Cloudflare Acquires Deno, and the Deno Runtime Gets One More Year of Monthly Releases

Cloudflare said on Friday, October 9, that the team behind Deno, the JavaScript and TypeScript runtime co-founded by Node.js creator Ryan Dahl and Bert Belder, is joining the company. TechCrunch reports it as an acquisition, says the financial terms were not disclosed, and notes that Deno had raised a total of $26 million, including a Series A led by Sequoia. Cloudflare’s announcement frames the deal around self-hosting: the team is joining “to radically simplify self-hosting Workers and Durable Objects.” What it means for current Deno users is in a separate post by Dahl on Deno’s blog. Cloudflare’s post points readers to it for “what’s happening to the Deno runtime,” and TechCrunch’s short report does not cover it. Dahl writes that Deno will keep releasing the runtime for another year and then end its development, and that Deno Deploy will shut down after six months.

I read both company posts and checked the repositories, documentation and earlier coverage they lead to. I did not run celld or workerd, so the technical claims below are the companies’ own.

What Deno Says Will Happen to Its Products

Dahl says the team has “decided to put our future development work into this shared platform rather than continuing to develop a separate runtime and hosting service,” and adds: “We know this is a consequential change for people who have built on Deno.” The concrete commitments are short:

Neither post gives calendar dates. Counting from the October 9 announcement, six months ends around April 2027 and a year around October 2027; that is my arithmetic, not a date Deno has stated. The wording covers bug fixes and security updates and does not promise new features. For scale, the GitHub release list shows eight releases from v2.9.0 on June 25 to v2.9.7 on September 17, and the last two gaps were three weeks each.

The Deploy on the six-month clock is probably the newer platform. Deno’s migration guide, last updated June 18, said Deploy Classic and the subhosting v1 API “will be shut down on July 20, 2026,” and sent customers to the new Deno Deploy. If that date held, the product now facing a six-month countdown is the one Deno had been steering customers toward this summer.

What the Announcement Leaves Out

Neither post mentions Fresh, Deno Sandbox, Subhosting, Claw Patrol or Deno for Enterprise, all of which appear in the product menu on the page that carries Dahl’s announcement. Neither names who would maintain the runtime after the year ends; Deno says only that “we welcome others who want to continue its development.” And in May 2025 Dahl wrote: “We’ve recently established an Open Governance Board and are actively working to transition JSR into an independent, community-driven foundation.” The October post says only that JSR keeps operating, with its infrastructure moving to Cloudflare.

How This Compares With Dahl’s May 2025 Message

On May 20, 2025, answering critics who said Deno was fading, Dahl wrote that since Deno 2 shipped the previous October, adoption had “more than doubled” by Deno’s monthly-active-user count. His post, covered by InfoWorld, closed: “We’re not winding down. We’re winding up.” The October 9 announcement comes a little over sixteen months later. It gives no usage figures. Dahl presents the move as a choice about focus (“Joining also means choosing where to focus our effort”) and as the next step in a progression: “The progression from Deno to Deno Deploy to celld explains why this move makes sense to us.”

What Cloudflare Gets: Durable Objects Outside Cloudflare

Kenton Varda, the Workers principal engineer who co-wrote Cloudflare’s post, says Cloudflare open sourced its Workers runtime, workerd, because customers such as Shopify said in 2022, when Cloudflare pitched them Workers for Platforms, that they could not build on it unless the runtime was open source. It is, he writes, “the same code we run in production.” The gap is Durable Objects: workerd implements them “only in a single-instance way, good enough for local testing, but unable to scale,” with TODO comments “going back to the original release.” Cloudflare’s production routing, Varda says, “is simply not the implementation any self-hoster would want.” Varda says he tried to build a self-hosting version himself: “I actually made an attempt last spring, but, embarrassingly, it didn’t work.”

That is the gap celld was built to fill. Dahl describes it as “one binary, written in Rust, with object storage as its only external service dependency.” Varda says Dahl and Belder will lead an effort that “will involve merging code and ideas from celld back into workerd,” with more announcements “in the coming months.” celld’s own site lists version 0.6.2 and labels it beta. Its repository is Apache-2.0 licensed and mostly Rust, while workerd is mostly C++. The site says an existing Cloudflare app “runs unchanged” across Workers, Durable Objects, KV, Queues, D1, R2, Workflows, Cron and static assets, with containers marked experimental and anything that needs Cloudflare’s network, GPUs or a browser farm out of scope.

The Lock-In Argument and a Cost Claim

Varda’s half of the post argues against “a popular theory” that Workers was made different from other cloud platforms to create lock-in. His answer is that “being open source and giving people an escape hatch is good business.” He also concedes: “Frankly, we haven’t done a great job of building out the ecosystem of adjacent services and tooling to make it really work.”

celld’s own site makes a cost claim. Its home page says “Orders of magnitude cheaper at scale,” and a cost chart lists about $0.02 per resident cell-month on $48, 8 GB nodes capped at 2,500 resident cells each, against $4.15 per resident cell-month for Durable Objects on the Workers Paid plan. The page says both base costs “exclude the compute and bucket usage from application traffic and writes.” Dividing the two gives roughly 200 times on a fully packed node (my arithmetic, using the page’s own numbers), and it assumes you run the nodes yourself. Cloudflare’s post does not mention those figures, and I have not benchmarked either system.

Where This Fits

Cloudflare’s Acquisitions tag page lists Astro joining on January 16 and VoidZero, “the team behind Vite, Vitest, Rolldown, Oxc, and Vite+,” on June 4. Its summaries describe continued investment: Cloudflare is “doubling down on making Astro the best framework for content-driven websites,” and “Vite stays open source, vendor-agnostic, and built for everyone.” Anthropic’s December 2025 purchase of Bun, another JavaScript runtime, came with a pledge that “Bun will remain open source and MIT-licensed.” Deno’s announcement differs on the point that matters most to its users: it ends development of the runtime itself after a year, while leaving the MIT-licensed code open for anyone who wants to carry it on.

What to Watch

• Calendar dates for the end of Deno Deploy and for the last runtime release.
• Whether anyone takes over the runtime before the year is up, since the post names no successor.
• What happens to Fresh, Deno Sandbox, Subhosting, Claw Patrol, Deno for Enterprise and the JSR foundation plan.
• The further announcements Varda promises about merging celld into workerd.

For more on Cloudflare’s developer platform, see our earlier coverage of its Artifacts git service for agents and its Traces beta.