Magic Secures $320M Investment to Transform Generative AI Coding

Groundbreaking Investment for AI-driven Coding Innovation

Magic, an innovative AI startup focused on revolutionizing code generation and software development, has successfully secured a massive investment of $320 million from a distinguished group of investors, including former Google CEO Eric Schmidt. This funding marks a significant milestone for the company as it aims to redefine the landscape of coding through advanced AI technologies.

Key Contributions and Total Funding

In a recent announcement, Magic revealed that this impressive fundraising round included notable contributions from major players in the tech industry, such as:

• Alphabet’s CapitalG
• Atlassian
• Elad Gil
• Jane Street
• Nat Friedman
• Daniel Gross
• Sequoia
• And others

This latest funding round elevates Magic’s total raised capital to nearly $465 million, positioning the company alongside other well-funded AI coding firms like Codeium, Cognition, Poolside, Anysphere, and Augment. Interestingly, Eric Schmidt is also backing Augment, further illustrating the competitive nature of this rapidly evolving market.

Valuation Insights: Exceeding Expectations

Previously, reports indicated that Magic was seeking to raise over $200 million at a valuation of $1.5 billion. The recent funding round surpassed these expectations, although the current valuation of Magic remains undisclosed. As of February, the startup had been valued at $500 million, hinting at its growing finance and industry worth.

Strategic Partnership with Google Cloud

In tandem with the funding announcement, Magic disclosed an exciting partnership with Google Cloud to develop cutting-edge “supercomputers.” The two proposed supercomputers, named Magic-G4 and Magic-G5, will utilize powerful Nvidia H100 GPUs and the upcoming Nvidia Blackwell chips, respectively. These advances highlight Magic’s commitment to leveraging the latest technology for the training and functioning of its generative AI models.

Impressive Computational Capabilities

Magic’s vision extends to scaling the system’s GPU clusters to accommodate “tens of thousands” of GPUs, collectively aiming to achieve a staggering 160 exaflops of computational power—equivalent to one quintillion operations per second. Magic’s co-founder and CEO, Eric Steinberger, expressed excitement about collaborating with Google and Nvidia, emphasizing that these technological advancements will significantly improve both inference and training efficiency for their AI models.

Founders with Vision and Experience

Magic was co-founded in 2022 by Eric Steinberger and Sebastian De Ro, both of whom bring a wealth of experience to the table. Steinberger’s early interest in AI and his background in computer science laid the groundwork for his role in this venture, while De Ro’s expertise in business process management further complements the company’s goals.

Revolutionary AI Tools for Software Development

Currently, Magic is developing AI-driven tools designed to assist software engineers with various coding tasks. These tools aim to:

• Write code
• Review existing code
• Debug issues
• Plan modifications

Operating as automated programming assistants, Magic’s tools continuously learn from the context of different coding projects.

Unique Features of Magic’s AI Models

What sets Magic apart from competitors—like GitHub Copilot—is its innovative model architecture known as the Long-term Memory Network (LTM). This distinctive approach allows for ultra-long context windows. The context window in a model refers to the amount of input data it considers before generating output, enabling more relevant and accurate results.

Achievements with Extremely Large Context Windows

Magic’s latest model, LTM-2-mini, boasts a remarkable context window of 100 million tokens—equivalent to approximately 10 million lines of code or 750 full-length novels. This achievement positions it as the most advanced of any commercial model available today. The closest competitor, Google’s Gemini flagship, features a context window of only 2 million tokens.

Thanks to this exceptional context allowance, LTM-2-mini was able to autonomously develop practical features, such as:

• A password strength meter for an open-source project
• A calculator using a custom user interface framework

Magic is currently focused on training a larger version of this groundbreaking model to further enhance its capabilities.

Exploring a Booming Market

Although Magic currently operates with a small team of about two dozen and has yet to generate revenue, it is targeting a colossal market projected to reach $27.17 billion by 2032. Investors recognize the potential for substantial returns in this rapidly expanding sector of AI coding tools.

Developer Enthusiasm for AI Tools

Amid ongoing concerns about security, copyright, and reliability surrounding AI-assisted coding, developers have shown a robust inclination towards adopting AI tools. Survey results from GitHub indicate that a significant majority of respondents have integrated AI solutions into their workflows. Moreover, Microsoft has reported that Copilot currently boasts over 1.3 million paying users and more than 50,000 business customers.

Aiming for Greater Goals: The Path to AGI

Magic’s ambition extends far beyond mere automation of basic coding tasks. On their official website, they outline a vision that includes achieving Artificial General Intelligence (AGI)—AI capable of solving complex problems with greater reliability than humans. To reinforce this objective, Magic has recently recruited Ben Chess, a former leader in OpenAI’s supercomputing division, and plans to expand its teams in cybersecurity, engineering, and research.