This is Part 3 of our AI Weapons series. Part 1: The Week AI Lost Its Conscience examined how ethics collapsed in a single week. Part 2: The $35,000 Question mapped the nine governance gaps. This part follows the money.
The Number That Changes Everything
On February 28, 2026, a $35,000 drone called LUCAS flew its first combat mission over Iran during Operation Epic Fury. Built by SpektreWorks, a small company in Phoenix, Arizona, the Low-cost Unmanned Combat Attack System was reverse-engineered from Iran's own Shahed-136 — the same drone Russia has been using against Ukraine.
The irony is surgical: America took Iran's design, improved it, and used it to strike Iranian targets.
But the real story isn't the irony. It's the number.
$35,000. That's what it costs to build one LUCAS unit. For comparison, a single MQ-9 Reaper — the drone that defined American aerial warfare for two decades — costs $30 million. That's an 857:1 cost ratio.
LUCAS went from first flight to combat deployment in 90 days. The F-35 program has been in development for over 20 years and has cost $1.7 trillion.
Something fundamental has shifted.
The Cost-Exchange Crisis
The Center for Strategic and International Studies calculated a number that should keep every defense minister awake at night: for every $1 spent on a drone attack, defenders lose approximately $30,000 in assets.
Let that sink in. A 1:30,000 cost-exchange ratio.
The numbers cascade from there:
| Attack | Defense | Ratio |
|---|---|---|
| $500 FPV drone | $82.5M F-35 fighter | 1:165,000 |
| $35K Shahed-136 | $3-4M Patriot interceptor | 1:86-114 |
| $38K drone | $3M SAM missile | 1:79 |
| $221K sea drone | $100M+ warship | 1:450+ |
In Ukraine, 70% of frontline casualties now come from UAVs. A $500 drone operated by a teenager with a VR headset can destroy a $2 million tank. A swarm of 343 drones targeted Moscow in a single operation.
This isn't a technology problem. It's a math problem. And the math is merciless.
When a $3 million Patriot missile intercepts a $38K drone, the attacker wins even when the attack fails. The defender has spent 79 times more money to neutralize a threat that cost almost nothing to produce. Scale this across hundreds of attacks per day, and air defense becomes economically unsustainable.
As RAND noted: modern warfare is no longer about technological superiority. It's about economic sustainability.
The Pentagon's Bet: 300,000 Drones
The Pentagon has responded with the most aggressive autonomous weapons procurement in history.
The FY2026 defense budget allocates $13.4 billion specifically to AI and autonomy — the largest single-year defense AI investment ever. For the first time, autonomy gets its own budget section. The Navy alone is spending $5.3 billion on autonomous systems, $2.2 billion more than the previous year.
But the headline number is the production target: the War Department has asked industry to produce more than 300,000 drones, quickly and cheaply.
The Drone Dominance program, launched in February 2026, pits 25 companies against each other in a competition to field 150,000 one-way attack drones at $2,300 per unit. That's not a typo. $2,300. Less than an iPhone.
The program structure is deliberately designed to bypass traditional defense procurement:
- Phase 1: 25 vendors, small production runs
- Phase 2: Down to 12 vendors, larger orders
- Phase 3: 5 vendors, 150,000 units, $2,300 target price
Software-first architecture. Rapid iteration. Performance in the air, not corporate overhead or lobbying power. This is Silicon Valley's playbook, applied to munitions.
The Replicator Hangover
This isn't the Pentagon's first attempt. The Replicator initiative, launched in 2023, promised "thousands" of autonomous systems by August 2025. It delivered "hundreds." The exact number is classified, but Congressional Research Service reports made clear: the program fell short.
Replicator has since been renamed the Defense Autonomous Warfare Group (DAWG) and is now focused on larger, longer-range systems. The Drone Dominance program represents a parallel track — one focused on mass, not sophistication.
The lesson: producing thousands of cheap drones isn't hard technologically. It's hard institutionally. The Pentagon's procurement system was built to buy 200 F-35s at $80 million each, not 200,000 drones at $2,300 each. Different supply chains. Different quality assurance. Different everything.
Anduril: The $60 Billion Insurgency
No company embodies this shift more than Anduril Industries.
Founded less than a decade ago, Anduril was valued at less than $2 billion in 2020. By February 2026, it's discussing funding at a $60 billion-plus valuation. That's 30x growth in six years. Revenue doubled from $1 billion in 2024 to roughly $2 billion in 2025.
For context: Anduril's $60 billion valuation puts it in the same territory as Northrop Grumman's market cap. A startup is now valued comparably to the company that builds B-2 stealth bombers.
The centerpiece of Anduril's strategy is Arsenal-1 — a 5-million-square-foot manufacturing facility in Ohio designed to produce tens of thousands of autonomous systems annually. That's a $1 billion investment creating 4,000+ jobs, with production beginning in July 2026.
And then there's Fury, Anduril's autonomous combat jet. Mach 0.95. 50,000-foot ceiling. Six-hour endurance. 700-mile range. Designed to fly alongside manned aircraft in combat.
All connected by Lattice, a software platform that integrates drones, sensors, and military assets into a unified AI-enabled network. It ingests data, identifies threats, and coordinates responses — in some cases autonomously.
Meanwhile, Shield AI has reached a $12 billion valuation for its autonomous aircraft technology. These aren't fringe startups. They're becoming the defense industry.
The traditional primes aren't dying — Lockheed Martin just reported $74.75 billion in annual revenue with a $179 billion backlog. RTX has a $251 billion backlog. But their growth comes from legacy platforms and missile production, not innovation. They're defending, not attacking.
Taiwan: The 50,000-Drone Hellscape
Nowhere is the economic logic of cheap autonomous weapons clearer than in the Taiwan Strait.
Taiwan cannot match China in conventional military platforms. The math doesn't work. So Taiwan has adopted what defense planners call the "porcupine strategy" — make invasion so costly that it becomes irrational.
The centerpiece: a two-year plan to procure 50,000 domestically built drones by 2027, across five categories, treated as consumable munitions. Like bullets. Use once, throw away.
Sea drones costing $221,000 can sink or damage warships worth hundreds of millions. Attack drones with autonomous terminal guidance can hit targets without human input, nullifying Chinese jamming attempts. The strategy is to exhaust China's interceptor missile stocks with mixed salvos of cheap drones and missiles, then attack the fleet with subsequent waves.
A recent report urged Taiwan to create a drone swarm "asymmetric hellscape" to blunt any Chinese invasion. The language is instructive. This isn't about winning a war. It's about making the economics of invasion untenable.
But implementation is lagging. Officials describe "an alarming lack of urgency." Taiwan's military establishment, like America's, was built for traditional platforms. Pivoting to mass-produced autonomous systems requires institutional transformation, not just procurement reform.
The Three Laws of Drone Economics
From the data, three principles emerge:
1. Attack is cheaper than defense — permanently.
This isn't a temporary advantage that better technology will close. A drone costs materials plus software. An interceptor costs materials plus software plus precision guidance plus launch system plus radar integration. The attacker's cost floor will always be lower than the defender's cost floor. The cost-exchange ratio may improve for defenders, but it will never reach parity.
2. Speed of iteration beats sophistication of design.
LUCAS went from concept to combat in 90 days. The Drone Dominance program selects winners by performance in competitive fly-offs, not by proposal quality. Anduril's software-first architecture allows updates in weeks, not years. In drone economics, the company that ships fastest wins — not the company that builds best.
3. Production volume is the new strategic advantage.
Ukraine's plan to produce 4.5 million drones in a year is the template. The U.S. procured 50,000 UAS in 2025 and plans 200,000 more by 2027. China's production capacity is classified but estimated to be massive. The strategic question is no longer "who has the best drone?" but "who can produce the most drones fastest?"
What This Means
The $35,000 LUCAS drone that flew over Iran on February 28 wasn't just a weapon. It was an economic argument.
The argument goes like this: the era of the $30 million drone, the $80 million fighter jet, and the $13 billion aircraft carrier is ending. Not because these platforms don't work, but because the economics no longer make sense when the other side can field 857 LUCAS units for the price of one Reaper.
Traditional defense contractors will survive — they have backlogs measured in hundreds of billions. But the growth, the innovation, and increasingly the strategic advantage will flow to companies that think in terms of $2,300 units produced at scale.
The Pentagon knows this. That's why it's asking for 300,000 drones. That's why autonomy has its own budget line for the first time. That's why Anduril is worth $60 billion.
In Part 1, we watched ethics collapse in a week. In Part 2, we mapped the governance vacuum. Here, we've followed the money.
The money says autonomous warfare isn't coming. It's here. And it's cheap enough that everyone can afford it.
That's the most dangerous part.
Top comments (0)