Space raises $2.4m pre-seed to build AI-native distributed filesystem
Space, a San Francisco-based startup founded in late 2025, has raised a $2.4 million pre-seed round led by a16z Speedrun, the accelerator programme run by Andreessen Horowitz. Golden Ventures and Northside Ventures also participated, alongside a dozen angel investors with backgrounds at companies including Parsec, Sentry, and Superwhisper. The funding will support development of what Space describes as an AI-native distributed filesystem, currently in private beta with roughly 100 teams onboarded.
The core proposition is deceptively simple: instead of syncing full files to a local drive before an application can open them, Space mounts a virtual filesystem directly above the operating system and streams only the precise byte ranges that any given application or agent actually requests. A video editor opening a multi-terabyte project, or an AI agent traversing an organisation's document store, would pull data on demand rather than waiting for an ingestion or download pipeline to complete.
How it differs from existing cloud storage
The release positions Space explicitly against Dropbox, Box, and Google Drive, arguing those tools are built around syncing entire files to local devices or routing workflows through web applications. Space, by contrast, operates at a lower level of the stack, appearing in the operating system's native file browser and remaining transparent to applications such as video editors, CAD tools, and code editors. No per-application integration is required.
Three co-founders built the first prototype in November 2025 after running into the same bottleneck at their previous company, where the team was moving terabytes of video footage each month for a content-production operation. Jason Zhao, one of the co-founders, had also accumulated dozens of terabytes of personal footage across drives during a decade of YouTube production.
Jonathan Lai, General Partner at a16z, said the team was "innovating with a new type of AI-native filesystem that can serve both human creators and AI agents with the same primitive: instant access to the exact data they need."
Market context and competitive landscape
The problem Space is addressing sits at the intersection of two growing pressures. First, the rapid expansion of agentic AI workflows has exposed a structural weakness in existing storage: most tools require a full file to be uploaded to a model or pipeline even when only a fraction of the content is relevant to the task at hand. Second, local disk capacity has not kept pace with the data volumes generated by modern creative, engineering, and AI-training workflows.
The distributed and virtual filesystem space has seen interest from a number of startups and open-source projects, but few have targeted the operating-system filesystem layer directly as a universal access primitive for both human users and autonomous agents. Space's closest conceptual antecedents are network-attached storage protocols and content-addressable file systems, though the company's framing around agentic access is more current.
Space says it will expand from its initial focus on video, marketing, and architecture, engineering and construction teams into AI training infrastructure, computer vision pipelines, and enterprise data systems. The company's longer-term vision, which it calls the Space Computer, is a machine that acts as a window into effectively unlimited cloud storage and compute rather than a standalone device.
At $2.4 million, the pre-seed is modest for a company pitching infrastructure-level ambition. Investors and enterprise buyers will be watching for disclosed performance benchmarks, named commercial customers, and evidence of security and compliance posture, particularly for regulated sectors where data residency and access-control auditability are non-negotiable requirements.