Anna Tong and Crystal Hu
(Reuters) – Magic, a U.S. startup that makes artificial intelligence models for software development, is in talks to raise more than $200 million in a funding round that values the company at $1.5 billion, just months after its last fundraise, three sources familiar with the matter told Reuters.
Investors including Jane Street are expected to participate in the round, which could triple Magic’s valuation from its previous round, despite the company having no revenue or product sales, according to the people, who asked not to be identified discussing private matters.
The startup, which has about 20 employees, was valued at $500 million after its last funding round in February, according to Pitchbook data. The company has raised $140 million since its founding in 2022 from backers including Nat Friedman and Daniel Gross’s NFDG Ventures and Alphabet’s Capital G.
The Magic declined to comment, and Jane Street did not respond to a request for comment.
Magic’s new funding round is the latest in one of the most promising applications of generative AI technology: Software developers represent a big cost for tech companies, making tools that can generate code or help developers program more efficiently very attractive.
The early success of Microsoft’s GitHub has further fueled investor enthusiasm: In April, coding assistant startup Augment raised $252 million, and Cognition, maker of the AI-powered coding assistant Devin, secured $175 million at a $2 billion valuation in a round led by Founders Fund.
Investors say they see the success of GitHub Copilot as evidence of the size of the market: GitHub reported a 40% increase in revenue year-over-year in its most recent quarter, thanks in part to Copilot, which has 1.3 million paying subscribers.
“Microsoft’s success has validated a commercial market for AI code assistants and led everyone to believe there is clear market demand and customer willingness to pay for the right product. The opportunity in this space is huge and there will likely be multiple winners,” said Brian Dudley, partner at Adams Street Partners.
Commercially available products like GitHub Copilot and OpenAI’s ChatGPT can suggest ways to complete a line of code, but the next frontier in coding assistants is designing and writing entire software applications without human help.
Magic and several other startups are working to achieve this goal by training their own large-scale language models for coding-specific tasks, but the effort comes at a steep cost in terms of purchasing data, chips and electricity.
Magic plans to spend the money on improving its models to support long context windows — AI systems that can process more data in a single query — the people added.
Magic’s ability to understand and process large amounts of context at once comes from an innovative design that goes beyond traditional “transformer models” commonly used in large-scale language models such as OpenAI’s GPT model, one of the sources said.
Poolside AI, a Paris-based startup with a similar approach to Magic, was in talks to raise $450 million at a $2 billion valuation, TechCrunch reported in June. Poolside also doesn’t have a commercially available product.
(Reporting by Crystal Hu in New York and Anna Tong in San Francisco; Editing by Rod Nickel)
