The Endless Fantasy of American Power
Neither Trump Nor Biden Aims to Demilitarize Foreign Policy
In late October, Germany and China began commercial-scale rollouts of 5G, the wireless technology infrastructure that is transforming the way the world computes. Machines and people will still talk to each other over the borderless network we call the Internet. But with 5G, a new networking infrastructure is emerging, dependent on the Internet but distinct from it and subject to much more government and private control.
With 5G it is possible to do enormous amounts of computing at very high speeds and without having to connect the input device—a cell phone, say, or a self-driving car—to a wire of any kind. But those high speeds are possible only if the rest of the system (signal towers, base stations, distributed servers, and the megascale centers that house the data and do a great deal of computing themselves) is physically near enough to these input devices. Having your phone, car, or pacemaker in constant contact with vast computational power in the so-called cloud sounds amazingly untethered and extraterritorial. Yet in its physicality and focus on location, the emerging system is more grounded than the Internet ever was.
Whether control over 5G will be exercised principally by states or companies remains to be seen. But the implications for surveillance, security, and national prosperity are enormous, and yet policymakers and business executives have hardly begun to address them.
The Internet has proved remarkably resistant to state governance. Its use can certainly be shaped by expensive government initiatives such as China’s Great Firewall or the European Union’s General Data Protection Regulation (GDPR). But multilateral attempts to control the Internet itself have so far failed, mainly because the deliberately impenetrable “global Internet community”—including Internet service providers and sui generis governance institutions such as the Internet Engineering Task Force and the Internet Society—is dedicated, in the best geek spirit, to avoiding state capture. That posture may change, but for now the community’s obdurateness and the jealousies of states, which have kept global Internet governance negotiations singularly unproductive for years, have combined to keep the Internet an open global platform.
In its physicality and focus on location, the emerging system is more grounded than the Internet ever was.
The Internet, of course, only moves data; it doesn’t store it. Servers in data centers do that. But even in the world of physical servers, location has been less and less important—at least, until the emergence of 5G. Take Amazon Web Services (AWS), today’s leading data storage company. Amazon established the subsidiary in 2006 in order to put its idle server space, built for holiday sales surges, to more regular use. The AWS idea was to link that server space into, in effect, a single “virtual” machine, available for rent in any place at any time. The actual machines involved could be anywhere, distributed (or not) across the globe.
Once AWS refined its product, this kind of computing—soon known as “cloud computing”—became very popular with business owners eager to reduce their budgets for information technology departments, servers, software suites, updates, patches, and then more servers. You could store your website and data entirely in the cloud and access it instantly from a laptop using the Internet. AWS soon became, and remains today, Amazon’s only major profitable unit.
This cost-efficient model proved a boon for innovation. Instagram, with 13 employees, was able to store some 100 million images on the cloud. Lyft, Pinterest, and Slack all relied on cloud computing to build their companies. And once phones could go, through apps such as Instagram, to the cloud and back, they became extensions of its vast computing and storage capacity. In fact, you could not only store your data but also do your computing in the cloud: services that would otherwise have been purchased as software and stored on dedicated servers were now performed remotely.
Google and Microsoft, and then the Chinese e-commerce and tech giant Alibaba, began their own cloud projects to compete with Amazon’s AWS. By 2011, major corporations were “migrating to the cloud.” In a landmark vote of confidence, the CIA decided to do the same (with AWS) in 2013. The U.S. Defense Department’s decision last month to go all in on cloud computing (using Azure, Microsoft’s answer to AWS) marks the model’s coming of age. The efficiency, redundancy, and security of cloud computing, and its capacity for encouraging and adjusting to technological innovation, are unmatchable. Last year, the value of the cloud-computing business was more than $200 billion.
So far so borderless; but physical space still matters. A movie may be stored on the cloud, but the closer the laptop on which it plays is to the WiFi router, the faster the movie streams. The speed at which a signal gets to the computer and back goes by the name “latency,” and the lower the latency, the faster the computing. Even with fiber-optic cable, there are latency advantages in physical proximity (see Michael Lewis’s Flash Boys). With radio signals, such as from a cell phone, latency is much more important, because radio waves carry vastly less information, at slower speeds, and with more interference than is the case with cable.
If proximity still matters for cloud computing, it is essential for 5G networks. The magic of 5G is its ability to connect within milliseconds a phone or car or sensor-equipped locomotive to the cloud, which can process the data, combine it with other data, and return to the end device. For the process to run smoothly, the latency must remain low. That is a principal reason why computing is returning to earth.
What the Trump administration is doing to Huawei, and China once did to Google, is what the EU might do to AWS.
The vast sums of money flowing into far-flung data centers suggest as much. Argentina last month invited AWS to invest $800 million in a data center in Buenos Aires. Google will invest $850 million in a new facility in Taiwan. That these companies and their two main competitors, Microsoft and Alibaba, are willing to invest on this scale indicates that they recognize that neither data-localization laws (which require citizens’ data to be stored in their own country) nor the underlying principles of data sovereignty (the notion, associated mainly with China, that cyberspace should be national space) are going away anytime soon. But building the data centers is also a way to gain market share by lowering latency for users. The new centers create what are called “availability zones,” where radio-based computing can take place at ever higher speeds. If cloud computing is global, then availability zones are like local clouds, built to accommodate the latency needs of companies making ever greater demands on cloud providers—and, especially, the latency needs of whatever is communicating with computers through radio waves, which is a lot, including cell phones, the sensors of autonomous vehicles, and the numberless nonwired sensors out there in the Internet of Things.
If the mega-multinationals don’t stake out this territory, others will. There is a parallel world of (mostly) smaller and less-known companies building data centers around the world. Google and AWS lead the market, but they are followed by Digital Realty Trust and Equinix, neither one a household name. China, as you might expect, favors domestic cloud data centers, although U.S. companies are often involved. Much of the data-center growth in India and Africa has been led by smaller companies, frequently in partnership with global players. The basic technology is reasonably accessible, if expensive to build and to cool. (AWS got an energy tax break for its Argentina project.) The data-center business is consolidating, but it is still fairly open compared with, say, search engines or social networking.
The combination of 5G’s need for low latency and the competition in the data-center business points to the likely emergence of more local clouds as the new network settles on the land. But the giants in the industry have responded to this idea defensively, because it threatens their ability to scale across borders and develop services for the largest-possible market. Last month, France and Germany announced plans to build a “European cloud” that would give local companies preference over AWS, Alibaba, Google, and other multinational cloud providers. A spokesperson for AWS retorted that such a project would lack the scale to compete with the industry’s dominant players. “We think that the idea of a ‘national’ cloud is interesting in theory,” an Amazon spokesperson said, “but in reality it removes many of the fundamental benefits of cloud computing.” Microsoft was blunter, hitting at Europe’s sensitive point: “Real sovereignty requires the most powerful cloud solutions—otherwise Europe will only cement its digital gap.”
But multinationals, even market-leading ones, do not always prosper by mocking the sovereignty of governments. Maybe someday Google’s search engine will be allowed to return to China; then again, maybe not. Facebook famously flamed out in India in 2016 after a board member, tech legend Marc Andreessen, was offended by a little subcontinental pushback and wrote, “Anti-colonialism has been economically catastrophic for the Indian people for decades. Why stop now?” The EU’s GDPR and a host of antitrust fines levied by the bloc against Google and Facebook have essentially been salvos in a public-private fight over data control. Russia is implementing legislation aimed at taming multinational tech companies, even as it is welcoming the Chinese telecommunications company Huawei as a political-economic alternative to Western platforms. The administration of U.S. President Donald Trump hasn’t yet turned its attention to these issues, but it has demonstrated a willingness to attack the tech sector when it feels that governmental prerogatives are disrespected.
The cloud-computing, data-center, and 5G industries—for they are really becoming one single industry—need to start thinking ahead. Currently, AWS, Google Cloud, Microsoft Azure, and Alibaba Cloud each provides its own separate cloud products. Migrating from one to the other, or combining the services of more than one, is not exactly easy: these are competitive companies that want to lock in customers and lock out competitors. But if the new global network these companies are building is to actually be global, products designed for it need to be able to move smoothly across these corporate borders. Otherwise, the cars and phones and intricate processes of the Internet of Things will not be globally scalable: at some point in their supposedly ethereal progress across the earth’s surface, they’ll cease to compute. In effect, they will need corporate visas to cross corporate borders. This amounts to a redistribution of sovereignty from state to corporation so far-reaching that states will push back—as indeed they already are. The techlash is not a blip. What the Trump administration is doing to Huawei, and China once did to Google, is what the EU might do to AWS. Silicon Valley’s “tomorrow belongs to us” period is ending.
At the same time, governments want to be in the best networks, all of which are private. The 50 or so companies that together make up the core of the industry must find ways to harmonize their operating needs with the security and non-monopolistic competition that most governments want. If they don’t, states will likely wage a long war of attrition against global data-center-cloud-5G platforms, levying fines, imposing regulations, protecting domestic markets, throwing out foreign competitors, subsidizing national champions, and so on. The disruption this would cause in today’s networked societies is immense. Without a more concerted effort by the corporate great powers of the new network, states’ reactions will grow ever more unpredictable as they confront their real and perceived vulnerabilities.
The United States Needs to Compete With Chinese Firms, Not Just Ban Them