Software Is Becoming a Weapon
Wars are now won and lost in update cycles. The Pentagon spent decades buying software like it buys ships — and that era is ending in real time.
Bottom Line Up Front
The decisive weapons of the last few years haven't just been drones and missiles. They've been the invisible things: the targeting software that finds targets in minutes instead of days, the jamming that blinds a weapon mid-flight, the code pushed to the front line before the enemy adapts.
Militaries used to upgrade weapons in decades. In Ukraine, effective lifetimes of some systems are measured in weeks, because electronic warfare — the fight over the radio spectrum — keeps breaking them.
The Pentagon has historically bought software the way it buys aircraft carriers: decade-long requirements, rigid milestones, delivery of a finished product. That process reliably produces obsolete software.
This is changing, visibly. Palantir's Maven targeting system went from experiment to a contract ceiling near $1.3 billion, and in 2026 the Pentagon moved to make it a formal program of record — the anchor of its joint command-and-control push. Palantir and Anduril's software alliance has become the de facto template.
The prize is enormous: whoever owns the military's software layer owns the highest-margin, stickiest position in the largest defense budgets since World War II.
The missile that stopped working
Here's a pattern that repeated throughout the war in Ukraine: a Western precision weapon arrives and performs brilliantly. For a few months it's devastating. Then, gradually, it starts missing. Nothing changed in the factory. What changed was the air itself — Russian electronic warfare units learned the weapon's guidance signals and began jamming or spoofing them.
At that point the weapon's fate depends on something no spec sheet captures: how fast its maker can diagnose the interference, rewrite the guidance logic, and push the fix to the field. Weeks-long update loops kept some weapons relevant. Systems that needed a years-long formal upgrade cycle became expensive scrap.
That's what "software-defined warfare" actually means. Not that software replaces the missile — the missile still has to fly. It means the missile's usefulness now decays at software speed, and only software processes can renew it.
Why the Pentagon buys software badly
To understand the problem, understand what the acquisition system was built for. Buying a destroyer is a hardware problem: freeze the requirements, spend years building, deliver a finished thing that serves for thirty years. The Pentagon's rules are superb at preventing waste in that world.
Software inverts every assumption. It's never finished, and its requirements can't be frozen, because the enemy votes on them continuously. A software program "delivered" after a seven-year acquisition cycle isn't late — it's dead on arrival. For decades, the government wrote detailed specs, hired contractors to code to those specs, and received exquisitely compliant systems that users hated and adversaries had already outrun.
The fix has been slow but real: a dedicated software acquisition pathway created in 2020, commercial-first buying, and — most importantly — a willingness to purchase working products instead of commissioning bespoke ones.
Maven, Lattice, and the new template
The clearest evidence is Maven. Born as a Pentagon AI project to sort drone video, it evolved into Palantir's Maven Smart System — software that fuses satellite feeds, intelligence, and unit data into a live targeting picture. The contract ceiling jumped from roughly $480 million toward $1.3 billion; and in 2026 the Pentagon directed that it become a program of record, the cornerstone of the department's push toward joint, AI-assisted command and control — the military's operating system, in plain terms: the thing every sensor and shooter plugs into.
Anduril's Lattice attacks the same problem from the edge, directly controlling autonomous systems, and the two companies formally allied in late 2024. Together they've defined the template: commercial software companies, iterating continuously, owning the connective tissue of the force while hardware makers plug into it.
Trace the chain and this story connects to everything: military software runs on the same chips, data centers, and electricity straining under commercial AI demand, and its talent competes with Silicon Valley payrolls.
A cardinal doesn't inspect every seed in the forest; it recognizes the pattern of where food reliably appears and ignores the noise. Targeting software earns its budget the same way — not by seeing more, but by recognizing what matters faster than the other side does.
Key Judgments
- Software will take a steadily rising share of defense spending, and its margins will exceed hardware margins for the same strategic contribution.
- Program-of-record status for commercial software platforms is the regime change: it converts episodic contracts into durable, compounding franchises.
- Vendor concentration is the next fight — expect open-architecture mandates and funded challengers within two to three years as Congress reacts to Palantir-Anduril dominance.
- Electronic warfare adaptation speed will become a formally tracked military metric within a few years.
Risks & Counterarguments
The honest case against runs three ways. First, lock-in: making one company's platform the military's operating system creates exactly the single point of failure — commercial, political, and cyber — that defense acquisition rules were written to prevent. A serious breach or outage of a dominant platform would reverse this trend fast. Second, hype: much of what's claimed for AI-enabled targeting is unverifiable outside classified channels. Third, software doesn't hold ground, deter fleets, or replace magazines of munitions — a force with brilliant software and empty magazines still loses.
Why It Matters
This is the rare case where a structural shift in government spending, a genuine change in how wars work, and a set of investable business models all point the same direction. If software becomes the military's connective tissue, then defense stops being purely an industrial sector and becomes partly a technology sector, with that sector's economics and concentration risks. Anyone tracking defense budgets with a hardware-era map will misread where the value is accruing.
What We're Watching
- Whether Maven's program-of-record transition survives budget review intact, and what open-architecture conditions get attached.
- Software and EW line items in the comptroller's budget justification books — the cleanest public measure of the shift.
- The first major program where a prime loses integration leadership to a software company, or vice versa.
- Reported update-cycle times for weapons in contested EW environments — the metric that actually decides relevance.
Sources: Office of the Under Secretary of Defense (Comptroller) budget justification books; Congressional Research Service reporting on defense software acquisition; GAO reviews of the software acquisition pathway; DefenseScoop reporting on Maven, CJADC2, and the Palantir-Anduril alliance; company disclosures. This is analysis, not investment advice.