Code Metal Secures $125M to Transform Defense With AI
▼ Summary
– Code Metal, an AI startup for code generation and translation, raised $125 million in Series B funding, following a $36 million Series A round.
– The company focuses on translating code between languages and verifying it, with early customers in defense and aerospace like L3Harris and the US Air Force.
– It addresses the industry challenge of modernizing old, legacy code, such as porting C to Rust, which is seen as a major future need.
– The startup’s platform uses test harnesses during translation to evaluate code and claims its pipelines can prevent errors by halting if no solution exists.
– Code Metal is part of a broader trend of startups securing venture capital to provide tools for testing and securing AI-generated code.
A Boston-based startup specializing in AI-powered code translation and verification has secured a substantial $125 million in Series B funding, signaling strong investor confidence in tools designed to modernize legacy software systems, particularly within the defense sector. Code Metal focuses on converting code between programming languages and ensuring its reliability, a critical need for industries managing outdated but essential infrastructure. This latest financial injection follows a $36 million Series A round just months prior, highlighting the rapid momentum behind companies providing the foundational technologies for the AI-driven transformation of software development.
The company is part of a broader trend where ventures like Antithesis, Code Rabbit, and Harness are attracting significant venture capital. These firms sell the essential tools, the so-called picks and shovels, for an industry increasingly reliant on automated code generation. While the long-term efficacy of some methodologies remains unproven, the market potential is driving considerable investment, with the bet that several of these approaches will yield substantial returns.
Founded in 2023, Code Metal has quickly carved out a niche by targeting the defense industry’s unique challenges. Its early client list includes major players like L3Harris, RTX, and the US Air Force. The startup is also collaborating with Toshiba and is in discussions with a prominent chip manufacturer to tackle code portability across different hardware platforms, though the specific company was not disclosed.
The core of Code Metal’s platform is its ability to translate code from higher-level languages such as Python, C++, and Matlab into lower-level or hardware-specific languages like Rust, VHDL, or Nvidia’s CUDA. This addresses a pervasive issue: vast amounts of critical software are written in obsolete languages, creating bottlenecks and security vulnerabilities. CEO Peter Morales, a veteran of Microsoft and MIT Lincoln Laboratory, points to the “big tentpole problems” of an industry on the cusp of an AI revolution. One major hurdle is the slow, costly process of porting old code for new applications, especially when expertise in legacy systems is scarce.
Morales references a recent observation by AI researcher Andrej Karpathy about the growing push to port C code to Rust, suggesting a future where massive software rewrites become commonplace. “That is all of what we do in one tweet,” Morales remarked, underscoring the company’s mission.
Investors echo this sentiment regarding the urgent need for modernization. Yan-David Erlich, a general partner at B Capital, notes that the code controlling vital communications and satellite systems is often antiquated and written in deprecated languages. However, he highlights the inherent risk: “But in the course of translation, you might be inserting bugs, which is catastrophically problematic.”
Code Metal asserts its proprietary technology mitigates this danger. Morales explains that their software generates a series of test harnesses at each translation step, creating a virtual environment to evaluate the code and provide continuous verification to the customer. When questioned about error rates, Morales states it varies with the complexity of the conversion. For their current operational pipelines, he claims the system is designed to avoid generating erroneous code altogether; if a translation cannot be completed correctly, the software simply reports that no solution is available.
While the startup remains guarded about the intricate details of its technical methodology, it is openly discussing its business model, particularly its pricing strategy. This transparency on commercial terms contrasts with its protective stance on proprietary algorithms, as it seeks to scale its operations following this significant capital raise.
(Source: Wired)