Actually, things go wrong even when the Pentagon does attempt to catch up to the advances that are prevalent in the commercial world. A classic case that Brose notes is that of DoD’s effort to bid out a contract for an enterprise cloud that would serve the entire Department as opposed to only some of its component parts. To do so, DoD sought to contract for what it labelled the Joint Enterprise Defense Infrastructure, or JEDI.

Several of the largest hi-tech firms—including Amazon, Oracle, IBM, Microsoft, and Google—expressed interest in bidding for what was to be a major $10 billion award. Significantly, Google dropped out of the bidding after its employees protested that working with DoD violated its “corporate values.” Although the award seemed imminent in the summer of 2019, it became subject to a combination of bureaucratic and political entanglements. First, Oracle protested the impending award to Amazon, for whom the contract appeared to be “wired.” The senior judge of the U.S. Court of Claims responded by putting the award on hold. In August 2019, several weeks before the winner was finally to be announced, President Donald Trump intervened to put the award on hold once more, ostensibly to have Mark Esper, who had been confirmed as secretary of defense a few weeks earlier, review the contract before the Department determined which bidder should win the contract.

As Brose notes,

it was only in October 2019 that the department finally awarded a contract to set up an enterprise cloud, which quickly became embroiled in official procurement protests stemming from Donald Trump’s attacks on one of the competitors, Amazon, and its founder Jeff Bezos.

Bezos owns the Washington Post, a frequent critic of the president, and no one mistook Trump’s intervention as motivated by technological rather than political concerns. On October 25, DoD awarded the contract to Microsoft.

Brose’s account of the JEDI mess ends there. In fact, however, the story is far from over. Amazon turned and protested the award to Microsoft. The following February, 2020, a day before the Microsoft system was meant to go live, a federal judge put a halt to any further activity until the protest was resolved. In response, the DoD promised the judge to quickly reopen bidding on part of the contract. That did not satisfy Amazon, however. Early in May, Amazon filed another protest, challenging the DoD’s handling of its reconsideration process and arguing that the department’s revision was too vague to permit Amazon to price its new bid with any degree of accuracy. Nevertheless, the DoD gave no indication as to whether it would revise its “reconsideration process.” Meanwhile, the Pentagon continues to operate without an enterprise cloud, years after it had first begun to identify the need for one.

The irony of the JEDI fiasco is that, as Brose notes, “the information revolution has moved beyond the cloud to what is called edge computing.” He buttresses his argument with the case of Nvidia, a firm whose graphics processing unit is at the heart of increasingly higher resolution video games and that is building mini-supercomputers for autonomous vehicles. He writes,

when equipped with well-trained machine learning algorithms, Nvidia’s computers enable vehicles to make sense of the myriad events that happen every second on congested roads and perform complex, time-sensitive actions, such as maneuvering through city streets … all of this information is being processed right where the vehicle collects it and needs it.

It is this process that is called edge computing; “a better description,” he notes, “might be machine brains.” 

Nvidia’s technology, and others like it, outstrips the most capable computers onboard the DoD’s most advanced and complex weapons systems. Brose points out that Nvidia’s DRIVE AGX Pegasus, which conducts 320 trillion operations per second onboard a car or truck, represents 800 times the computing power of the F-35 fighter, the DoD’s most advanced aircraft. What is even more troubling is that, as Brose observes, “compared to the rest of US military programs, when it comes to being an intelligent system the F-35 is light years ahead.” Most weapons systems cannot themselves process the information they collect. That takes place elsewhere, and is a lengthy and painstaking process that humans, not machines, carry out.

One might have thought that DoD would be eager to exploit Nvidia’s technological breakthroughs. The Pentagon continually assures both the Congress and the public that it is seeking the closest possible relationship with the commercial high-tech sector. Nevertheless, Nvidia does not do any business with the Department of Defense. This isn’t because of Nvidia’s lack of interest or ideological misgivings about working with the military. On the contrary, Nvidia would very much like to partner with DoD. Rather it is because the DoD has yet to develop a consistent formula for interacting with smaller companies and start-ups that are not part of its traditional industrial base.

AS A result, it is China, not the United States, that has devoted massive resources to capitalize upon these technological developments. Even as the Pentagon has been fixated on the “war on terror” for the past two decades, China has begged, borrowed, and stolen many of the American-developed hi-tech capabilities in the process of reinventing its military to compete on equal terms with that of the United States and to break the American “kill chain.”

Brose is at pains to point out that machines will not completely replace humans on the battlefield. Instead, machines will enable humans to focus on strategy and to optimize both operations and tactics. He also posits that it is critical that new technologies be subjected to sufficient operational testing to warrant the trust of troops in the field. Brose concedes that machines will be no better than the algorithms that control them and that they will be subject to error. Nevertheless, with adequate testing, machines will be far less prone to errors than humans who most likely would be overwhelmed by the mass of information they must evaluate in the shortest possible time. Indeed, in the absence of high-speed connective networks, humans generally would have to do so without the ability to fuse their findings with material that other systems might be generating at the same time.

Even as the DoD has yet to exploit commercial high-technology to its fullest, it continues to acquire ever more costly systems at increasingly lower rates. While Brose acknowledges that the U.S. military is more capable than ever, it is also shrinking at a frightening pace. Moreover, whereas in the past three decades the United States has had time to build up its forces, faced only third-rate opposition on the battlefield, and benefitted from unchallenged command of the skies and the seas, that will not be the case when facing Russia and especially China. Quality is important, of course, but numbers matter as well, and Brose demonstrates that America is simply ill-equipped to generate in a timely manner the force levels that would be required to face its peer competitors.

Brose offers a number of reasons for the Pentagon’s inability to exploit the technologies that could be made available to it. To begin with, in the past the DoD was prepared to place big budgetary bets on a relatively small number of systems without hamstringing developers with a myriad of rules and regulations that the bureaucracy, or Congress, or both, might have generated. Brose writes of General Bernard Schriever, who, together with the former Nazi scientist Wernher von Braun (whom Brose does not mention), led the development of long-range ballistic missiles despite numerous test failures and bureaucratic opposition thanks to President Dwight Eisenhower’s unstinting budgetary support. “And,” notes Brose, “they did it all, from start to finish, in just five years.”

Similarly, money was not an issue nor regulations a barrier for Hyman Rickover, father of the nuclear-powered submarine and—not mentioned by Brose—the nuclear-powered aircraft carrier. Kelly Johnson, who headed Lockheed’s so-called “Skunkworks,” led the development of several aircraft, including the SR-71 reconnaissance plane, the fastest ever flown. All of these developments took place during the Cold War, when the Soviet Union posed both a military and technological threat to the West; it was the Soviets who launched the first satellite and made Sputnik a name as widely recognized worldwide as Michael Jordan decades later. But then, as Brose archly notes, “this was how America acted when it was serious.” He argues forcefully and convincingly that this is how America should act again.

Why, then, has the Pentagon failed to engender the same spirit of dynamism and experimentation that it once did? Why is its acquisition system so hidebound that, as Brose recounts, the DoD spent two years and $17 million to test a pistol? Brose offers a powerful and cogent series of explanations that alone would make the book a worthwhile read. He points to the flaws in the DoD’s programmatic processes and the prevailing bureaucratic caution that penalizes experimentation and risk taking. And he notes that both are buttressed by Congressional preferences for maintaining job-creating programs whether or not those programs best support the military’s efforts to retain its superiority over all comers.