How Amazon Is Beating Antitrust Before It Happens

Jan Antonin Kolar/Alexa

Information technology is the largest sector of the stock market, worth more today than the entire market in the trough of the recession, and contains the market’s five highest valued companies.

Amazon, jockeying in the top five, is in a class of its own.

In 2018, the company reached a valuation of $1 trillion, the fastest ever to do so, earned $241.5 billion in revenue—exceeding Facebook and Google combined—and employed 613,000 people. It’s now the second largest employer in the United States. The retail startup has become, as Lina Khan writes in Amazon’s Antitrust Paradox, a “marketing platform, a delivery and logistics network, a payment service, a credit lender, an auction house, a major book publisher, a producer of television and films, a fashion designer, a hardware manufacturer, and a leading host of cloud server space.” The firm stands alongside history’s titanic conglomerates.

The pillars of Amazon today are Marketplace, Amazon Web Services (AWS), and Prime membership. Marketplace includes Amazon.com, the familiar retail platform, and Fulfillment by Amazon, providing warehousing and shipping to retailers. AWS is a suite of computing infrastructure services, including cloud computing, and storage. Prime is its annual flat fee membership for free 2-day shipping and other perks.

Amazon can’t be comprehensively described as a retailer, or even a technology firm. It’s more like a utility for commerce comprised of infrastructure service platforms. Like other large firms, Amazon overcomes high fixed costs and low profit margins with massive volume and economies of scale.

The main innovation at Amazon was not technological but organizational. The company is uniquely adapted to the economic logic of the information age, and its structure and conduct lend themselves well to defense against internal and external threats to growth, especially antitrust enforcement. As Khan notes, Amazon is structured in such a way that it’s like “Bezos charted the company’s growth by first drawing a map of antitrust laws, and then devising routes to smoothly bypass them.” The CIA’s computing-intensive operations, for example, are heavily tied into AWS. This is likely not an accident. Aside from providing crucial services to dark parts of the government, Amazon cuts off antitrust angles by obsessing about customer experience. Antitrust requires proof of customer harm, after all. No harm, no antitrust—at least in theory.

Rather than take profits and issue dividends, which Bezos specifies as Amazon’s “fundamental measure” of success, the company aggressively re-invests in new growth. Amazon’s value has climbed as investors have rewarded this strategy, enticed by the promise of even greater dividends when the company’s greater ambitions are realized. Management has been characteristically secretive about what that might be.

For a corporation pursuing growth and survival, the ideal position to achieve would be that of the megacorporation. The megacorp is a company that uses economically or technologically derived power to establish a fortified position in society, outside the state’s reach, from which to exploit the population. We’ve seen the trope of the social control state manifest in China; perhaps recent technological perturbations could give rise to a genuine, malicious megacorp.

The notion that social and economic structures under state socialism and large-scale corporate capitalism begin to resemble each other is not a new one. Back in the 1940s, the economist Joseph Schumpeter predicted that the success of corporate power itself would lead to the demise of entrepreneurial markets. Likewise, the unorthodox Hegelian philosopher Alexandre Kojève predicted that a Cold War victory by either American or Soviet power would lead to a fundamentally similar society, where bureaucratic management displaced the older class structure. The same idea can be seen in the familiar cyberpunk trope of a hegemonic and dystopian corporation which embodies state-like powers, unchecked by law or justice.

Amazon’s dominance has invited the comparison. But this might actually underestimate the company’s ability to foresee and adapt to the obstacles of public resentment, particularly where it might have political fallout. Under Bezos, the company’s structure and conduct lack much resemblance to the malicious corporate powers of science fiction. Such criticisms also have to contend with the fact that Amazon’s power has been built off its ability to build customer loyalty. True, workers have had to deal with grueling warehouse work hours and companies using Amazon platforms have been put under pressure. But all these actions are undergirded by a logic which has benefited consumers. And since many of those adversely affected by Amazon in one capacity benefit from it in another, this throws a major wrench into the state’s task of deciding how to deal with its growing power.

The Nature Of Amazon

What’s Amazon’s endgame? It’s impossible to know without first looking at what Amazon is as a company and why it was created that way. In his first letter to shareholders, Bezos describes his aim: “extend and solidify our current market leadership position…in terms of the metrics most indicative of our market leadership: customer and revenue growth, the degree to which our customers continue to purchase from us on a repeat basis, and the strength of our brand.” This sounds entirely banal, but what if a company actually followed the ‘core values’ that normally make half its employees roll their eyes? That company is Amazon.

Amazon is also obsessed with guarding against internal and external threats to growth, primarily organizational sclerosis and antitrust enforcement. These are the interests around which Amazon—and every facet of the company—was devised.

As Bezos writes, Amazon is “a distinctive organizational culture that cares deeply about and acts with conviction on a small number of principles.” These are “customer obsession.., eagerness to invent…willingness to fail…patience to think long-term, and…operational excellence.” These precepts are designed to forestall the forces that kill companies. This approach may seem trivial, but management’s operational excellence—their consistently demonstrated discipline to execute on it—is rare.

As Amazon scales, it can afford the fixed cost of more fulfillment centers near large populations, expediting shipping and therefore increasing sales. Amazon’s business model is built on the use of scale. This is the key to its defensibility.

Amazon invests heavily in customer experience to drive this flywheel. No one will ever ask for higher prices or slower delivery, Bezos often jokes, so its investments in the short-term will pay dividends indefinitely.

In early years, Amazon burned hundreds of millions of dollars selling popular holiday items and electronics at a loss to build a reputation for vast selection. Its automated systems proactively issue refunds if a price drops after pre-order, or playback quality on a movie rental is poor. Given the amount of data Amazon has on price elasticity, it could easily decide to set profit-maximizing prices. But Amazon offers lower prices to build trust, which really means it forgoes profit to snuff out competitors. If customers trust that Amazon will put their interests first, they’ll start shopping searches at Amazon.com, and they won’t bother to compare prices elsewhere.

The key to bringing this operation together is Prime. This service is Amazon’s bread and butter. Prime exists to lock-in Amazon’s growth. The sunk cost and better service makes members more likely to only shop at Amazon.com, and additional perks like video streaming and Whole Foods discounts strengthen Amazon’s brand relationship with members by increasing engagement throughout Amazon’s ecosystem. Prime members shop in more categories and spend considerably more. Crucially, “less than 1% of Prime members are likely to consider other mass-market retail sites.” One analyst estimates that half of U.S. households will have Prime by 2020.

Amazon’s public goal is to be a pro-social monopolist.

Amazon’s Organization

This blueprint is difficult to follow. Successful firms are plagued by internal maladies. A startup with energy and focus exploits an initial high-growth opportunity, but this culture is commonly lost as the organization expands. Deadweight employees, divisional fiefdoms, and bureaucratic procedures produce malaise and politics as the company grows complacent on the rents from its original success—much like with Microsoft and its country club reputation. Vertical integration insulates some operations from competition, so they grow sub-rate. The company becomes blind to orthogonal threats and opportunities. It fails to develop new businesses because it become too conservative to invest, too jealous of what it has to risk cannibalizing its businesses, and too large to be patient with initiatives that start small.

This pattern kills great companies.

Bezos has taken great pains to architect Amazon, so as to prevent this internal entropy. Workers are organized in teams small enough to be fed by two pizzas to maintain a lean culture. Each team has a metric for success, approved personally by Bezos, and competes with others for resources.

The most interesting organizational idiosyncrasy at Amazon is that around 2002, all teams at Amazon, without exception, at the cost of great disruption in the short-term, began to expose their data and functionality through APIs. All communication between teams occurs through those interfaces, making intra-company collaboration structured and legible. Most importantly, these interfaces are designed to one day be externalized for use by third parties as services; every two-pizza team is potentially incubating a revenue-generating product. When internal services are exposed to external competition, they need to stay the best, as indicated by financial results.

This structural shift preceded a strategic shift to selling infrastructure services to producers. They knew that, just as the aggregate demand for the “long tail” of non-best-sellers dwarfs demand for big titles, their major opportunity was in selling the millions of other products people might want to buy online. But their speed expanding selection was limited. Their solution was emblematic of their style—they allowed other sellers on Amazon.com. Opening their shelf space to competitors, for a fee, massively increased selection. Amazon became the best place for online shopping. Then in 2006, Amazon offered its fulfillment infrastructure to third party sellers, and launched Amazon Web Services, which provides computing infrastructure to developers.

In industry parlance, Amazon created “platforms.” Platforms are uniquely lucrative because they create enormous value and are highly defensible. Many of the greatest tech businesses in recent decades, like Windows and iOS, were platforms. They provide capabilities that third parties can use to make new things and serve customers. Platforms tend to become monopolies because of high fixed costs and increasing returns to scale. For example, bringing competitors onto Amazon.com created a new feedback loop to the flywheel, where more traffic brought more sellers, and that increased selection increased traffic.

By providing underlying infrastructure, Amazon is insulated from changing consumer tastes. Like the organelles that found the ultimate evolutionary niche by integrating into cells for mutual benefit, Amazon has achieved an endosymbiotic position in e-commerce and cloud computing. It is embedded in other business’ operations at such a basic level that it can’t be removed without great cost and disruption.

This structure did not come about through pure theoretical analysis; Bezos’ leadership is a foundationally important feature of the organization, which Brad Stone described in his company biography, The Everything Store, as “scaffolding built around his brain.” Many initiatives, especially the most successful like Prime and Marketplace, were strongly opposed internally but championed by Bezos, who lost widespread employee and investor confidence at times. It would be easy to understate the significance his savant intuition for the dynamics of internet business has had for Amazon.

Bezos also has a talent for condensing Amazon’s values and practices into stories and mnemonic phrases—“It’s always Day 1,” “it’s all about the long term”—and drawing attention to important ideas with surprising word choices. Each year has a theme at Amazon, with its own phrase. “Get Big Fast Baby” preceded “Get Our House In Order.” Bezos is sometimes theatrical, such as by making a show of leaving a chair open “for the customer” in meetings. The door desk, a DIY desk made cheaply from a door and 4x4s, is a potent symbol of startup scrappiness from Amazon’s early days widely invoked in the company and elsewhere. Stone recounts a story in which Bezos noticed that TVs had been installed in conference rooms—indicating to him wasteful spending and communication—and had them removed, but left up the wall mounts as a reminder of the episode. His most enduring legacy will surely be a symbol of long-term thinking installed on his ranch with the Long Now Foundation, a clock designed to keep time for 10,000 years.

Amazon’s Investments

Amazon’s flywheel widens the moat around its existing businesses by improving experiences, and its endosymbiotic position protects it from changes in demand, but Amazon has to proactively build new businesses to stave off irrelevancy—Day 2, Bezos calls it.

Bezos wants the company biased towards investment by having multiple paths towards greenlighting a project, having conviction that “big things start small,” and being willing to continue an undertaking as long as it has a single “high-judgment” champion. Amazon spent more on R&D than any firm in 2017 and 2018.

Its primary areas of investment are in retail and logistics, smart home and voice, and internet infrastructure. Other recent initiatives include expansion into India and the Middle East, the acquisition and production of media properties for Prime Video, a dramatic increase in advertising in recent years, and a healthcare experiment with Berkshire Hathaway and JP Morgan, along with other healthcare acquisitions like PillPack, a pharmaceutical delivery company.

Expediting shipping is a primary long-term goal because whether a given product makes sense to buy online often depends on how quickly it can arrive; this is why Amazon sells books, rather than hot cups of coffee. A more than doubling of Prime shipping coverage in recent years has brought products closer to consumers. In 2019, Prime moved from two-day to one-day shipping. What else? Amazon has developed drone delivery, a cashierless corner store, free two hour delivery on select items, and invested in expanding its own ground and air shipping fleet. At the warehouse level, Amazon has experimented with robot automation, worker tracking, and artificial intelligence for operations management.

Amazon has over 10,000 employees working on Alexa, its digital assistant. Through AWS, Amazon offers natural language processing, speech recognition, and speech synthesis to developers. The long-term goal is for Alexa to become like the computer in Star Trek. However, the short-term motivation is likely defensive. Dominant tech firms are often harmed when the primary computing platform changes because it disrupts their channels for reaching customers. For example, Google now has to pay Apple billions of dollars annually to be the default search provider on iOS. Amazon wants to own the dominant voice computing platform to protect its access to customers.

Alexa will manage the smart home—over 28,000 Alexa-compatible devices and Amazon’s door locks, home security cameras, and recently acquired Ring smart doorbell cameras. The smart home has retail synergies, such as aiding delivery and automatically ordering consumable household items.

With internet infrastructure, Amazon wants to remain integral to developers’ technology stacks. AWS has dramatically outpaced competitors in adding new services, especially machine learning. In 2019, Amazon acquired Eero, a home WiFi solution, and announced plans to provide global satellite internet.

While Amazon employees have been widely discussed in the context of their intensive work environment, they also play a key part in testing new products and services as consumers. Dogfooding—a company’s own use of its products—is a significant part of employee culture. Advanced access to new products is generally seen as a major perk of employment and is a way for employees to signal status. As a result, many end up investing portions of their own income back into the company itself. Amazon’s Seattle HQ was the trial site for services like Amazon Go, drone delivery, Prime Now, Amazon Fresh, Grocery Pickup, and Amazon Scout. Sites like the Amazon Books and Amazon Go stores have even become something like tourist attractions.

Amazon’s ability to roll out new services faster than both large companies and many startups is what keep this culture moving. This would be impossible without its twin features of an internal startup culture and a large-scale corporate operation. Consumers and employees are regularly confronted with a new launch that generates hype and increases the company’s market power. Prime Now is a classic example: released only 111 days from concept. The goal had been 90.

Amazon’s Long Term And Its Inevitable Politics

Amazon burns, foregoes, and invests money in the short-term to accelerate the growth flywheel, planning for an even greater payoff in the long-term that will have monumental social consequences.

Some suggest that Amazon aims to acquire a monopoly and then begin charging above competitive market prices. Once barriers to entry become insurmountable, they’d be safe from price competition. Arguably this is already the case—and yet Amazon’s prices remain fairly competitive.

Amazon’s size and dominance attracts distrust, but there are certainly long-term outcomes for the company to target that would pay off their investments solely by continuing to benefit consumers.

Amazon’s primary industries, e-commerce and cloud computing, still have a lot of room to grow. E-commerce makes up not quite 10% of spending in the U.S. And there’s still room for improving the experience. Nearly any good could be available instantaneously at just above manufacturing cost. Demand prediction and home monitoring could deliver goods before customers realize they’re needed. End-to-end autonomy would create massive efficiency in time, dollars, and energy. Alexa could provide ambient, frictionless computing. Internet access could be free and ubiquitous, subsidized by Amazon’s profit interest. AWS providing powerful capabilities to the public and making development ever simpler will increase access to the most potent creative tools ever.

In this best case scenario, Amazon would be a pro-social monopolist. The company could be wildly profitable by taking a small rake on the economic activity it enables.

But moving towards even this positive symbiotic endgame builds arbitrary power for Amazon. Economic power, past some threshold, becomes political and cultural power. Even at its present scale, Amazon must be modeled as a political actor. Its decisions influence society and sometimes involve political interests and calculations as much as economic ones. As its influence grows, Amazon must pursue political strategies—if only to protect itself from other political actors. It may also wish to obtain things that can’t be purchased except with the exercise of power.

Amazon’s enormous amounts of capital can be deployed politically. As one of the country’s largest employers, it’s used its fulfillment centers as leverage when negotiating with state governments over online sales taxes. Major cities adulated Amazon for a year, proffering extravagant perks, competing to be the site of the company’s second headquarters.

Amazon’s reservoir of data yields power. It has logs for a great deal of internet, logistics, and retail activity. Its smart home products like Ring doorbell cameras, with computer vision and facial recognition, could form a massive network of surveillance data of public and private spaces collected legally by individuals wishing to monitor their property. As an industry leader in AI, Amazon is well-equipped to extract insights from this data, in order to predict macroeconomic trends and build highly accurate profiles of individuals.

Amazon’s control of infrastructure is a source of power. Marketplace is the key distribution channel for thousands of large and small businesses. AWS is mission critical to the majority of the internet, including Netflix, Airbnb, Apple, and the CIA. It has the ability to quickly deny service to anyone on its platforms. In 2017, the infamous crash of Amazon’s S3 service for merely four hours—all due to a typo—caused widespread disruption and demonstrated how vital Amazon services are for large swathes of the internet, including major websites.

The 2018-2019 government shutdown shows how powerful this capability can be. Congress was in a deadlock until air traffic controllers used their control of a bottleneck on air transportation, by not going to work, to delay service at two major airports. The White House moved to end the shutdown within hours. Amazon’s grip on key infrastructure could similarly be used coercively, but it’s deterred by the possibly irreparable blowback. The air traffic controllers in the recent shutdown were protected because the shutdown was publicly unpopular, and they gave themselves cover by calling in sick. Their outcome was different from that of the unpopular 1981 ATC union strikers, whom President Reagan fired and blackballed. To exercise infrastructure-derived power, Amazon would need public support and plausible deniability.

Power does not lie dormant; it is used, or it is acted against. The power afforded to Amazon by its capital, information supply, and control of infrastructure won’t be ignored for long. A job of the state is to coordinate society’s power entities for mutual benefit, and rising powers must be integrated or disintegrated. Antitrust enforcement, which Amazon has adeptly outmaneuvered so far, is one of the state’s primary tools. If Amazon doesn’t want its wings clipped by antitrust agencies, it must decide whether to exercise its power aggressively and decisively to remain independent, or to broadly align itself with the government. If it opts for the latter, this presents the opportunity to hijack the state’s power through regulatory capture, while preserving its own. Rival powers within the state itself might therefore be inclined to pre-empt the possibility by working to pass more stringent and updated antitrust laws. Under the Trump administration, Google may soon be undergoing a similar probe from the Department of Justice.

Amazon And The State

There are historical analogues to the megacorporation. During the Age of Discovery, trading companies were granted monopolies by patron governments and took on many of the qualities of states. The British East India Company occupied India with an army manned by the local population and fought wars with governments and other trading companies. During Japan’s industrialization following the Meiji Restoration, a small group of mostly family-owned firms, called zaibatsu, came to dominate Japan’s economy. With mining monopolies or government-granted favors like government contracts or cheap capital, these zaibatsu got a head start that they used to diversify into conglomerates. At the height of their power, they had great influence in government affairs, including close ties to political factions and branches of the armed forces. History’s megacorps have derived their power from government favor, but the leverage provided by easier access to capital and new technological capabilities makes bootstrapping power more feasible for private organizations today.

Realistically, no organization without the ability to challenge the state’s monopoly on the use of force can transcend its traditional legal framework without permission. Power must be granted by government or taken via regulatory capture. Even in the wildest scenarios, Amazon has no ability to influence government against strong resistance.

To avoid antitrust action, Amazon must integrate with the state, and in doing so, would seek to come off the better—winning more new power than it loses in the coalition. If it coordinates with the state, it could protect its existing businesses from competition, stave off antitrust enforcement, and influence national policy. Even if Amazon doesn’t wish to use a coalition with government advantageously, it is still under pressure to coordinate to protect itself from government action.

Amazon and Bezos have taken action in recent years that can be interpreted as attempting to integrate with the state. Amazon and Blue Origin, Bezos’ rocket company, have contracts with government agencies. As one of the country’s top lobbying spenders, Amazon has sway with elected officials. The company’s second headquarters will be built across the Potomac from the National Mall. Employing a planned 50,000 people, this would make Amazon the largest non-federal employer in the DC area. This could build significant influence over the long-term by bringing the DC social scene within Amazon’s gravitational sway. A similar objective may motivate Bezos’ recent purchase of a Washington museum, which he plans to convert into a private residence and, the Washingtonian reports, “a veritable Death Star of Washington entertaining.” Most notably, Bezos purchased the Washington Post in 2013.

The Antitrust Battle

Whether antitrust enforcement can check Amazon’s growing economic, political, and cultural power is unclear. Antitrust policy is not static; judicial interpretation, public opinion, and contemporaneous economic perspectives influence how it is applied. Each era in the history of business, since the Sherman Antitrust Act of 1890, has been accompanied by a distinct antitrust interpretation.

Since the 1980s, the dominant perspective has been the Chicago School—advanced by legal scholar Robert Bork. This view holds that the only normative objective of antitrust is maximizing consumer welfare, which amounts to economic efficiency as measured by lower prices and greater output. It supports the use of microeconomic theory to evaluate the results of industry structure and conduct. It’s generally assumed that charging monopoly prices is difficult because of new entry into the market, so price cuts and vertical mergers that would previously be considered predatory are seen as pro-consumer.

Under this paradigm, there is no basis for antitrust action against Amazon. Amazon’s mere size is immaterial because agencies are only concerned with its conduct and how it affects consumers, and Amazon has been good for consumers as measured by the prices it charges.

Not everyone agrees with the Chicago School’s assumption that economic efficiency should be the chief concern of regulators. Others care about the interests of workers, competitors, partners, and consumers. Excessive dominance of a single firm creates unhealthy systemic risk. A shift in public, judicial, and economic opinion may elevate other concerns and evolve the evaluation criteria of antitrust agencies. The last change in antitrust thought came from Bork’s The Antitrust Paradox, his arguments before the Supreme Court as Solicitor General, new Supreme Court judicial interpretation, and regulatory agencies revising merger guidelines.

Amazon has been criticized for advantaging itself over third party sellers on Amazon.com. When Amazon identifies a successful product, it may go directly to the manufacturer and undercut the seller. This ruthlessness is off-putting, but it’s allowed because it saves consumers money.

Amazon’s hyper-competitiveness is often mistaken for anti-competitiveness. Amazon is certainly brutal to partners and competitors. In his book, Stone describes employees quitting because they couldn’t stomach the negotiations. But consumers have never been collateral damage in this conflict. They’ve been the chief beneficiary.

Previous interpretations of antitrust would view Amazon’s actions as predatory pricing—charging below cost to kill competitors, gaining a monopoly, and then charging monopoly prices. The Chicago School holds that in practice this would attract new competitors with lower prices. Amazon successfully made aggressive price cuts to force Diapers.com and Zappos into a sale, but they have not charged monopoly prices, though they have since reduced or ended the discounts.

Amazon is unusually patient, which changes how it monetizes its market power. The scope of its operations provides hidden ways to flex power without raising prices—namely by using its size to secure better terms from partners and improve profit margins. Amazon’s staying power is such that it surely has many decades over which to recoup losses. Even if Amazon never recoups losses in specific categories, having the largest selection is critical to its model of becoming the only place to shop for anything online. Viewed this way, these price reductions are like grocery stores’ loss leaders to get people in the door.

Whether a firm can sustain monopoly pricing depends on the ease of competitive entry into their market. Amazon has built such immense moats that credible competition is unlikely any time soon. It’s vertically integrated, so competitors need to compete not just on the retail front, but also with Amazon’s logistics and Prime advantages. Firms competing with Amazon in one area might be reliant on them in another. For example, Amazon could deny service, increase prices, or reduce quality of Fulfillment by Amazon for competing retailers who need Amazon’s logistics infrastructure. Amazon can also use its power in one industry to increase its power in other industries, which is the principle behind its flywheel.

But incumbents are never as secure as they appear. It just takes missing one new technology, or becoming too adapted to a certain way of doing business, to age into irrelevancy. If Amazon began to harm consumers with little possibility of competition, this would certainly move popular sentiment toward a relevant reinterpretation of antitrust policy, and there’s growing support for this even now. Unless Amazon is sufficiently coordinated with the government or is otherwise able to prevent it, regulators may step in.

The subject of regulation and antitrust action against Amazon has recently gained momentum, but no perspective has been raised that mitigates the threats posed by the firm’s dominance without harming consumers. Senator Elizabeth Warren’s proposal to, among other things, reverse Amazon’s 2009 merger with Zappos is unmoored from a theory of consumer harm resulting from the merger. Extracting Zappos from the Amazon umbrella would negligibly affect Amazon’s overall selection and market power, but would disrupt Zappos’ operations, harming customers.

Legislators may instead choose to address more narrow issues with Amazon’s dominance. There may be public support for establishing oversight of how platforms are operated, such as how search results are shown, or what the rules of participation are. Facebook’s recent decision to establish an independent review body for content moderation decisions shows that near-monopolist platforms may prefer abdicating responsibility for tough calls to external groups. Smart home devices and artificial intelligence, especially facial recognition, present new challenges to individuals’ privacy. Walling off data between Amazon’s divisions may mitigate some privacy concerns without impacting consumer welfare.

Amazon may preempt government action against it. In October 2018, Amazon raised the minimum wage paid to workers to $15/hr—which would likely have been forced on the company within a few years—to put wage pressure on competitors and buy political goodwill. Because of the organization’s service-oriented architecture, its teams can continue to cooperate as before, at least in principle, if they were formally broken up. It’s been suggested that Amazon may choose to break up on its own terms if it felt antitrust action were coming. The company is deepening its approach of structuring various departments like separate firms, a continuing strategy which appears to be coming from the top down. Services like AWS are treated as if they were external clients. At the same time, dependency on truly external services, from Oracle software to shipping carriers, are kept at a minimum. Such a move would place Amazon ahead of both the U.S. government and its fellow more-centralized tech giants, like Facebook, in the event of a large anti-tech crackdown.

* * *

With Bezos still at the helm, Amazon retains much of the genius and culture of its founding, but drift is inevitable—as is Bezos’ departure. The pathologies of companies will befall Amazon.

Amazon may continue toward benevolent monopoly, sustaining pro-consumer, positive-sum businesses with continued innovation, or it may attempt to reap monopoly profits and gouge customers. As Amazon navigates its relationship with the U.S. government, it must decide what it wishes to gain, and what it is willing to lose. Amazon’s future is still firmly in Bezos’ hands. He has the confidence of investors and employees now more than ever before. But the long run will depend on the strategy he sets, how he manages the evolution of the organization, how he handles succession, and the legacy that he wants for himself.

He’s presented a dream, which first captivated him as a child, of a solar system filled with one trillion humans in space habitats, with the Earth reserved for wildlife and leisure. There are science fiction writers who could fit a malicious corporate power into such a world, but it would sully an otherwise hopeful vision of the future. Which future Bezos’ corporate empire ultimately works towards for the remaining decades of his life depends on the nature of his ambition.

Chris Gillett is a data scientist and writer studying economics at Texas A&M. He can be reached on Twitter at @ChrisRGillett.