On October 9, Ryan Dahl announced that the entire Deno team is joining Cloudflare, that Deno Deploy will run for six more months before shutting down, and that the Deno runtime gets one last year of maintenance releases before Deno stops developing it. Terms were not disclosed. Cloudflare's joint post with Dahl presents the deal as a merger of Deno's celld into workerd, the open-source runtime behind Cloudflare Workers. Ten months after Anthropic bought Bun, both of Node.js's best-known challengers now live inside larger companies.
The timelines that matter
Deploy keeps running for six months, which by simple arithmetic puts the end at roughly April 2027, and Deno promises migration support for paying customers moving to Cloudflare Workers. Nothing is promised to free-tier users. As of this writing, the Deploy product page shows only a "joining Cloudflare" banner, with no shutdown date and no migration guide.
The runtime gets a year of monthly releases with bug fixes and security updates; after that, Deno ends its development, keeps the code open source and invites others to carry it on (Deno). The license stays MIT (The New Stack). A follow-up on October 15, as reported by Socket, set out the mechanics: monthly 2.9.x LTS releases through October 2027, continued security triage, no new features, and feature pull requests closed unmerged. One loose end: Deno's stability documentation still listed the 2.9 LTS window as ending January 31, 2027 when we checked.
JSR, the package registry, keeps operating, with its infrastructure moving to Cloudflare. Deno will also keep supporting rusty_v8 and work toward integrating it into workerd.
What Cloudflare is actually buying
Celld, released in August, is Deno's open-source take on the Workers and Durable Objects model (Simon Willison): by Cloudflare's description, one Rust binary whose only external dependency is object storage. Dahl and Deno co-founder Bert Belder will lead an effort to make self-hosted workerd a first-class way to run Workers apps, merging celld's code and ideas into it. The New Stack reports the combined project should stay open source, likely under celld's Apache 2.0 license, and DevClass reports that Dahl and 12 colleagues made the move, which makes this as much an acquihire as an acquisition.
Cloudflare is candid about why it wanted them. Its post concedes that workerd's Durable Objects support was limited to a single instance and was never production-ready for self-hosting, and it argues that open source as an escape hatch is simply good business. The thesis, in its own words: "we believe we've simply created a better programming model for cloud software" (Cloudflare).
Why the challenger stopped challenging
The arc is familiar. Dahl seeded the ideas behind Deno in a June 2018 JSConf EU talk on his regrets about Node); Deno 1.0 shipped on May 13, 2020; the company raised a $4.9 million seed in 2021 and a $21 million Series A led by Sequoia in June 2022, for $26 million in total. Deno 2 arrived on October 9, 2024, pitched on access to more than 2 million npm modules, two years to the day before the Cloudflare announcement. Adoption more than doubled after Deno 2 by Deno's own monthly-active-user metric, and DevClass puts the base at 400,000 monthly users as of August. That same 2025 post disclosed that Deploy had shrunk from 35 regions to six, citing cost and usage.
Dahl's own diagnosis is the cleanest. On Hacker News he wrote that Deno "has been sucked into the gravity well of node compatibility" (via Simon Willison). The npm compatibility that made Deno 2 adoptable also made it harder to justify as a separate thing.
There is unfinished business, too. In November 2024 Deno petitioned to cancel Oracle's "JavaScript" trademark on grounds of genericness, abandonment and fraud, and expected discovery to stretch into 2026. Deno's announcement does not mention the petition, and we found no current status.
Consolidation, with honest counterpoints
Put the two deals side by side. Anthropic's purchase of Bun, announced December 3, 2025, also came with undisclosed terms and a promise that the runtime stays MIT-licensed; Bun then counted more than 7 million monthly downloads. Node.js keeps its steady cadence: v26.0.0 shipped May 5, 2026, and the 26.x line is now nodejs.org's latest LTS. The challengers kept the ideas, Node kept the ecosystem, and the ideas moved into companies that sell something else.
Standards work helps explain why absorption is cheap. In January 2025, WinterCG's work began moving to Ecma as TC55, chartered to standardize a minimum common API for server-side runtimes, with Deno's Luca Casonato as a co-chair. Portable code lowered the cost of trying Deno. It now lowers the cost of leaving.
Not everyone is calm. Socket documents developers who felt misled by how the wind-down was presented, early forks such as Ende and Dodo, and calls for a Deno Foundation or a home at the OpenJS Foundation. Simon Willison called it bad news for Deno users. Dahl points to io.js as precedent for a community carrying a runtime forward, and Cloudflare's self-hosting promise is still a plan, with more announcements due in the coming months.
The 30,000-foot read
What can be controlled is the calendar. If you run on Deno Deploy, plan for roughly April 2027, and note that the only promised help is for paying customers moving to Workers. Start the inventory now; the announcement says nothing about Deno KV, so export anything Deploy-hosted early rather than waiting for tooling. If you run the Deno runtime, a year of patches is a runway, not a home. The irony is that the Node compatibility Dahl calls a gravity well is also the exit ramp, to Node 26 or to Bun. JSR users have the least to do, since the registry continues on Cloudflare infrastructure, though publishing to npm as well is a cheap hedge.
What cannot be controlled is who owns the runtime. Watch for a Deploy shutdown date and migration tooling, a foundation or fork that earns trust, the first release of the merged workerd and its license, and the fate of the Oracle petition. The quieter lesson: independent runtimes were a good idea that never found a business. They found landlords instead, and the rent is paid in roadmap.
Sources: Deno: "Deno is joining Cloudflare" (team move, six-month Deploy window, migration help for paying customers, one year of runtime releases then end of development, JSR, rusty_v8; primary); Cloudflare: "Deno is joining Cloudflare" (joint post by Kenton Varda and Ryan Dahl: celld into workerd, Dahl and Belder roles, celld architecture, workerd's self-hosting limits, Cloudflare quote; primary); The New Stack (undisclosed terms, MIT license, likely Apache 2.0 for the merged project, io.js precedent); Socket (October 15 LTS details, seed round, forks, community reaction); DevClass (12 colleagues, 400,000 monthly users); Simon Willison (celld's August release, Dahl's Hacker News comment, his assessment); Deno docs: stability and releases (2.9 LTS window); Deno Deploy (no shutdown date posted); Deno 1.0 (release date); Announcing Deno 2 (release date, 2 million npm modules); Reports of Deno's Demise Have Been Greatly Exaggerated (adoption doubling, Deploy regions); TechCrunch ($21 million Series A, Sequoia, $26 million total); Wikipedia: Deno) (date and venue of the 2018 talk only); Deno v. Oracle, part 3 (trademark petition); Anthropic (Bun acquisition date, terms, license, downloads); Node.js download archive (v26.0.0 date, latest LTS); Ecma International (WinterTC/TC55). CAVEATS: the 12 colleagues and 400,000 monthly users figures are single-sourced to DevClass, and the $4.9 million seed to Socket (consistent with TechCrunch's $26 million total). The October 15 LTS details come via Socket's report; we could not load Deno's original notice, and Deno's docs still showed a January 31, 2027 end for 2.9 LTS, which conflicts with the October 2027 horizon. Deno's own post says "open source" without naming a license; MIT is per The New Stack, and Apache 2.0 for the merged project is The New Stack's expectation, not a Cloudflare commitment. Dahl's Hacker News comment was read via Simon Willison's quotation, not directly. "Roughly April 2027" is our arithmetic; no shutdown date has been posted. The Deno KV advice is our recommendation, based on the announcement's silence. Correction to our discovery notes: Deno ends runtime development after the support year, not merely a year of fixes, and its post does not itself name the MIT license.