The sector which shall not be named

With all eyes on Yann LeCun, one of the most celebrated AI scientists of the last 20 years, after his unceremonious exit from Meta, it’s a bit disappointing that one of his first public posts covers "AI-powered weapons to defend liberal democracies." It comes in the backdrop of defense-tech funding jumping from less than 1% of VC funding in 2020 to a solid 10% of funding today, in a trend which I find deeply disturbing.

To understand why this enthusiasm from one of AI's foremost practitioners and theorists is particularly disturbing, we will answer 3 questions today:

Is Defense-tech actually used "preventive logistics / readiness" or in "active offense"?

To answer this, lets look at the 4 biggest defense tech success stories of the last decade and how they actually generate revenue:

• Palantir: $2.9 B revenue @ ~$60 B valuation, with ~65% from Gotham & AI Platform (active ISR, targeting, command in battle zones) vs ~35% from Foundry & Apollo (preventive logistics and readiness).

• Anduril: $1 B revenue @ $30.5 B valuation, with ~70% from autonomous drones and C-UAS systems actively deployed in combat zones vs ~30% from AI software and maritime/space programs focused on readiness and control.

• Helsing: €200–400 M revenue @ ~$12 B valuation, with ~80% from combat drones and AI EW systems in active battlefields vs ~20% from autonomy and air-combat readiness projects.

• Shield AI: Undisclosed revenue @ $5.3 B valuation, with ~75% from Hivemind autonomy software and aircraft integrations used in contested operations vs ~25% from ISR and training applications supporting preventive defense.

Similarly, lets try understand the most promising investment themes from VC theses:

a16z, How the US can rewire the pentagon for a new era :

• Lux Capital, Q2-2024 report :

• Bessemer Venture Partners, Roadmap: Defense tech

Take-away: Defense-tech, both from a current revenue split and investment theme perspective, is thought of as the engine of "modern warfare," with a focus as much, if not more, on proactive offense as on preventive defense.  

Even if defense tech is used in active offense, is it used for defense of liberal democracies fighting against oppressive regimes?

To understand this lets first look at where majority conflicts have taken place in the last 50 years:

• Africa has hosted the highest number of state-based conflicts in recent years e.g., 2023: Africa 28, Asia 17, Middle East 10, Europe 3, Americas 1 (PRIO’s Conflict Trends: A Global Overview, 1946–2023)

• “Electing to Fight: Why Emerging Democracies Go to War” published by Belfer Center for Science and International Affairs, HKS finds that emerging democracies (i.e., states in transition) with weak institutions are more likely to initiate or be involved in war

Now lets look at who procures majority of weapons and defense systems:

• 2020–24 top importers (share of global imports = 35%): Ukraine, India, Qatar, Saudi Arabia, Pakistan. Regional shares of all imports: Asia-Oceania 33%, Europe 28%, Middle East 27%, Americas 6.2%, Africa 4.5% (SIPRI, trends in international Arms transfers, 2024)

• Although there are checks and export controls in place there have been instances re-export / diversion to conflict prone areas eg. U.S.-made Javelin missiles discovered in a pro-Haftar depot near Tripoli (NYT, 2019)

Takeaway: Defense tech, particularly "budget-friendly" weapons and systems, are frequently acquired and used in conflicts across Africa and nations with fragile democracies, often skirting export regulations.

Even if it is used indiscriminately, does it spur innovation and long term wealth generation?

This, I believe, is one of the strongest arguments in favor of defense tech spending, given that everything from semiconductors to the internet owes its development to the fabled defense industrial complex.

To understand this lets try to understand the extent of R&D spillover effect:

• Only 20–30% of defense-funded patents historically show spillovers into civilian technology sectors. In most peacetime buildups (e.g., post-1970s U.S., 1980s Soviet Union), high defense budgets coincided with slower total-factor-productivity growth despite massive outlays. (NBER Working Paper 26483, Defense R&D, Innovation and the Economy, 2020)

• A London Business School study (2023) found that only R&D-weighted defense spending yields productivity gains; procurement-heavy spending often crowds out private innovation and investment

Now lets weigh these benefits against potential risks:

• Multiple papers published in World Bank, Journal of Peace Research show that exogenous increase in one country’s military spending is more than doubled once you account for the reaction in that country and its neighbor

• Global military expenditure hit $2.4 trillion in 2023, up 7% year-on-year, the largest real-term jump in a decade (SIPRI 2024). The UN Development Programme cautions that such surges “fuel mistrust and heighten risk of confrontation,” especially absent corresponding diplomacy or conflict-prevention mechanisms (UNDP Report, 2024)

Take away: While defense spending does drive genuine innovation diffusion, the actual impact is far less significant than anticipated. This shortfall carries a serious risk of escalating mistrust and igniting an arms race.

Conclusion

In the presence of ample funding and the smartest brains gravitating to sectors like these, it's only a matter of time before they hit "success." Historically, these returns have come at the expense of stability in Africa and emerging economies. As an investor and a person who belongs to these geographies, this is incredibly distressing, especially given that there are so many ways AI can be used to bring people together or improve their lives. If we choose to prioritize these sectors over advancements in sectors like healthcare, we only have ourselves to blame if the world 10 years from now is more divided and conflicted.