Space Raises $2.4M to Build an AI-Native Filesystem

Space has raised $2.4M in Pre-Seed funding to build an AI-native distributed filesystem that lets people and software agents work with cloud-hosted files as if they were stored locally. IT Brief Canada reported that a16z Speedrun led the round, with Golden Ventures, Northside Ventures, and angel investors from companies including Parsec, Sentry, Stan, Superwhisper, and Modem participating.

The financing matters because cloud storage solved the price and capacity problem more cleanly than it solved access. Teams still lose time downloading, syncing, indexing, copying, and compressing files before an application or AI agent can do useful work, which means the cloud often behaves like an enormous warehouse attached to a narrow loading dock.

Space, founded by Matthew Ao, Arihant Bapna, and Jason Zhao, is rebuilding that access layer. Its filesystem mounts cloud storage as a native volume and streams only the byte ranges an application requests, allowing a computer to become a window into remote data instead of a container that must hold every file in advance.

What Happened

The $2.4M Pre-Seed gives Space institutional backing from a16z Speedrun alongside Golden Ventures, Northside Ventures, and experienced software operators. The investor group connects early-stage product development with practical experience in creative software, developer infrastructure, monitoring, audio, and large-file workflows.

The first Space prototype was built in late 2025, and the team later based itself at Founders Inc. in San Francisco while developing the product and raising the round. The company says its private beta has onboarded about 100 users and teams, and it reports an audience of more than 80,000 across platforms. Those figures are company-reported rather than independently audited, but they provide an early measure of attention around a stubborn infrastructure problem.

Why File Access Is the Real Product

Conventional cloud drives generally sync files to a device or move the user into a browser-based workflow. Space works beneath that interface at the filesystem layer, where existing applications see a familiar drive even though the data remains remote until a particular byte range is requested.

That architecture is designed for work that overwhelms local storage. A video editor can open a project larger than the laptop, an architect can work with a substantial building model, and a developer can navigate a large repository without first copying the whole asset to the machine. The same approach can let an AI agent inspect only the relevant parts of a data store instead of forcing a separate upload or full ingestion pipeline.

The useful distinction is not simply that Space offers more storage. It changes when data needs to move, which can reduce duplicated files, long transfers, and the collection of connectors teams currently maintain between storage systems and applications.

The Founding Problem Was Operational

Space grew from the founders' own experience managing large volumes of media and company data. Jason Zhao encountered the problem while handling dozens of terabytes of YouTube footage, and the issue widened at the founders' previous company as the team moved terabytes of footage each month and lost time to uploads, downloads, and compression.

That history gives the company a credible starting wedge. Video, marketing, architecture, engineering, and construction teams already feel the cost of local capacity limits, while creative applications are often difficult to move into fully browser-native workflows. Space can meet those users inside the tools they already know rather than asking them to rebuild the workflow around another portal.

Why AI Changes the Filesystem Question

AI agents have made an old storage problem more visible. An agent may need one configuration file, a few frames from a video, or a narrow slice of a dataset, yet many systems still require a complete copy or a separate indexing process before the task begins.

Space is betting that people and agents should share the same live filesystem primitive. If that model works, applications can ask for exactly the data they need while the storage layer handles location and transfer behind the scenes. That creates a cleaner foundation for agentic work than adding one connector for every tool and another copy for every workflow.

The company identifies media, training-data infrastructure, computer vision, and enterprise data systems as longer-term areas of expansion. These markets have different buyers and security requirements, but they share the same underlying pressure: data volumes are growing faster than the local disks and manual transfer habits built around them.

Competitive Context

Dropbox, Box, and Google Drive made cloud storage mainstream, while network filesystems and content-delivery technologies have long separated physical data location from the user experience. Space is not entering an empty field, and established vendors have distribution, trust, compliance programs, and years of infrastructure investment.

Its opportunity is to make the filesystem itself feel native across human and agent workflows without forcing users into a browser or a full local sync. The company must still prove reliability, security, collaboration behavior, economics, and performance under real production loads. Infrastructure products earn trust one uneventful file operation at a time, which is a slower and less glamorous test than a product demo.

What the Round Signals

Jonathan Lai, the a16z general partner focused on Speedrun, framed Space as a challenge to the assumption that a file must exist on a device before anyone can work with it. That thesis fits a broader infrastructure shift in which compute, storage, and software agents are increasingly distributed, while the user still expects an immediate local experience.

For operators, the lesson is simple: capacity is no longer the only storage metric that matters. The next layer of value may sit in reducing the time between identifying data and acting on it, especially when the user is software working at machine speed. Space now has $2.4M to prove that access can become the new capacity without making reliability, security, or control someone else's problem.