AI for Automation
Back to AI News
2026-03-21AI codingdefense techCode Metalstartup fundinglegacy code

This AI rewrites military code — now worth $1.25B

Code Metal raised $125M to translate decades-old military software into modern languages using AI that mathematically proves every line is correct. The U.S. Air Force and Raytheon are already customers.


The U.S. military runs critical systems on software written in programming languages from the 1960s and 70s. Updating that code manually takes months and costs millions — and one mistake could be catastrophic. Code Metal just raised $125 million to let AI do it instead, with mathematical proof that every translated line works identically to the original.

Code Metal cofounders Peter Morales and Alex Showalter-Bucher

Translating code the military still depends on

Imagine trying to run a modern app on a computer from 1985. That's essentially the problem Code Metal solves — but for fighter jets, missile systems, and satellite networks. Many of these systems still run code written in COBOL, Ada, and Fortran — languages most programmers today have never touched.

Code Metal's AI translates this legacy code into modern languages like Rust and C++, then uses a technique called formal verification (a mathematical proof that the new code behaves exactly like the old code, line by line) to guarantee nothing breaks. As CEO Peter Morales put it: "Speed without proof does not work in regulated fields."

The $125M Series B at a glance:
Valuation: $1.25 billion — making Code Metal a unicorn
Lead investor: Salesforce Ventures
Other backers: Accel, B Capital, RTX (Raytheon), Shield Capital, J2 Ventures
Revenue: Already profitable with eight-figure annual revenue
Customers: U.S. Air Force, RTX (Raytheon), L3Harris, Toshiba, Boeing

How the AI actually works — in plain English

Code Metal uses what's called a neuro-symbolic approach — a combination of AI language models (the kind that power ChatGPT) and traditional mathematical logic. Here's the four-step process:

1. Divide & Conquer — The AI analyzes the entire codebase and breaks it into smaller, verifiable chunks

2. Instrument & Test — It automatically writes tests for every piece of code to establish a baseline

3. Translate & Verify — The AI translates old code to modern languages and mathematically proves the translation is correct

4. Optimize & Benchmark — The new code is tuned for the specific hardware it will run on

The key difference from tools like GitHub Copilot or ChatGPT: those tools generate code that might work. Code Metal generates code that is mathematically proven to work — a requirement when the software controls weapons systems or aircraft.

AI-powered software ecosystem diagram showing code translation workflow

From MIT Lincoln Lab to billion-dollar startup

The founders aren't Silicon Valley outsiders guessing at defense needs. CEO Peter Morales led AI research at MIT Lincoln Laboratory — the U.S. government's go-to research center for national security technology. Co-founder Alex Showalter-Bucher spent a decade at the same lab, working directly with the Navy, Army, and Department of Homeland Security.

The company's new President and COO, Ryan Aytay, previously ran Tableau (the data visualization tool Salesforce bought for $15.7 billion). His hiring signals Code Metal is ready to scale beyond defense into automotive, semiconductor, and other industries where software reliability is life-or-death.

Why this matters beyond the military

Code Metal isn't just a defense story. The same problem — ancient code running critical systems — exists across industries:

Banks still run core systems on COBOL — the language behind most ATM transactions
Automakers need to modernize embedded software for electric and autonomous vehicles
Chipmakers need code optimized for new processor architectures
Healthcare systems run on legacy software that's expensive and risky to update

Code Metal claims it reduces deployment time from months to days and cuts manual coding effort by at least 80%. For defense contractor L3Harris, a code translation that previously took weeks now finishes in days.

As Salesforce Ventures partner Rob Keith put it: "Mission-critical industries cannot deploy what they cannot verify. Code Metal's approach solves this by mathematically proving code is correct."

With the U.S. government increasingly pushing to modernize its technology infrastructure — and AI code generation tools raising trust concerns everywhere — Code Metal sits at the intersection of two massive trends. The $1.25 billion valuation suggests investors agree.

Related ContentGet Started with Easy Claude Code | Free Learning Guides | More AI News

Stay updated on AI news

Simple explanations of the latest AI developments