A startup that most people outside Silicon Valley have never heard of just raised $100 million to bet on an idea that would have seemed absurd two years ago: that the future of software development belongs not to elite engineers, but to anyone with an idea and the ability to describe it in plain English.
Code Metal, a San Jose-based company founded in 2016, closed its Series B round at a valuation that signals deep investor confidence in the emerging practice known as “vibe coding” — a term coined by former OpenAI co-founder Andrej Karpathy in February 2025 to describe the act of building software by conversing with AI rather than writing code by hand. The round was led by Maverick Silicon, with participation from Playground Global and others, according to a report by Wired.
From GPU Compilers to the Vibe Coding Gold Rush
Code Metal’s origins have little to do with the AI-assisted coding movement it now champions. The company was originally focused on building compiler technology for GPUs and specialized hardware accelerators — the kind of deep systems work that attracts PhDs, not product managers with a ChatGPT subscription. But the company pivoted sharply as large language models began demonstrating an unexpected talent for generating functional code from natural language prompts.
The pivot appears well-timed. According to Wired, Code Metal now offers a platform that allows users to build, test, and deploy applications using AI agents that handle the actual programming. The company’s thesis is straightforward: as AI models get better at writing code, the bottleneck in software creation shifts from technical skill to imagination and domain expertise. Code Metal wants to be the infrastructure layer that makes this shift practical and scalable for enterprises.
What Vibe Coding Actually Means — and Why It Matters Now
The term “vibe coding” has become something of a lightning rod in the tech industry. Karpathy described it as a process where you “fully give in to the vibes, embrace exponentials, and forget that the code even exists.” In practice, it means using tools like Cursor, Replit, or Claude to generate entire applications from conversational prompts, often without the user reading or understanding the underlying code. The programmer’s role shifts from writing syntax to directing intent.
Critics have been vocal. Security researchers have warned that AI-generated code frequently contains vulnerabilities, and experienced developers have pointed out that vibe-coded applications often break in unpredictable ways when they encounter edge cases. A March 2025 study from Stanford’s computer science department found that code produced by LLMs contained security flaws at a higher rate than code written by junior developers, raising questions about whether vibe coding is ready for production environments. Yet the momentum behind the movement is undeniable. GitHub reported earlier this year that more than 40% of code on its platform is now AI-generated, a figure that has been climbing steadily since the launch of Copilot in 2022.
The Investor Thesis: Software Creation as a Mass-Market Activity
The $100 million raise reflects a broader investor conviction that software development is about to become radically more accessible. Maverick Silicon, which led the round, has been making concentrated bets on AI infrastructure companies. The firm’s partners have argued publicly that the total addressable market for software creation tools expands by orders of magnitude if non-programmers can build functional applications.
Code Metal CEO and co-founder Walid Najjar told Wired that the company’s platform is designed to handle the complexity that individual AI coding assistants cannot. Rather than simply generating code snippets, Code Metal’s system orchestrates multiple AI agents that handle different aspects of the development process — architecture, implementation, testing, and deployment. The goal is to produce software that is not just functional but maintainable and secure, addressing the primary criticisms leveled at vibe coding.
A Crowded Field With High Stakes
Code Metal is far from alone in chasing this opportunity. The AI coding tools market has become one of the most competitive sectors in technology. Cursor, developed by Anysphere, raised $900 million earlier this year. Replit has positioned itself as the go-to platform for building applications with AI assistance. Cognition’s Devin, marketed as the first AI software engineer, attracted significant attention — and skepticism — when it launched. Poolside, another AI coding startup, raised $500 million in 2024.
What distinguishes Code Metal, according to the company, is its focus on the full application lifecycle rather than just code generation. Many existing tools excel at producing individual functions or components but struggle with the architectural decisions, dependency management, and integration testing that define real-world software projects. Code Metal’s compiler heritage gives it a technical foundation in understanding how code executes on hardware, which the company argues translates into better optimization and fewer runtime errors in AI-generated applications.
The Enterprise Angle: Where the Real Money Lives
While consumer-facing vibe coding tools have grabbed headlines — stories of teenagers building apps in hours, or retirees creating custom software for their hobbies — the serious revenue opportunity lies in enterprise adoption. Large companies spend billions annually on custom software development, and many have internal backlogs of requested applications that their IT departments cannot build fast enough.
Code Metal is targeting precisely this pain point. The company’s enterprise platform is designed to let business analysts, product managers, and domain experts build internal tools and applications without waiting in the IT queue. According to Wired, several Fortune 500 companies are already piloting the platform, though Code Metal has not disclosed specific customer names. The enterprise focus also helps address security concerns: Code Metal’s platform includes built-in code review agents and security scanning that aim to catch the vulnerabilities that plague simpler vibe coding approaches.
The Talent Paradox: Will Vibe Coding Kill Programming Jobs?
Perhaps no question in the current AI discourse generates more heat than whether tools like Code Metal’s will eliminate programming jobs. The data so far presents a mixed picture. While some companies have reported reducing their engineering headcount after adopting AI coding tools, others have found that AI assistance simply allows their existing engineers to be more productive, taking on projects that previously would not have been economically viable.
Najjar has taken a diplomatic stance on this question, telling reporters that Code Metal’s tools are designed to “augment human capability, not replace it.” But the company’s own marketing materials tell a more aggressive story, highlighting cases where a single product manager accomplished in days what previously required a team of developers working for weeks. The tension between these two narratives — AI as assistant versus AI as replacement — is one that every company in this space must manage carefully, particularly as the technology improves.
Technical Challenges That Remain Unsolved
For all the enthusiasm surrounding vibe coding, significant technical hurdles persist. Current AI models struggle with long-context reasoning, meaning that as applications grow in complexity, the quality of AI-generated code tends to degrade. Debugging remains particularly challenging: when an AI writes code that a human never reads, identifying and fixing problems becomes a novel kind of engineering challenge in itself.
There is also the question of intellectual property. Code generated by AI models trained on open-source repositories exists in a legal gray area that courts have not yet fully resolved. Companies deploying vibe-coded applications in production face potential exposure to copyright claims, a risk that enterprise legal departments are only beginning to assess. Code Metal has said its platform includes provenance tracking for generated code, but the broader legal questions remain open.
What the $100 Million Signals About the Market’s Direction
The size of Code Metal’s raise — $100 million for a company that has pivoted from its original business model — says something significant about where venture capital sees the software industry heading. Investors are not simply betting that AI coding tools will be useful; they are betting that the fundamental economics of software creation are about to change. If building an application becomes ten times cheaper and faster, the number of applications that get built could increase by a factor of a hundred.
Whether Code Metal specifically will be the company that captures this value remains an open question. The AI industry is littered with well-funded startups that failed to find sustainable business models despite strong technology. But the underlying trend — software creation becoming accessible to a vastly larger population of builders — appears durable regardless of which companies ultimately win. For the software industry’s incumbents, from Salesforce to ServiceNow, the rise of vibe coding represents both a threat and an opportunity that demands immediate strategic attention. The $100 million flowing into Code Metal is just one data point in what is shaping up to be one of the most consequential shifts in how the world builds technology.