A few thoughts on the national security justification for industrial policy
To: Lockheed Martin Government Affairs Team
From: Your friends at the YLS School of Military Strategy
Date: February 20, 2025
Subject: Software eats the world
The United States Department of Defense is both the largest employer and the government agency with the largest budget in the entire world. As the regulatory and business outlook for software companies’ traditional lines of business (consumer advertising) becomes less compelling as a source of continued growth, it is reasonable for the tech industry to look toward military contracting as a new economic opportunity.
I am concerned about some of the ideas promoted by tech companies pursuing the strategy I described in the previous paragraph. I’m harmless, well-behaved, and deep inside the Overton window of American politics, so I of course have no problem with corporations discharging their responsibility to create value for shareholders in whatever legally-compliant way they see fit. However, I do want the fine readers of my blog to be equipped to identify the weaknesses associated with the burgenoning “defense-tech” and “industrial policy” discourses. It is in that spirit that I would like to discuss several concerns I have with the arguments put forth by tech companies and their lobbyists on why they should be cut into our government’s massive defense spending.
Ball bearings for freedom
Since the 2016 election and Washington’s conversion to universal hawkishness on China, there have been calls from across the political spectrum to address the United States’ lack of industrial capacity to engage in extended war. As far as I can tell, the idea here is basically that the United States has lost the ability to physically make a lot of the components necessary for producing enough materiel to replace stocks destroyed by extended combat with the PLA.
Doyens of the right like Tom Cotton have emphasized the importance of building up a defense-industrial base to fight the Chicoms.
This is a bipartisan publication. Center-left people like Jake Sullivan have made similar noises:
Any war with a country like the PRC, a military like the PRC, is going to involve the exhaustion of munition stockpiles very rapidly. So a big part of the answer to a healthy defense industrial base over time is the ability to regenerate, to surge, to build during a conflict, not just to build before to prepare for a conflict. And that’s got to be a key lesson that we take away from what we’ve seen over the last three years on the battlefield in Ukraine.
I find the idea that we would ever choose to engage in extended all-out combat with the PLA, and a fortiori that we should order our economy to be able to do this, absolutely deranged. I wonder if those invoking this possibility in order to lobby for DoD handouts understand how many millions of Americans would be killed, dismembered, and impoverished by such a destructive prospect. I know some people say we have to endure such warfare in order to defend one-party states like Japan and Vietnam from illiberal CCP agression, while others would admit that all-out conventional war with the PLA is a terrible idea but say we have to be ready for anything.
I agree that we should be ready to defend ourselves and our interests in case the PRC decides to threaten either, but that doesn’t mean we have to either engage in extended conventional warfare with the PLA or have an industrial base capable of supporting such warfare. More importantly, I believe this conclusion holds no matter how much you believe China is an aggressive expansionist power bent on destroying the international world order.
Allow me to illustrate my view with a case from 20th century history. When President Dwight D. Eisenhower took office, everyone agreed that the Soviet Union possessed conventional military superiority in Europe. In plain English, that meant that if the Soviets decided to launch a ground invasion of Europe, the United States government, and Moscow itself, believed the Soviets would certainly win.
Ex-CIA Director and Secretary of Defense James Schlesinger recounts:
We, in the judgement of the administration, could not stand up against the hordes of Soviet soldiers that would be sent against the West…[for] NATO intelligence, the prevailing presumption was …that the Soviet Union would be bending every resource to build up their military forces against the West as rapidly as possible.
I think this situation is a lot like the one defense-industrial policy proponents allege the United States currently is in with respect to China. We have a rising superpower with 1) undoubtedly expansionist aims 2) a commitment to gaining territory in which the United States has a strong interest, and 3) superior industrial capacity to fight and win an extended war to acquire that territory.
Eisenhower, like Trump today, faced calls from what was not yet then known as the “military-industrial complex” to massively build up the defense-industrial base in order to respond to the Soviet threat. But Eisenhower–wisely, as I will argue–declined these calls. He even decreased the size of the standing military and balanced the budget by reducing defense spending. Eisenhower was able to succesfully advocate for these things because he understood the strategic importance of the United States’ most destructive military asset: nuclear weapons.
Beginning with NSC 162/2, Eisenhower depended on a new strategy based on countering the Soviets by threatening massive nuclear retaliation should they invade Europe:
The result of the Eisenhower review was the emergence of a deeper dependence on nuclear weapons and long-range airpower to deter war. Eisenhower chose not to maintain all of the very large Army and Navy that had fought the Korean War. He chose, rather, to invest more heavily in airpower, especially Strategic Air Command, in large part because that kind of defense could be built for lower cost.
IR Professor George Quester puts a finer point on it:
Massive retaliation, the way Dulles and Eisenhower used it, in a way was a suicide pact. It said if the Russians invade Western Europe, “we’re going to hit back with everything we have. We’re not going to waste a lot of resources on trying to build tank defenses”…Depending on massive retaliation made European economic growth continue longer…we’re all richer today because [of it].
I hope the reader can see why Eisenhower’s choice was a wiser move than trying to go pound for pound with the Soviets in Europe. The Soviet Union was a non-democratic command economy that could coercively suppress its population’s consumption in order to fund military buildups. Another problem was that the United States had to defend both itself and many non-European theaters, meaning that forces committed to Western Europe could only ever be a fraction of total force availability. With Europe still devastated in the 1950s, both of these factors meant that any effort to keep up with the Soviets in Western European military capabilities would be outrageously expensive for the United States.
Samuel Huntington recounts:
The United States, the President is reported to have argued, must have sufficient military strength, but a “prodigal outlay of borrowed money on military equipment could in the end, by generating inflation, disastrously weaken the economy.”
Defense-industrial policy advocates are right to identify that an industrial base is a compoment of a nation’s military strength. But that doesn’t mean that we need to have as big of an industrial base as China or that we should try to compete with them in their own backyard. Eisenhower was not a pushover when it came to the Soviets, but even he could see that it would be better to draw a clear line beyond which they could expect massive retaliation and thereby avoid massive military expenditures of our own.
Even in a nuclear age, it is still important for us to have powerful conventional capabilities to deal with situations that require military force but do not rise to requiring a nuclear reponse. But the United States is already spending a highly abnormal portion of its GDP on defense, which is to say we already have a massive reserve of conventional capabilities that can inflict devastation anywhere in the world. I don’t see why we really need, on top of this, to spend money building factories that can churn out munitions in year five of an extended war with the PLA. Any threat dangerous enough to justify that kind of warfare would be better addressed through credible nuclear deterrence.
Chips and AI don’t blow things up
An argument related to the idea to the PLA defense-industrial base stuff is the idea that computer chips and software have some great military significance that justifies subsidizing domestic corporations’ uncompetitive production lines for chips and other technology products. The supposed military importance of chips–and their necessity for developing AI–is often discussed in light of drones, so let’s begin our analysis there.
Everything I’m about to say on drones is mostly a summary of this excellent article Why Drones Have Not Revolutionized War: The Enduring Hider-Finder Competition in Air Warfare from ETH Zurich’s Mauro Gilli, so you should just read that if you want a truly rigorous exposition of my point.
Gilli’s analysis begins with the general kinetic features of air warfare. Dropping ordnance from the air allows militaries to inflict enormous and well-targeted destruction, but in order to be able to do this, aircraft must evade an enemy’s air defenses. As General William DePuy put it: “what we see we can hit; what we hit we can kill.” The only force limiting the operation of this maxim is enemy air defenses.
These factors give rise to what Gilli calls a “hider-finder competition”: a technological arms race between an enemy’s ability to detect aircraft and the aircraft’s ability to evade enemy detection. The driving force in the former case is radar development, whereas in the latter, aircraft hide by reducing their radar cross sections, being fast/maneuverable enough to evade enemy anti-air fires, and by locating and/or destroying enemy air defense systems.
What that means is in order to succesfully penetrate enemy airspace, aircraft must have extremely sophisticated properties that allow them to accomplish all of those tasks. They must also carry enough ordnance to be able to destroy their targets. If one imagines a highly fast and manueverable low-RCS aircraft, equipped with sophisticated radars and the ability to communicate with ISR assets, and capable of carrying heavy ordnance, then one is basically thinking of a contemporary strike fighter.
Drone proponents believe that drones will outperform and obviate these strike fighters because they are cheaper and don’t need a pilot. On the first point, I hope it is clear that all of the features that make fighters expensive (e.g. supersonic engines, low-observable airframes, datalinks to ISR, low probability-of-intercept radars) would have to incorporated into a drone if it is to succesfully penetrate enemy airspace. In other words, drones are currently cheaper than strike fighters because they are slow and are unsophisticated, but being fast and sophisticated, and therefore expensive, is a necessary condition of being able to conduct modern air warfare. If drones are ever to become an important part of offensive airpower, they will need to take on the expensive characteristics that strike fighters currently have.
On the issue of pilots, there are two important points to note. First, none of the combat drones currently in operation are autonomous, so they still require pilots, just not in the cockpit. It’s reasonable to assume that the costs of training these pilots would be similar to manned aircraft. It’s also worth pointing out that these uplinks are vulnerable to electronic warfare, an issue which has been so difficult in the Russia-Ukraine war that both sides have been known to connect drones to their operators with physical wires.
But even if one assumes that the electronic warfare issue will be solved or that autonomous combat aircraft will be developed, there’s no reason to think that will be a particularly big deal. It costs about $10 million to train a fighter pilot for the U.S. military, and one can assume that the aircraft have to be a little bigger to accomodate a cockpit. That’s a lot of money, but compared to the total flyaway cost (which doesn’t even account for R&D expenditures or maintenance costs) of a modern strike fighter like the F-35, one is only looking at 10-15% cost savings by avoiding the need to train a pilot. That would be very nice, but it’s not exactly a strategic gamechanger. So, the idea that we need chips or cutting-edge AI labs because drones are about to revolutionize warfare doesn’t make a lot of sense.
One can abstract these observations beyond from the case of air warfare and drones. When we see that warfare is defined by kinetic realities and not whether a technology product sounds cool to gullible podcast listeners, we can justify a more skeptical attitude towards the military uses of popular technologies like generative AI or uncompetitive domestic chip production lines. Of course, chips are necessary in military systems like radars and fly-by-wire, but the number of these systems that the military needs are numbered in the thousands or tens of thousands, not the 8-figure annual production numbers achieved by superstar firms like TSMC. There’s really no reason to believe that existing production capacity won’t be able to meet the comparatively miniscule (and non-cutting edge) production requirements for military uses.
At this point grifters people will take issue with my dismissiveness about the imminence of the AI-singularity or will posit other technological frontiers that we need to start worrying about. They will say that although the technologies of which I’m being skeptical haven’t revolutionized warfare yet, they will soon, so we have to invest in them now in order to stay ahead of the curve. That’s an impossible argument to refute, since it relies on premises that it acknowledges aren’t true right now, so all I’ll do is note that defense spending is almost half of all government discretionary spending. I’m a W-2 wagie without access to any offshore tax shelters, so anyone saying we need to “invest in transformative technologies for national security” is effectively asking for 17% of my salary (37% federal income tax * 47% share of defense spending in discretionary budget). If you want me or other voters to give you that money, you’re going to have to do a lot better than simply invoking millenarian sci-fi tropes like transhumanism and superintelligence.
An American Affairs article I read today provides an excellent example of this pattern of thinking. The entire article is about how our lack of chips production capacity is a national emergency that requires large taxpayer expenditures to address. But the article concedes that the military currently does not need very many or very advanced chips (“the existing DoD inventory of weapons and systems largely does not require the most advanced chips used in commercial equipment”). So then why is a lack of domestic chips production a national security emergency? The authors offer us the following incredible passage:
It is increasingly clear that the emerging technologies that will decide future conflicts will once again require the most advanced digital logic and memory chips. Artificial intelligence, quantum computing, advanced data centers, and autonomous drones, for example, are now realities, and will require ever faster, ever more powerful and ever less power-hungry chips, and nations with early and assured access to such chips will be able to seize technology dominance.
- There are no fully autonomous drones in military use today, so they are certainly not “now realities”.
- Quantum computing is 1) as of now totally unusuable for anything practical and 2) has limited connection to the transistor-based chip production they propose to subsidize.
- As for the other “emerging technologies that will decide future conflicts,” the authors provide no argument for how artifical intelligence (apart from drones) or “advanced data centers” will be militarily useful.
This little bit of specious reasoning (“increasingly clear”) is all the authors have to offer as to why it’s a national security necessity to spend our money to maintain domestic cutting-edge chip production.
For now, conventional military dominance is defined by one’s ability to produce and fire physical ordnance like gravity bombs and artillery: the two technologies–which, despite all the hooplah about drones–still make up the vast majority of fires employed in the Russia-Ukraine War. As I said in the previous section, I do think we need to maintain a limited level of production of these items to address situations in which nuclear retaliation would be an overreaction. But I certainly don’t think taxpayers need to fund a bunch of other stuff related to chips and AI, at least not until there is an actual demonstrated military use for them. National security is too serious of a business for vague techno-optimist speculation.
Industrial defense production won’t uplift the American worker or elites
If you buy my arguments that the national security case for ramping up subsidized production lines for industrial equipment and chips doesn’t really make sense, then we have to ask why these arguments are so popular. Of course, as the opening of this post describes, there are lots of people who have an economic stake in getting the DoD to fund their companies. But lots of intellectuals with no such economic exposure to DoD contracts are also promoting it. Why are they continuing to do this even though the military case for it doesn’t make much sense?
I think the answer has a lot to do with the fact that some intellectuals implicitly regard industrial policy not as a demand of national security, but rather as a way to free American labor from the grinding inequality of a debt-financed tech-platform-dominated service economy. I’m sympathetic to this aim, but I don’t think that the kinds of production that would be subsidized under the most ambitious defense-industrial policies would actually accomplish it.
I currently live in New Haven, which is in many ways an excellent metaphor for an American economy made up of disempowered laborers whom deindustrialization has reduced to serving lattes and delivering takeout to a small class of knowledge workers. Even the “elite” pole of this economy doesn’t have a great deal, as they spend their time nervously competing to access an ever-narrowing band of prestigious credentials and firms. Most of this competition is experienced as compulsive anxiety interspersed with repressed thoughts of how meaningless all their work is. YLS Professor Daniel Markovits’ book The Meritocracy Trap, is an excellent description of the sociological situation I’m trying to describe.
It’s hard for me to see how defense-industrial policy would make this situation better. The obvious mechanisms would be 1) laborers can get good jobs in the factories and 2) directing artillery shell production would be more fulfilling for elites than doing financial transactions or technologically-sophisticated advertising. The problems with (1) are, first of all, that I’m not sure those jobs would actually be that great.
The United States has a romantic attachment to an age of industrial production in which workers were paid considerable monopoly rents in addition to their marginal productivity, but if you assume that defense-industrial production doesn’t obey such monopoly dynamics, then working for long hours on one’s feet in toxic and dirty environments is not actually a great deal.
But even if these jobs are great for the people doing them, that doesn’t make them a net benefit for the country. That’s because however you do the accounting, to the extent a given production line requires government subsidies, then the rest of the country has to pay a cost in order to have it exist. Whatever economic benefit accrues to these workers beyond their marginal productivity would be borne by the rest of the economy. The cost for everyone else would at least balance out and even potentially exceed the benefits to the small percentage of workers employed in such work. And before anyone says the rich would bear the tax burden, they should recall that America already has one of the most progressive tax systems in the world, making it so that any large expenditures would inevitably be borne in large part by the middle class through higher taxes, inflation, or both.
It’s important to distinguish this observation from a general critique of industrial policy. One very well might think that it makes sense to subsidize certain industries in order to give them time to develop comparative advantages against global competitors. But that’s not what’s being proposed here. No one is saying that American factories just need some time in order to get good at producing materiel for the market; the argument is simply that we should produce ad nauseum for our own military so we can fight the PLA at some point. That is to say, what’s being proposed is not a limited period of subsidies eventually leading to the emergence of a competitive industry; it’s just a perpetual transfer from taxpayers to firms, a set of cashflows of which there’s no reason to think workers will ever see a substantial cut.
Another argument is that elites will have more fulfilling careers in defense-industrial production than they will in a post-industrial service economy. I won’t defend the current elite offerings of private equity, plastic surgery, and optimizing algorithims for screen time addiction, but two wrongs don’t make a right. There’s no reason to think defense-industrial production is a wise use of resources just because existing options are bad. I personally don’t think I would find very much fulfillment in funelling taxpayer money to uncompetitive firms who survive by making tools for Americans and Chinese teenagers to murder each other on beaches in a competition over which nationality of oligarchs will extract IP rents from TSMC.