Clipto has raised $15 million at a $250 million post-money valuation to index the video, audio, images and documents already sitting on a user's own machine. The San Francisco company, which also has teams in Singapore and Hong Kong, reached $15 million in annual recurring revenue in early 2026 and says it is profitable with a little over 20 employees. HSG — formerly Sequoia China — GL Ventures, EnvisionX Capital, Palm Drive Capital, Hans Tung, Lu Zhang and 522 Ventures took part. The bet runs against the direction of almost everything else funded this cycle: not making more files, but finding the ones people already have.
The valuation is 16.7 times revenue, which is unremarkable for AI in 2026 and quite remarkable for a profitable twenty-person company. Those two facts pull in opposite directions. A business that already covers its costs at $15 million ARR does not need $15 million; a business raising at that multiple is being priced on something other than its current revenue.
What Clipto does is narrow and easy to describe. It indexes the video, audio, images, meetings and other files on a computer, so that instead of walking through folders the user describes what they are looking for — or hands the search to ChatGPT, Claude or another AI tool. Everything is processed on the device. No cloud service is involved.
Founder Henry Kan has been circling the same problem for roughly twenty years. In 2006, as a doctoral student at Carnegie Mellon, he worked on robots that recorded their surroundings, recognised objects and remembered where to find them again. His first company applied the same core idea to a wardrobe, using AI to keep an inventory of a user's clothes and suggest outfits. His second, ZenVideo, worked on making video easier to produce; Kan says Tencent acquired it in 2020. He founded Clipto in 2023 with several members of that founding team.
Kan's framing is that the AI era's problem is surplus, not scarcity: users' machines have accumulated more video than anyone will ever go back to. While the rest of the industry spent those three years on generating new material, Clipto worked on finding and reusing what already existed — a position that looked eccentric in 2023 and looks like patience now.
The user mix has drifted a long way from the original plan. Clipto was built for video creators managing material scattered across computers and external drives; creators are now only about a quarter to a third of the app's audience. The rest are lawyers, doctors, researchers, marketing specialists, HR staff, teachers and students. More than 30 million people have used the company's apps since launch, and paid subscribers number in the hundreds of thousands. Kan declined to give an exact subscriber count or average revenue per user, but said a substantial share of users stay subscribed for more than two years.
The event that actually explains the round happened about two weeks before it. Clipto added support for the Model Context Protocol, the standard that lets AI applications connect to outside sources of information. Access to the index requires an active request and the user's permission, and an AI application can only reach what falls inside the scope the user has granted.
That two-week gap is the whole story, and I doubt it is a coincidence. Indexing a hard drive is a feature, and a feature Apple and Google give away. Being the thing an agent calls when it needs to know what is on a hard drive is a position. Clipto did not raise as a search app; it raised as local-file infrastructure for models that cannot see local files. The new money going into AI models and the on-device compute infrastructure to run them, plus connections to other AI agents, is consistent with that reading and not much else.
The numbers Kan withheld are partly constrained by the ones he gave. Hundreds of thousands of paying subscribers against $15 million in ARR puts average revenue somewhere in the range of roughly $20 to $75 a year — consumer pricing, not the per-seat professional pricing you would expect from a user base of lawyers and doctors. That gap is the most interesting thing in the disclosure. Training and serving models on users' own devices is an infrastructure cost structure, and it is being carried by a consumer subscription. Either the pricing moves or the model work stays small.
Kan's case against the incumbents is that Clipto searches video, audio, images and documents at once, while Adobe, Apple and Google each search inside their own service — Adobe offers AI search in Premiere, Apple Photos and Google Photos find images and video from natural-language descriptions. He is right about the current shape of those products. But that shape is a product decision on their side, not a technical limit, and each of them can widen its own search without asking anyone to install anything. Clipto's advantage is that it already crosses the boundaries between those services. Its exposure is that the boundaries belong to someone else.