Magic builds frontier AI models for software engineering and research. The company develops LTM (long-term memory) models with 100M-token context windows and an AI software engineering coworker that assists with code generation, debugging, and complex development tasks. Magic partners with Google Cloud and operates thousands of GB200s to power its AI infrastructure.
Founded in 2022 in San Francisco, Magic has raised over million in total funding across Seed, Series A, and Series B rounds. The company publishes research on large context models and collaborates with partners to deploy its AI coding assistant for enterprise and research use cases.
The AI-assisted software engineering market is experiencing rapid growth as enterprises seek to accelerate development cycles and reduce technical debt. Magic operates at the frontier of this space with its LTM models that can process 100 million tokens, far exceeding the context windows of most competing solutions. This positions the company to capture share in the expanding market for AI coding assistants and autonomous software engineering agents.
Competition is intensifying from well-funded players including GitHub Copilot, Cursor, and Anthropic, as well as from open-source alternatives. However, Magic focus on ultra-long context understanding and its deep partnerships with cloud providers create durable differentiation. The total addressable market for AI developer tools is projected to reach tens of billions of dollars within the next five years.
Magic differentiates itself through frontier long-term memory (LTM) models capable of 100M-token context windows, enabling deep comprehension of entire codebases and complex research documents. This technological leap allows Magic to outperform conventional coding assistants on large-scale software engineering tasks.
The company has secured strategic partnerships with Google Cloud and assembled a world-class team of researchers and engineers. With over million in funding from top-tier investors including CapitalG, Sequoia, and Eric Schmidt, Magic has the capital and compute infrastructure—thousands of GB200 GPUs—to train and deploy next-generation AI models at scale.
