Dropbox is spending $350 million on Tigris: the file-sync veteran wants to be the warehouse for AI agents
Dropbox agreed to acquire Tigris Data for about $350 million in cash and unveiled DBX Agent Cloud, a storage layer for AI applications and agents. The deal is expected to close in Q4 2026.

For many people, Dropbox is still "that blue box where files live". On November 5, it spent $350 million to tell everyone otherwise: it no longer wants to be just your filing cabinet. It wants to be the warehouse for AI agents. On the same day, it launched a new business, DBX Agent Cloud, and a new version of Dropbox that works directly on your files. After 19 years of selling sync, Dropbox has finally decided to sell separately the thing it is actually best at — operating storage at massive scale.
The facts first. On November 5, Dropbox announced it had signed an agreement to acquire Tigris Data Inc. for approximately $350 million in cash, subject to customary adjustments, with the deal expected to close in the fourth quarter of 2026. By several outlets' accounts, it is the largest acquisition Dropbox has made since it was founded in 2007. Tigris is a Sunnyvale startup founded in 2021 by former Uber engineers, building globally distributed, S3-compatible object storage that charges no egress fees. It has raised only about $31.9 million in total — a $25 million Series A in October 2025 led by Spark Capital, with Andreessen Horowitz and Basis Set participating. In other words, Dropbox is paying roughly eleven times everything Tigris ever raised, buying a product, a team, and developer relationships that already work.
" alt="Dropbox official blog illustration: orange artwork with files, charts and a lock floating over a grid surface">
The other half of the announcement is DBX Agent Cloud. According to Dropbox's official blog, it is a new infrastructure business "for companies and developers building AI applications and agents", starting with standalone S3-compatible object storage that can be bought independently, with filesystem and content-processing capabilities to follow. Co-CEO Ashraf Alkarmi put it more bluntly on the earnings call: a "new avenue for growth" that puts "the infrastructure we've built over nearly two decades" to work. Ovais Tariq, Tigris's co-founder and CEO, said: "AI is driving storage demand at a scale this industry has not seen before." Put the two quotes together and the message is clear: Dropbox believes storage demand from AI applications will be big enough to justify a standalone business.
That judgement is not made up. What AI applications and agents need from storage is a different thing from file sync. Training datasets, model artifacts and multimodal content have to flow between clouds and compute providers; agents need to read and write files persistently across multi-step tasks instead of re-uploading everything each time. The Dropbox blog post lands the key point in one sentence: "AI tools, models, and compute environments will continue to change. The content those systems depend on, along with its permissions, history, and context, needs a consistent home." DBX Agent Cloud wants to be that home: a storage and data layer that works alongside the clouds, compute providers and developer platforms customers already use, instead of asking them to move house.
There is a small irony tucked inside the deal. Dropbox spent years moving most of its user data off Amazon S3 and building its own storage, which is exactly how it acquired the expertise it now wants to sell — including to developers currently running AI workloads on S3. Tigris's pitch fits the same logic: S3-compatible, so developers can use their existing tools; globally distributed, so data sits closer to compute; no egress fees, which is real money saved when AI workloads routinely move terabytes of training data around. Of course, DBX Agent Cloud is still just a beginning: no pricing has been published, no service-level agreement, no launch date, and the filesystem and content-processing layers are still in the "coming next" column. What buyers can actually buy today is a direction and a team.
The other announcement from the same day is, in practice, closer to ordinary users. Dropbox launched a new experience in which AI works directly on the files you already have: finding information, answering questions about files and projects, summarising folders, comparing versions, drafting documents, editing PDFs and images, transcribing recordings, trimming video clips — and carrying out several steps from a single request, such as building a project folder structure or moving and renaming files. The company says the experience has already rolled out to more than 70,000 teams, starting with eligible Standard and Advanced teams in the US, UK and Australia, in English first, with broader availability planned for 2027. Control is the centrepiece of the pitch: Dropbox shows a plan for approval before making changes, works within existing permissions, lets you undo, and keeps a record of actions taken. Models can be switched within a session, and files do not need to be uploaded and downloaded repeatedly. Dropbox files can also be opened directly in ChatGPT, Claude and Muse, with the finished work landing back in the original folders.
The earnings numbers add the other half of the story. Dropbox's third-quarter revenue was $637.3 million, up just 0.5% year over year; total ARR was $2.555 billion, up 0.8%; paying users totalled 18.20 million, a net gain of only 20,000 in the quarter; but GAAP net income was $116.1 million, operating margin 27.5%, and cash, equivalents and short-term investments stood at $1.142 billion. This is a company whose growth has nearly stalled but which is solidly profitable — which is exactly why $350 million for Tigris reads as a serious bet: the ceiling on the core business is already visible, and new growth has to be found outside it. Alkarmi noted in the release that this was the fifth consecutive quarter of year-over-year revenue growth excluding FormSwift. The subtext: the old business is stabilised; now comes the new story.
The new story has formidable competition. The object-storage arena already holds Amazon S3, Cloudflare R2, Backblaze B2 and a wave of newer players aimed at AI workloads. Dropbox is betting its differentiation on two things. First, it has genuinely operated storage at enormous scale — 700 million registered users, around 180 countries, nearly two decades of operations — which no young startup can buy. Second, it holds the "content" card: DBX Agent Cloud is not just selling "cheap places to put things" but storage whose contents carry permissions, history and context, ready for agents to use on arrival. That logic is especially attractive to indie developers and small teams: what your agent application most often lacks is not compute, but a place where files survive across sessions without turning into a mess.
But a cooler look is warranted: today, DBX Agent Cloud is still an announcement plus a not-yet-closed acquisition agreement. The deal is only expected to close in Q4, the product has no pricing and no launch timetable, and the filesystem and content-processing capabilities are still on the way. The Tigris team has to be integrated into Dropbox, the product has to be genuinely sold to developers, and developers have to be willing to entrust their AI workloads' data to a company famous for "file sync". All of that still has to be proven. Dropbox's largest acquisition ever may also be the largest IOU on its path to reinvention.
For VibeFix readers, the point of this story is not the $350 million figure but the trend it confirms: the infrastructure of the agent era is surfacing layer by layer. The models exist, the agent frameworks exist, and now the question of "where does an agent's data live, and how does it stay alive across sessions" is being taken seriously. Dropbox's answer is that content needs a home, and that home is best independent of any single cloud vendor. Whether that answer is right will only be known once the product ships, pricing is published, and developers vote with their usage. But one thing has already changed: the "blue box for files" has formally started reintroducing itself for the AI era.
Comments (0)
Sources
Related articles
On 18 November 2026, Google Research and co-authors posted BuggyMLE to arXiv: bugs injected into 222 working ML repositories, and 14 models slide from 95% on the easiest tier to 9% on the hardest. A 9-billion-parameter model trained easy-to-hard beats Claude Opus 5.5 on the 10-bug tier.
On 17 November 2026, Microsoft launched MAI Code 2 and Code 2 Flash: the larger model has 77 billion active parameters and a one-million-token context, averaging 63.2% across four benchmarks — close to GPT-6.1 Sol and Claude Opus 5.5 — at $0.309 per task. Read the three caveats first: promotional pricing, the lowest score of the three, and no long-horizon work.
On 11 November 2026, AWS open-sourced UAEF: adapters turn traces from LangGraph, LangChain, Strands and Bedrock AgentCore into one format, then 36 metrics across 7 dimensions score them, flagging regressions past 5%. The judge model is hardened too — scores must cite verbatim evidence, and never decide safety.