AI-native filesystem startup Space raises $2.4 million led by a16z Speedrun

San Francisco-based startup Space has raised $2.4 million in pre-seed funding to build an AI-native virtual filesystem. The round was led by a16z Speedrun, marking a push to eliminate the traditional requirement of downloading files locally before they can be accessed.

The funding round included participation from Golden Ventures and Northside Ventures, alongside approximately a dozen angel investors from companies including Sentry, Parsec, Modem, Superwhisper, and Stan. Space is currently operating in a private beta, supporting roughly 100 users and teams.

Moving beyond the sync model

Traditional cloud storage providers, such as Google Drive and Dropbox, generally rely on a "sync and download" model where a full file must be transferred to a local device before an application can open it. While specialized tools like LucidLink have provided streaming capabilities for the film production industry, Space aims to standardize this approach for all computers and AI agents.

The technology functions as a virtual filesystem that streams only the specific byte ranges required by an application in real-time. This allows users to interact with files that exceed their local disk capacity. Co-founders Matthew Ao, Arihant Bapna, and Jason Zhao—who previously worked on Eden and Kortex—were driven to solve this problem after struggling to manage terabytes of video footage.

Implications for AI and enterprise data

As the volume of data creation continues to outpace hardware storage capacity, the traditional download model has become a bottleneck for large-scale AI workflows and high-resolution media. By creating a filesystem primitive that AI agents can navigate directly, Space removes the need for separate indexing or ingestion steps.

"The computer no longer needs to hold every file," co-founder Matthew Ao said, describing the device as a "window" into any file regardless of size or location. Co-founder Jason Zhao added that neither humans nor agents should have to wait for files to be copied before work begins. Jonathan Lai, a general partner at a16z, noted that the company is disrupting the long-held computing assumption that a file must exist on a device to be usable.

What's next

Space is currently focused on expanding its private beta as it refines the streaming primitive for broader application. The company's success will likely depend on its ability to maintain low latency across diverse network conditions, ensuring that the "window" into cloud files remains seamless for both human users and autonomous AI agents.