April 14, 2026
The New Arsenal
Tracking the technologies, infrastructure, and industrial scale shaping modern warfare
"The power of the military is no longer gauged by the number of soldiers or the weapons they have. It is now gauged by the ability of a country to process information."— The New Arsenal · Issue 16 · April 2026
In This Issue
- The Number That Changes Everything — Maven, 1,000 targets per hour
- The Budget Is the Thesis — $54.6B for autonomous warfare, $1.5T total
- Maven and the First AI War — Operation Epic Fury as live proof of concept
- The Cost Asymmetry Problem — And the only structural answers
- Five Things Happening Right Now on the Distributed Battlefield
- The Numbers That Matter This Week
- The Original Insight — What nobody is writing about
- What To Watch
There is a number that changes everything, and almost no one is writing about it correctly.
On February 28, 2026, the first day of Operation Epic Fury against Iran, the Maven Smart System processed satellite imagery, drone feeds, signals intelligence, and 150+ additional data sources and presented US military commanders with more than 1,000 strike options. What once required days of human analysis was completed in minutes. By the end of the first 24 hours, CENTCOM had executed approximately 900 strikes across Iran — nearly double the scale of the 2003 shock-and-awe campaign in Iraq.
That is the headline number. But the number that actually matters is this one:
The next-generation Maven is being built to process 1,000 targets per hour.
Not per day. Per hour. That is not an upgrade. That is a different kind of warfare — a different relationship between human decision-making and machine execution. A different answer to the question of what it means to be at war.
We are now living inside the thing The New Arsenal has been tracking since Issue 1. The distributed battlefield. The AI kill chain. The compression of decision cycles from days to hours to minutes. All of it went live this spring in the skies over Iran. And this week, the US government told us what it actually thinks about all of this. The FY2027 defense budget dropped. What it revealed is staggering.
Section One
The Budget Is the Thesis
The FY2027 defense budget request — $1.5 trillion in total, the largest in American history, exceeding the Reagan-era buildup in inflation-adjusted terms — is not primarily a defense document. It is a thesis statement about the future of warfare.
The Defense Autonomous Warfare Group received $225 million in FY2026. Its FY2027 budget request: $54.6 billion. A 24,166% increase in a single year. The proposed DAWG budget exceeds the entire GDP of Zimbabwe. It is 15% of the entire $350 billion reconciliation package. Almost no one had heard of the DAWG before this week.
The spending plan signals the transformation of the DAWG into a joint Autonomous Warfare Command — the American equivalent of Ukraine's Unmanned Systems Force and Russia's Rubicon complex. The template exists. The US is now funding it at civilizational scale.
Beyond DAWG, the budget structure reveals the full thesis in budget line items:
The overall theme, as one Pentagon official described it: "Shifting from a sustainment-based force to a force that is investing in productive expansion of the industrial base and delivery of new innovation." Translation: the US government has concluded that it has been maintaining legacy systems while its adversaries built the next generation. This budget is the course correction.
Section Two
Maven and the First AI War
The Iran war is the live test of everything the defense technology ecosystem has been building for a decade.
Since February 28, US and Israeli forces have executed more than 11,000 strikes against Iranian targets. At the center of this operation is the Maven Smart System — Palantir's AI platform that one analyst described as "Google Earth for war, a map with white dots, infused with information about elevation, coordinate, what is precisely there, whether it's friendly or foe." Maven fuses satellite imagery, drone feeds, radar data, signals intelligence, and classified data to classify targets, recommend weapons, and generate strike packages in near real time.
Before AI integration, human analysts could process fewer than 100 targets per day. With Maven, that number grew to 1,000 targets per day. With the integration of large language models, including Anthropic's Claude embedded within the Palantir platform to summarize intelligence and simulate strike scenarios, that number is projected to reach 1,000 targets per hour in the next iteration.
Iran responded by striking AWS data centers in the UAE and Bahrain — a direct attack on the commercial cloud infrastructure supporting US AI systems. This is the first documented targeting of commercial AI infrastructure as a military objective. It will not be the last. Every nation that watched this war is now modeling US dependency on commercial cloud compute.
The Anthropic dimension requires careful tracking. On March 4, the Trump administration designated Anthropic as a supply chain risk after the company refused to allow its technology to be used for fully autonomous weapons and mass domestic surveillance. Claude is being phased out of classified military systems within six months. What the episode reveals is that the boundary between commercial AI development and military AI deployment has effectively collapsed — and the governance frameworks for managing that collapse do not yet exist.
Section Three
The Cost Asymmetry Problem — And the Only Structural Answers
A Shahed-136 drone costs between $20,000 and $50,000 to manufacture. A Patriot interceptor missile costs more than $4 million. Using a Patriot to shoot down a $35,000 Shahed costs the United States 100 times what it costs Iran to launch it. US and allied interceptor inventories are being drawn down at a rate that sustainment pipelines cannot match. The Pentagon has acknowledged this explicitly.
The math the Pentagon is trying to solve: how do you sustain a war against an adversary with 100:1 cost advantage in saturation weapons? There are only three structural answers.
Reduce the cost of defense to approach cost parity
Iron Beam does this at $3 per shot versus tens of thousands for kinetics. The $2 billion directed energy allocation in FY2027 is the explicit investment in scaling this answer. Israel fielded the world's first operational directed-energy air defense system in December 2025. The race now is to the next range threshold.
Attack the adversary's production capacity before weapons are launched
Ukraine proved this works. Striking 6,000 FPV drones in a single container facility eliminates weeks of Russian offensive capacity in one operation. The industrial targeting doctrine — attacking the supply chain, not the weapon — is the most cost-efficient answer to the saturation problem. It requires deep-strike capability and persistent ISR to execute.
Build autonomous offensive systems cheap enough to approach cost parity
This is what the DAWG's $54.6 billion is funded to do. Mass-produce attritable autonomous systems at a cost that makes the offense-defense ratio sustainable. Ukraine building 5 million drones per year from zero in four years is the model. The US is building the institutional infrastructure to replicate that at American industrial scale.
All three answers are now funded. All three are in active development. The race is which of them scales to operational relevance before the next peer conflict begins.
Section Four
The Distributed Battlefield: Five Things Happening Right Now
Ukraine launched more attack drones than Russia in March 2026 — for the first time in the war. Production has grown from 800,000 units in 2023 to approximately 5 million projected for 2026. Ukraine is out-producing Russia in the category Russia invented. A nation a fraction of Russia's size, under live bombardment, built a defense industrial ecosystem from nothing in four years and is now winning the production race.
The "high-low mix" doctrine is now embedded in NATO procurement planning. High-end platforms like Storm Shadow saw dramatically increased effectiveness in Ukraine only when complemented by cheap drone swarms and electronic warfare that overwhelmed defenses first. The exquisite weapon requires a low-cost wrapper to survive. This is no longer theoretical — it is being written into program requirements.
Autonomous orchestration is now a $100M+ procurement priority. DIU's Autonomous Orchestrator Challenge, Breaker's voice-commanded fleet control, Scout AI's Fury foundation model, Edgerunner's WarClaw — the architectural race for how one operator commands hundreds of autonomous systems is no longer a research program. It is an active contract competition with a funded requirement.
Iran struck commercial AI infrastructure as a military target. AWS data centers in the UAE and Bahrain were targeted specifically to impair US AI targeting capability. The $9 billion requested in FY2027 specifically for AI-dedicated military data centers is the direct institutional response to this vulnerability being exposed in combat.
The software margin thesis is being validated at scale. Lockheed Martin earns 8–10% gross margins. Palantir's Maven platform earns software margins. Anduril's Lattice runs at 40–50%. The companies setting the AI kill chain architecture are not competing in the hardware procurement cycle — they are defining the protocol that hardware operates within. At $1.5 trillion in annual defense spending, 1% of software margin advantage compounds into generational wealth.
The Numbers That Matter This Week
◆ The Insight Nobody Is Writing About
The Iran war has not validated the AI kill chain. It has validated the first generation of the AI kill chain. And the first generation is already being superseded.
Maven can process 1,000 targets per day. The next version targets 1,000 per hour. The version after that — if the trajectory holds — will operate at speeds where human authorization becomes a formality rather than a meaningful control. The question of what happens to the laws of war, to the doctrine of distinction, to the entire framework built around human judgment at the moment of lethal decision — that question is not hypothetical anymore.
The DoD requested $580 million for the counter-drone governance task force. It requested $54.6 billion for autonomous offensive warfare. The ratio of investment between the governance problem and the capability problem is 100:1 — the same asymmetric ratio that makes the Shahed so effective against the Patriot.
The defense investment thesis of the next decade is real. The urgency is real. The strategic necessity is real. And the fastest-moving technology in human history is being deployed in the domain where the consequences of failure are the most permanent.
◆ What To Watch
The "skinny" FY2027 budget shows toplines. Detailed justification documents publish April 21. The DAWG's $54.6B will either come with a program breakdown or remain classified. Which outcome tells you more than the number itself.
The progression from 1,000/day to 1,000/hour is the most consequential technology development in current defense. Watch for DoD announcements about program of record structure and AI governance frameworks being built around it.
VP Vance confirmed the US has achieved stated objectives and may begin scaling back. The ceasefire negotiation determines whether this conflict becomes a controlled precedent or an uncontrolled escalation. Either outcome reshapes threat modeling for every allied nation.
The creation of a joint Autonomous Warfare Command — the logical successor to the DAWG at $54.6B scale — would be the most significant military organizational development since the creation of US Cyber Command. Watch for the announcement.
Claude is being phased out of classified systems within 6 months. The replacement has not been publicly announced. Whatever fills that gap becomes the foundational LLM for US military AI targeting. That is not a small contract.
The budget is the clearest signal a government can send about what it believes is true.
The FY2027 budget says the US government believes autonomous warfare is the decisive domain of the next conflict. It says the cost asymmetry problem is existential, not marginal. It says the AI kill chain is not a future capability — it is a present operational reality that must be funded at transformational scale immediately.
The answer to that question is still being written.
Until next week.
◆ Stay Briefed
The New Arsenal
The definitive newsletter tracking the defense technology, modern warfare, and the industrial scale shaping national security. Published by Emeka Alozie.