Let’s talk about a problem that is simultaneously incredibly boring and existentially terrifying: old software. Not just the outdated operating system on your grandad’s laptop, but the ancient, creaking code that runs some of the world’s most critical defence systems. For decades, the Pentagon and its contractors have been wrestling with this beast, a giant hairball of legacy code that’s expensive to maintain, nearly impossible to update, and a massive security risk. So, when a startup lands a cool $125 million to tackle this specific headache, you have to sit up and pay attention.
Enter Code Metal, a Boston-based startup that just hit a $1.25 billion valuation. Their mission? To be the digital Rosetta Stone for the military-industrial complex. This isn’t just another shiny app; it’s a fundamental piece of the puzzle in the grand, necessary push for defense software modernization.
The Multi-Billion Pound Proving Ground
What exactly is defense software modernization? At its core, it’s about dragging critical national security systems into the 21st century. Many of these systems run on programming languages that were popular when the internet was still a niche academic project. This isn’t just an inconvenience; it’s a strategic vulnerability. You can’t just plug modern AI capabilities or advanced cybersecurity protocols into a system built on digital foundations from the Cold War.
This is where legacy system translation comes in. Think of it like swapping out the engine of a classic car while it’s speeding down the motorway. You need to replace the old, inefficient parts with a modern, high-performance engine without crashing. Code Metal’s platform is designed to do just that, automatically translating high-level code like Python or C++ into the specialised, lower-level languages like Rust and VHDL that interact directly with military hardware. For clients like L3Harris, RTX, and the US Air Force, this is a game-changer.
Can We Trust the AI Butler to Write the Code?
Now for the critical question: if humans writing this code created bugs, what happens when an AI does it? This is where the conversation turns to military AI, and not in the way you might imagine. This isn’t about autonomous drones, but about using AI for the painstaking, detail-obsessed work of writing and, more importantly, verifying code.
The fear of AI-generated errors is very real. As Code Metal’s Yan-David Erlich told WIRED, “in the course of translation, you might be inserting bugs—which is catastrophically problematic.” A misplaced semicolon in a banking app is an annoyance; in a missile guidance system, it’s an international incident.
This is why secure code verification is the other half of Code Metal’s value proposition. Their system doesn’t just translate; it rigorously tests and validates the new code at every step. According to the company, their process is designed to be foolproof. If a perfect, bug-free translation isn’t possible, “The software will just say, ‘There’s no solution for this’.” For the people in charge of billion-pound defence projects, that kind of certainty is priceless.
The New Breed of Defence Tech Startups
Code Metal is part of a growing wave of defense tech startups attracting serious capital from Silicon Valley heavyweights. The $125 million Series B round was led by Salesforce Ventures, with previous investors like Accel and B Capital chipping in. This follows a $36 million Series A, signaling immense confidence from investors who typically chase consumer apps or enterprise SaaS.
What’s changed? Two things. First, the geopolitical climate has made national security an undeniable priority. And second, the technology, particularly in AI, has finally matured to a point where it can solve these deep, intractable problems that have plagued the defence sector for years. These startups aren’t just building better weapons; they’re refactoring the very infrastructure that supports them.
A Radically Simple Pricing Model
Perhaps one of the most interesting things about Code Metal is its business model. In a world of per-seat licenses and complex usage metrics, their approach is refreshingly direct: they price their service based on the development time they save their clients.
This is a clever strategic move. They are not selling software; they are selling a guaranteed outcome—faster, cheaper, and more reliable modernisation. By aligning their revenue directly with their customer’s cost savings, they remove the friction from the sales process. It becomes a simple calculation of return on investment for massive defence contractors, making it an easy “yes.”
The Inevitable Rewrite of Everything
So, what does this all mean for the future? Former Tesla AI chief Andrej Karpathy once tweeted, “It feels likely that we’ll end up rewriting large fractions of all software ever written many times over.” Code Metal’s CEO, Peter Morales, saw that and essentially said, “That is all of what we do in one tweet.”
He’s not wrong. The work being done by Code Metal is a precursor to a much larger trend. As AI gets more capable, this process of translation, verification, and modernisation will accelerate, not just in defence but across finance, healthcare, and energy—any industry built on a foundation of aging code. We are at the very beginning of a great software migration, moving decades of digital legacy onto modern, more secure, and AI-ready platforms.
The journey of defense software modernization is long and fraught with challenges. But with highly-capitalised startups like Code Metal bringing a new approach to the problem, the path forward is becoming clearer. It’s less about revolutionary new weapons and more about the quiet, crucial revolution happening in the code itself.
What do you think? Is AI-driven code translation the key to unlocking the next generation of defence technology, or are we placing too much trust in algorithms for our most critical systems?
