Unicorns for Uniforms: On the Problematic Allure of VC Investments in Defence

[Elke Schwarz is a Professor in Political Theory at Queen Mary University London, specialising in ethics of war and ethics of technology with an emphasis on unmanned and autonomous / intelligent military technologies and their impact on the politics of contemporary warfare]

The second REAIM summit was underway in Seoul last week. The event, co-hosted by the governments of the Republic of Korea and the Netherlands, seeks to advance the agenda of “Responsible AI in the Military Domain”. Invited are representatives from governments, military organisations, civil society members and industry representatives who will have an opportunity “to demonstrate the practical application of AI in the military domain”. Among them are often a number of startups which, together with their VC funders, have set their eyes on disrupting the defence market with their products. Given that the private sector is increasingly affecting military practice and culture, it is high time the possible upshots of these new influences are examined with a sober eye. Without scrutiny of the aims, interests and logics of these new actors, we may be missing a crucial piece to the puzzle for governing military AI effectively. 

Military AI is a lucrative and growing market. In 2023, it was estimated to be worth just under 9.2 billion USD and it is projected to increase to 38.8 billion USD by 2028. This growth is driven by swelling defence budgets across the globe, which are, in turn, driven by growing geopolitical tensions. As so often is the case, the U.S. is setting the pace for AI adoption. According to a 2024 Brookings Institute study, federal contract allocations for military AI have nearly tripled between 2022 and 2023, reflecting a potential increase in the value of contracts by a staggering 1,200%. Where there are new AI technologies, the world of venture capital, is seldom far. VC is involved in providing startup capital and funding for ventures that are promising but too risky for banks, for example. And, indeed, in recent years, the VC sector has turned its attention to military technology startups, in the hopes to benefit from current global defence spending worth 2.4 trillion USD, of which the United States corners nearly 40%. It is a market of significant growth in an era where the typically extraordinary returns for technology startups for the commercial sector are beginning to dry up. 

Although military endeavours, Silicon Valley and venture capital share an original set of DNA, emerging in the 1950s, venture capital firms have not been particularly interested in the defence sector in the recent past. Military acquisitions were tightly regulated and the market was dominated by the five primes who held a firm grip on the procurement and requisitions process for military needs. Additionally, there was a sense that it was frowned upon to profit in this way from the defence sector, so there was a moral cost to placing bets on war and warfare in this way. However, the speed with which potential investors put aside their qualms about this kind of profiteering from what is often euphemistically termed “kinetic” defense technology in the past few years suggests that this norm was not particularly robust to begin with. It is more likely  that the reputational cost was too high and the possible gains too little for investors and VC managers to spend time on defence. This has now begun to change, with potentially significant consequences for the business, and practice, of warfare.

VC investment in military startups is booming. VC money funnelled into defence startups in the US alone surged in the years from 2019 to 2022, doubling in size from 16 billion USD to 33 billion USD, with VCs putting 35.8 billion USD in startups across 800 deals in 2022. To this end, the Russia Ukraine conflict and other global tensions have been instrumental in demonstrating to observers “that you can adopt technology and experiment at a much faster pace”. Since 2021, the defence technology startup sector has been injected with nearly 130 billion USD in funding.  These are not small sums. Many VC companies place their hopes on unicorn startups like Anduril and Shield AI, among others. They are often comprised as the so-called SHARPE cohort, a number of unicorns on which the VC-for-defence market pins its greatest hopes for the creation of a VC friendly defence ecosystem. 

And although VC money for startups is still small fries compared to the overall market, I would argue that with the mandate to disrupt defence for gains, VC interests are starting to have an outsized bearing on the direction of military practice and culture which, effectively, benefits all with financial stakes in defence.

Defence Disrupted

Venture capital has different needs than other financial instruments in the defence markets. The primary aim for VC companies (VCs) is to make extraordinary returns for investors. VC firms are entities that raise money from so called limited partners (LP) which can be pension funds, endowments, insurance companies, corporations and high-net-worth individuals – to invest in promising startups or other venture funds. Investors provide money in exchange for equity which cashes out once the company becomes publicly traded or gets bought by a bigger fish in the sector. VCs do not deal in long term investments, but they do deal in high risk high reward opportunities. VC investment likes to present itself as supporting young innovative entrepreneurs in creating a better world. That may be true, but every VC portfolio is ultimately also a high-stakes bet. 

As many as 90% of startups fail, and that is built into the VC high-risk – high- reward structure, following the so-called power law. The few startups that make it big generate such outsized returns that it all works out quite spectacularly for the VC sector, or at least has in the last two decades. Every VC investor hopes to invest in what becomes a unicorn. Unicorns are startup companies that are valued at over 1 billion USD. The valuation process is not an exact science. 

Valuations are assigned in relation to people’s expectations about the company, it is a projection based on a number of parameters, usually a mix of qualitative and quantitative factors – market size, sales potential / current sales, management structure and others, that come together in a convincing story about the current and future value of a company. A successful startup must demonstrate ambitions to grow fast and be able to raise repeated rounds of funding via the VCs usually in 12 to 18 month intervals. With each round of successful funding valuations should go up and set the company on a path to hypergrowth before the VC fund comes to an end (typically 10 years). To attract VC funding and secure contracts, startups need to think big and need to declare big, ambitious and possibly unattainable goals. Key words here are ‘revolutionary, ‘ground-breaking’, ‘changing the nature of warfare’ type terms. The mandate for growth disincentivises prudence in start-ups and encourages making big promises that often cannot be realized and, importantly, the goal is to convince customers of the same. The VC logic plays a significant part in the current AI hype cycles.

Key here is that the product VC companies offer is not a particular technology, or the startup business-as-business, but rather the product VCs offer to their clients is growth. Growth in the valuation of a startup, to be specific. VC managers receive between 2-3% of the portfolio’s value annually for managing the funds, and, upon successful exit, they receive between 20 and 25% of the gains made. Growth at scale and speed is how VCs and their investors make money. That is the business. 

The temporal and structural requirements for VC firms, geared toward growth, create an incentive to over-value a firm – that is to say not evaluate a company’s worth based on its operating profits but based on some speculative, future-oriented value. This has been the norm within the commercial startup environment in the past decade, where VC companies have been able to make extraordinary gains. However, such over-valuations are not the usual practice in the defence sector, contracting in defence is a long, multi-year process to come to fruition – the opposite of the fast-paced commercial digital environment. The temporal and structural requirements for VCs, geared toward growth, in the defence space, needs to create an incentive to get customers to enter into contracts quickly, which helps bolster the valuation. This can best be done by shaping the defence sector in the image of Silicon Valley itself. 

In order to be profitable for the VC investors in the order of magnitude the commercial technology sector has provided, defence, as a sector, must be disrupted. Successful startups will need to be ambitious – to become a new prime, to gain a monopoly position which will allow for the perpetual extraction of monopoly rents. With this in place, valuations can soar, and with that, financial gains. AI focused companies like Palantir and Anduril, among others, make little secret of these ambitions. But as we know from technologies such as Uber or AirBnB, disrupting markets comes at cost for stakeholders and for regulative environments. Uber could only be successful (despite running billion dollar losses) by undermining existing labour laws, exploiting precarious worker’s rights to healthcare and unemployment and disrupting the taxi industry. Similarly, AirBnB shook up the hotel industry, drove up rental prices in popular tourist destinations and so on. In other words, disruption, and establishing a monopoly position often comes with wider social and political costs. 

Manifesting Myths 

VCs are often quite hands on in making this positioning possible and employ a tried and tested mix of modes of influence – unicorns don’t just happen, they are made. One such approach is to enlist a roster of former military and government officials to help with lobbying efforts to influence legislative and policy environments. Millions are spent by defence-oriented VCs in lobbying efforts. Another is to utilise the instrument of law to disrupt existing practices. Both SpaceX and Palantir and their funders have sued the U.S. government in pursuit of contracts, Anduril has reportedly been spending most of its money on lobbying and lawyers. The third and perhaps most potent effort is through a series of narratives – about the U.S. not being ready for the wars of the future, about the need for agility and speed in all aspects of military endeavours, first and foremost procurement, about the extraordinary historical significance of AI weapons, about the looming threat of a great power war, and about the need to win the technology rivalry with China at all cost. 

And perhaps the most potent narrative of all is that the new startup products will herald the much anticipated possibility of perfect knowledge; of knowing exactly who is friend and who is foe, and where they are, of a clean and transparent war; smooth and effortless connectivity; real-time insights, of tools of war as democracy, winning wars swiftly and decisively with AI. A fantasy of omniscience and omnipotence. It is terribly alluring. But it does not bear out against the history of warfare or the character of war as a human affair which abounds with subterfuge, surprise and subversion. But with these fantasies, the new military contractors shape ideas of possibility and create desires that everything that moves across the globe can be captured and controlled. These desires are manifesting in programmes like Joint-All-Domain-Command-and-Control (JADC2), Maven, and other multi-domain data connectivity projects which serve to realise ideas of being able to identify and intervene for potential threats and “irresponsible behaviour” across domains and geographies within minutes, rather than hours, thanks to AI. It is, perhaps no surprise that the U.S. DoD draws an analogy with Uber to illustrate the desired aim for the JADC2 programme. More generally, the demands of the VC-funded startup sector and its products refocuses attention, and funding onto systems that thrive on and with AI. Systems of surveillance and lethality delivered at speed and scale. 

The efforts invested by VCs and their startups to ““infuse [the Pentagon’s] clogged arteries with the nimble, agile DNA of Silicon Valley” are beginning to show signs of success. U.S. acquisitions processes are presently undergoing significant structural procurement reforms in order to better accommodate the products and timelines of various new defence software startups. Making the acquisitions process more agile, more flexible and faster is the name of the game. This does, however, come at the cost of accountability and oversight. 

Also well underway is a refocusing of defence culture toward experimentation, prototype warfare, the embrace of risk and error, to ““reward[] and recognize[e] trailblazers and mavericks”.  “Agility” has long become a buzzword for the U.S. DoD and it increasingly embraces the need for disruption. U.S. Deputy Defence Secretary Kathleen Hicks gave the keynote this year at an event organised by the VC company Andreessen Horowitz. At the event – the evocatively titled American Dynamism summit – closed her speech by bringing defence in line with Silicon Valley’s most iconic motto: “yes, moving fast and breaking things is necessary to win wars” (Hicks 2024) – the only thing that must never be broken is the law and the U.S. Constitution.

It is little wonder, then, that ethical considerations about war and its human consequences are shifted increasingly into the margins. Ethical concerns can neither be appropriately be accommodated by the aims foregrounded in VC endeavours, nor can they accommodate the mandate of growth-at-all cost and at high speeds. The technology sector is not particularly well known for its appreciation of, or adherence to regulation or ethical constraints. Quite the contrary. The often flippant ethos that comes with the technology sector and its startups, and the sometimes outright irresponsible attitude to violence and harm displayed by key influential figures in this space is a poor role model for the utmost seriousness of military engagement and the high moral stakes that are involved therein. And it might behove policy makers to resist the allure and promises new military startups and their VC backers make, or at the least take them with a grain of salt. Because the VC business model does not serve the public at large, it primarily serves the LPs and themselves. Such is the VC logic. 

Last year’s REAIM summit featured Alex Karp, CEO of Palantir, a highly successful former defence startups, in a plenary conversation with a Financial Times journalist. In the conversation, Karp oscillates between making disparaging remarks about Silicon Valley and PowerPoints, dropping pop-philosophy references and engaging in enthusiastic praise for the values of the West and its magnificent AI products for offensive targeting, made by Palantir. But his core message was this: more money needs to be spent on software products, now. It is urgent. Europe must get serious about investing in AI and opening its doors to Palantir business. The proclamations are always the same: to win a war, you need AI. His company’s market cap has since doubled from just over 30 billion USD in 2023, to just over $80 billion USD at the time of writing and Palantir has just been included in the S&P500. 

The wars on Ukraine and Gaza rage on, with mounting death tolls and no end in sight.