Today we are excited to announce CapitalG’s investment in Magic. Magic is developing an “AI colleague” for software engineering to transform the way all developers write code. We are thrilled to support Magic and partner with co-founder and CEO Eric Steinberger, his co-founder and CTO Sebastian De Ro, and their extraordinary team.

AI Unlocked

2022 was a watershed year for AI: Decades of hardware improvement, research, data collection, open source collaboration, and invested capital coalesced into breakthroughs that sparked the imaginations of not just technologists, but of the broader public as well. The future of software is being reimagined before our eyes.

Large language models (LLMs), in particular, made exciting leaps in performance in 2022. Tools ranging from chat bots to summarization engines seemingly demonstrated AI's ability to truly understand human language and generate compelling content. Though it is early innings today, it feels as though every interaction with software has the potential to be re-considered as an automatable task that can be abstracted away and unlocked using natural language commands.

At the heart of AI’s ongoing momentum has been the consistent drumbeat of contributions from some of the brightest minds at Google. From the early days of embedding ML in search, through to the invention of the transformer architecture and the scaling of production LLMs, Google has consistently proven to be one of the epicenters for AI ingenuity and progress. By partnering closely with incredible visionaries at Google like Jeff Dean, Eli Collins, Andrew Moore, and others, we’ve seen first-hand where AI has been and where it’s going, and we’ve developed deep conviction in AI as a transformational force that will pave the path for the creation of exciting new generational companies.

Software Engineering’s Talent Crunch

Over the last few years, the high cost (on average about $150K+ per year) and scarcity (IDC predicts a shortfall of 4 million engineers by 2025) of software engineering talent have fueled the need to optimize developer productivity. Individual developers and CTOs alike have evaluated every possible tool and avenue to eke out developer productivity. Thus far, these tools have been largely ineffective: Teams spend upwards of 25% of their time integrating and maintaining their toolchains. As a result, organizations find themselves right where they started: looking to hire as many developers as they can afford. In fact, despite an uncertain economic outlook, CoderPad’s recent annual Tech Hiring Survey revealed that over 80% of technical recruiters have developer hiring budgets at the same or higher levels than last year.

It will come as no surprise then that one of the most promising domains in which to apply LLMs is software engineering. Structurally, code is the perfect use case for an LLM–it is text-based and largely practical, and thereby under less scrutiny for human-like features like style or emotion. Furthermore, the ROI is clear–developer time is one of the most expensive and scarce human resources in the world today. Any AI-powered productivity augmentation for developers can be almost directly translated into incremental dollars saved or earned.

Introducing Magic

Enter Magic. Magic is leveraging proprietary LLM technology to develop an AI-powered software engineer. Far beyond simple code completion or next line prediction, Magic is building a true AI colleague that acts as a pair programmer and is able to understand and continually learn more about the context of the work and user. Eric and his team’s incredible research on neural memory and reinforcement learning allows for a truly differentiated product that will unlock significant value and open up a new market of use cases.

When we first met Eric, it was clear that he is a brilliant technologist, but it was his lifelong passion for AI and effusive excitement for what he was building that pulled us into his vision. We believe that Magic has the potential to revolutionize software development as we know it and to act as a genuine force multiplier for the productivity of every developer worldwide. Given that there are 25 million developers around the world today and are predicted to be 40 million by 2027, we can only begin to imagine what they will build for the world with Magic.

CapitalG’s Investment in Magic

We’re proud to lead Magic’s $23 million Series A round and to work closely with Nat Friedman, the former CEO of GitHub and inventor of Github copilot; Elad Gil, the serial entrepreneur and investor; and Mike Dauber, general partner at Amplify Partners, who are joining us in this investment. We are absolutely thrilled to support Eric and the entire Magic team as they push the boundaries of what we believe to be possible–and as they make magic happen one developer at a time.