square-foot building. To keep up, Valdez, Clark, and Roth leased some six hundred tractor trailers, stocked them with overflow, and parked them around the tiny town (population: nine thousand). “It was pure survival,” Valdez said.

I also had a fulfillment center do this 25+ years ago, but luckily they only needed a couple of trailers.

Bezos didn’t want another empathetic business philosopher to replace Onetto as the head of Amazon’s operations; he sought an uncompromising operator who could gain operating leverage by slowing down the growth of costs in the FCs relative to Amazon’s skyrocketing sales. Fulfillment expenses had jumped by 58 percent in 2011 and 40 percent in 2012. Amazon hired fifty thousand temporary workers in its domestic FCs over the 2012 holidays alone; those numbers would keep going up and up and up to meet anticipated increases in sales. The new operations leader would have to take a hard-nosed run at the entire supply chain and figure out how to use technology to get more efficient.

The day after the 2013 holiday meltdown, Clark called Michael Indresano, a former FedEx executive who had joined Amazon as a transportation director, and asked how many “sortation centers” they could build before the next holiday peak. These were the facilities that consolidated packages by zip code and injected them into the U.S. Postal Service for last-mile delivery to people’s homes. Indresano estimated that he could open sixteen by the end of 2014. “Build ’em all!” Clark replied. Inside Amazon, the rapid creation of sort centers in cities like Atlanta, Miami, and Nashville was dubbed “the Sweet 16.”

That was the unofficial birth of the service that would come to be known first as Prime Air, then Amazon Prime Air, and finally, Amazon Air. The confusion was born from Bezos using the former name in his infamous 60 Minutes interview in 2013 to announce that the company was working on aerial drones to ferry individual packages to people’s backyards. Several operations executives told me they were embarrassed by that stunt (which seven years later, had progressed no further than private tests). They used an old internal Microsoft term to describe it: “cookie licking,” or the act of claiming to do something before you actually do it, in order to capture notoriety and prevent others from following.

team members appeared to cast blame on Stephenson, who would leave the company a year later to become the CFO of Airbnb. “The tone of it was ‘how did you miss it?’ ” said another executive privy to the discussions. “But we had been missing it together for this whole time.”

Advertisers accustomed to getting rich sets of demographic data about customers from Silicon Valley firms were also rebuffed. No, advertisers couldn’t access Amazon data about its customers’ age, ethnicity, and shopping habits. No, Amazon wouldn’t allow companies like Adobe and Acxiom to put their third-party software tags onto ads and track their performance, a common practice elsewhere on the web. Advertisers would have to settle for getting reports about their ads’ effectiveness directly from Amazon.

Inside the advertising group, a few of these battles became notorious. Over one holiday, Paul Kotas vetoed the particular hue of blue in ads by the Ford Motor Company, because the display campaign felt “Sunday circular.” Amazon also told the wireless carrier T-Mobile that its trademarked magenta-pink logo was too bright and distracting. And it informed Sony Pictures that a banner for the James Bond film Skyfall violated the policy on showing weapons. The studio “was like, ‘screw you!’ ” an Amazon ad executive recalls. “Who is James Bond in silhouette without a gun? Literally, he’s just a random dude.”

When sponsored ads were prominently displayed, there was a small, statistically detectable short-term decline in the number of customers who ended up making a purchase. The longer-term effects were unknown. The scorpion wasn’t killing the frog, but lightly nicking it—and it wasn’t clear yet whether the bite was poisonous.

“Moving ads to the top of search pages was the game changer,” said an Amazon computer scientist who worked on the ad business. “Sponsored products would be nothing close to what it is today if that decision wasn’t made, and Jeff was the one who signed off on it.” Once Bezos showed a willingness to convert the relative meritocracy of search results into a domain that prioritized Amazon’s commercial interests, the possibilities were endless.

unpredictable. In the days before the finalists were announced, the HQ2 team divided the list of more than two hundred cities that hadn’t made the cut and placed phone calls to local officials, alerting them to the bad news. Most asked why, while expressing disappointment with the amount of time and resources they had put into the failed effort. Amazon employees responded with blasts of data. “Your metro area only has 375,000 people and of those only 10 percent have advanced degrees,” went a typical explanation. “Sorry, that’s not enough of a labor pipeline.” City officials mostly agreed that Amazon’s outreach was conscientious, and that Holly Sullivan in particular spent more time than she needed answering questions and preserving relationships that might be helpful to the company in the future.

Now Amazon was about to validate that pessimism: one of the richest companies in the world, led by the world’s richest person, was expanding its presence in the political and financial capitals—cities where Jeff Bezos owned lavish homes. Adding to the awkwardness, on the morning of Tuesday, September 4, Amazon’s shares were worth as much as $2,050—briefly tipping the company’s market capitalization over the momentous threshold of a trillion dollars, before the stock price retreated.

Lina Khan, who had authored an article for the Yale Law Journal accusing Amazon of anticompetitive behavior and the nation’s antitrust laws of being woefully out of date.

and 4 years later became head of the FTC

Holly Sullivan called Dallas mayor Mike Rawlings to deliver the bad news. The financial incentives from Dallas and the state of Texas had totaled $1.1 billion, considerably higher than the $573 million in cash grants offered by Arlington County and Virginia, though not quite as much as the city and state of New York’s enticement of $2.5 billion in tax credits and rebates. It was also about 40 percent cheaper to build in Dallas than on the East Coast—but in the end, that hadn’t mattered at all. “Help me here, why did you put us through all of that if this is where you were going all along!” an exasperated Rawlings asked Sullivan.

Several employees argued that the helipads were a terrible idea but were told the request came right from the top and wasn’t going to be rescinded. “The helipad was the worst thing they could have ever asked for,” mourned Mitchell Taylor, bishop of Center of Hope International church in Long Island City and a supporter of HQ2. “Why do you have to put that front and center? They could have had a helipad after the fact.”

Lauren Sanchez flew helicopters

Amazon employees were now just as perplexed as their counterparts at Blue Origin were when a company called Black Ops Aviation and its cofounder, former television personality Lauren Sanchez, started appearing at New Shepard launches in West Texas to record promotional videos for the secretive space firm.

That September, AMI had signed a non-prosecution agreement with the Department of Justice over allegations that it tried to buy and bury negative stories about Donald Trump. The deal required its executives to cooperate with the federal investigation of Trump lawyer Michael Cohen and to operate in the future with unimpeachable honesty.

Now Pecker was hemmed in; if they didn’t run the story, or if it broke elsewhere, they would have wasted the large sum and exposed the company to another possible “catch and kill” allegation. After Pecker exploded in anger at Stracher over lunch at Cipriani Wall Street in lower Manhattan, the veteran lawyer walked out of the restaurant, essentially quitting on the spot. That elevated his recently hired deputy: Jon Fine, who—another unlikely coincidence—had previously worked at Amazon for nine years.

Sanchez also tipped the paper off to the couple’s travel plans; and when he had dinner with them at the Felix Trattoria restaurant in Venice, California, on November 30, two reporters were stationed at tables nearby and the tabloid’s photographers were clicking away surreptitiously.

Michael Sanchez, Lauren's brother.

Michael Sanchez believed he was cleverly manipulating the Enquirer.

After signing a $25,000-a-month contract with his sister to help her navigate the descending insanity, he called Dylan Howard to announce that he was acting as his sister’s representative and suggested that he come to New York to review the paper’s reporting (which, of course, he had provided). Confident in the promise of confidentiality from AMI, Michael Sanchez was now playing both sides. Bezos, meanwhile, involved his longtime security consultant,