Magic, an innovative AI firm focused on creating solutions to automate coding and various software development processes, has successfully garnered a significant investment from a group of investors, notably including former Google CEO Eric Schmidt.
In a recent announcement, Magic revealed it has completed a funding round of $320 million, with significant contributions from Schmidt, CapitalG (Alphabet’s investment arm), Atlassian, Elad Gil, Jane Street, Nat Friedman, Daniel Gross, Sequoia, among others. This influx of funds increases the company’s total capital to almost $465 million, positioning Magic among the elite AI startups focused on coding, such as Codeium, Cognition, Poolside, Anysphere, and Augment — with Augment also enjoying backing from Schmidt.
As reported by Reuters in July, Magic aimed to raise upwards of $200 million, targeting a valuation of $1.5 billion. Ultimately, the funding exceeded expectations, although Magic’s latest valuation remains undisclosed; as of February, the startup was valued at $500 million.
Moreover, Magic announced on Thursday its collaboration with Google Cloud to construct two “supercomputers” on the Google Cloud Platform. These include the Magic-G4, powered by Nvidia H100 GPUs, and the Magic-G5, which will utilize Nvidia’s forthcoming Blackwell chips, expected to debut next year. GPUs are crucial for running complex AI models due to their parallel processing capabilities.
Magic intends to expand this infrastructure to encompass tens of thousands of GPUs, aiming for a combined processing power of 160 exaflops, equating to one quintillion operations per second.
“Our collaboration with Google and Nvidia signifies a major step forward in building our next-generation AI supercomputer on Google Cloud,” declared Magic’s CEO and co-founder, Eric Steinberger. “The [Blackwell] system by Nvidia will significantly enhance our models’ efficiency, while Google Cloud presents the most rapid scalability option along with a comprehensive suite of cloud services.”
Magic was co-founded in 2022 by Steinberger and Sebastian De Ro. Steinberger shared with TechCrunch his early fascination with AI, recounting how he experimented with machine learning on his high school’s computers. Steinberger embarked on a computer science degree at Cambridge but left after a year, eventually joining Meta as an AI researcher. De Ro came from FireStart, a German business software firm, where he became CTO. The duo met through the environmental non-profit ClimateScience.org, co-founded by Steinberger.
Magic is developing AI tools intended to assist software developers in coding tasks such as writing, reviewing, debugging, and planning changes. These tools act as an automated coding assistant, striving to continuously understand and adapt to the context of different coding projects.
While similar platforms exist, Magic sets itself apart with its model’s exceptionally broad context windows, branded as the “Long-term Memory Network” or “LTM.” This capacity allows models to maintain awareness of extensive amounts of input data without losing relevance or accuracy.
The standout model, LTM-2-mini, boasts a context window of 100 million tokens, equivalent to about 10 million lines of code or 750 novels, marking it as the most extensive among commercial models, even surpassing Google’s Gemini flagship models at 2 million tokens.
Thanks to this expansive context, LTM-2-mini has successfully automated complex tasks like implementing a password strength meter and developing a calculator with a custom UI framework autonomously.
Despite its small team and current lack of revenue, Magic is pursuing a market projected to reach $27.17 billion by 2032, according to Polaris Research, attracting significant investor interest.
Amidst various concerns surrounding AI-driven development tools, their adoption continues to grow, as highlighted by GitHub’s recent survey showing widespread use among developers, with Microsoft’s Copilot boasting over 1.3 million paying users and significant business adoption.
Far beyond automating mundane tasks, Magic has its sight set on advancing towards AGI (Artificial General Intelligence), aiming for AI that can outperform human problem-solving capabilities.
As part of these efforts, Magic, based in San Francisco, recently welcomed Ben Chess, formerly of OpenAI, and is expanding its teams across cybersecurity, engineering, research, and systems engineering.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
