Andy Garcia · freelance dev
FRENES
Back to the blog
Open Source

Cloudflare Gets the Deno Team. Deno Deploy Gets Six Months, and the Deno Runtime One Year of Fixes

October 19, 2026 6 min read

On Friday, October 9, 2026, Ryan Dahl, who created Node.js and then Deno, announced that the entire Deno team is joining Cloudflare. His post on the Deno blog says what that means in four lines. The Deno runtime gets another year of monthly releases with bug fixes and security updates, then the team ends its development; it stays open source, and others are welcome to continue it. Deno Deploy, the hosting service, keeps operating for six months and then shuts down, with migration support for paying customers moving to Cloudflare Workers. JSR, the package registry, keeps running, on infrastructure that moves to Cloudflare. And rusty_v8 stays supported, with work toward integrating it into workerd, Cloudflare's open-source Workers runtime.

The same day, Dahl and Kenton Varda of Cloudflare published a joint post on the Cloudflare blog. workerd and celld, Deno's self-hosted take on Workers, are being merged, and Dahl and Bert Belder will lead the effort to make self-hosting workerd "a first-class supported way" to run apps on the Workers model. Neither post uses the word acquisition or names a price. The New Stack described it as Cloudflare buying the startup and reported that terms were not disclosed.

I orchestrate my work in Claude Code, which ships as a Bun executable, so what happens to a JavaScript runtime when its company changes hands is not an abstract question for me. I read both posts, the Deno Deploy pricing page, the migration docs and the release history. The clocks are clear. The data is not mentioned.

What Cloudflare is getting

The asset in this story is celld. Deno published its first GitHub release, v0.1.0, on August 5. It is a daemon written in Rust that runs a Cloudflare Workers application on your own machines, from the wrangler.json you already have: Workers, Durable Objects, KV, Queues, D1, R2, Workflows and Cron Triggers. Its long-term state lives in an S3-compatible, Google Cloud Storage or Azure bucket. The license is Apache 2.0.

That fills a gap Varda admits in his half of the post. Cloudflare open-sourced workerd after pitching Workers for Platforms to customers like Shopify in 2022, but its Durable Objects run "only in a single-instance way, good enough for local testing, but unable to scale." Cloudflare's own production routing is, in his words, "a beast" that no self-hoster would want. He told The New Stack why the fix matters to Cloudflare: "If the model is exclusive to us, that narrows our TAM [total addressable market], and ultimately makes us less competitive versus the legacy clouds."

The Deno runtime is not part of that plan, and Dahl does not pretend otherwise. In a Hacker News comment quoted by Simon Willison, he wrote that Deno "has been sucked into the gravity well of node compatibility, which forces it to behave exactly as Node does. Why reimplement Node? It works."

Two runtimes, two buyers

On December 2, 2025, Bun announced that it had been acquired by Anthropic. Its post promised that Bun stays MIT-licensed and "extremely actively maintained," and gave the reason in two sentences: "Claude Code ships as a Bun executable to millions of users. If Bun breaks, Claude Code breaks."

Deno's runtime is MIT-licensed too; the license file in the repository still says so. The license did not predict the difference between the two outcomes. What predicted it is whether the buyer ships its own product on the runtime. Anthropic ships Claude Code on Bun. Cloudflare's Workers platform runs on workerd.

In September I wrote about DuckDB, whose creators sold their company to Amazon while the license and a separate non-profit foundation stayed where they were. Deno is the version of that story where the people move and the code stays behind, with a public end date for its maintenance.

The clocks, with dates

Deno Deploy in its current form went generally available on February 3, 2026, under the promise "No adapters. No build config. No vendor-specific config files." On October 19, Deno's documentation still carried the guide for moving onto it from Deploy Classic, which was set to shut down, along with the subhosting v1 API, on July 20, 2026. That guide warned that existing KV data was not migrated automatically and that Deno queues were not supported on the new platform.

Teams that moved off Deploy Classic this summer now face a second migration. The announcement gives no calendar date for the shutdown. Six months from October 9 is around April 9, 2027. The runtime's year runs to around October 2027.

The release history shows what "monthly" means. Deno published eight releases on GitHub in the twelve weeks from 2.9.0 on June 25 to 2.9.7 on September 17. As I write on October 19, 2.9.7 is still the latest tag. One release a month, fixes and security only, is maintenance, not development.

The arithmetic of the move

The hosting bill is not the obstacle. Take an app serving 8 million requests a month at 7 milliseconds of CPU each, the per-request figure Cloudflare uses in its own pricing examples.

On Deno Deploy Pro, $20 a month includes 5 million requests, then $2 per million: $20 + 3 × $2 = $26. CPU is 8 million × 7 ms = 56 million ms, about 15.6 hours, inside the 50 hours included, assuming memory time also stays inside its allotment.

On Workers Paid, the $5 monthly minimum includes 10 million requests and 30 million CPU milliseconds, then $0.02 per additional million CPU milliseconds. The CPU above the allowance is 56 million minus 30 million, or 26 million ms, so 26 × $0.02 = $0.52. Total $5.52, and Cloudflare does not charge for egress.

On requests and CPU alone, the destination is cheaper. KV, storage and, on Workers, Durable Objects are metered separately, so treat this as a floor, not a quote. The real cost is code. A request handler written in TypeScript mostly ports. Deno.openKv, Deno.cron and the KV queue methods are Deno APIs; on Workers, the equivalents (KV and Queues bindings, Cron Triggers) are declared in configuration and have their own interfaces. And the migration support is promised to paying customers. The October 9 post says nothing about apps on the free plan, up to 1 million requests a month.

What I would do this month

If I had a Deno app in production, five things, in this order.

Inventory the Deno-specific calls. Search the code for Deno.openKv, Deno.cron, enqueue and listenQueue, the @deno/sandbox SDK and any deployctl script. Each hit is a migration item. Everything else is TypeScript.

Export the KV data while the service still answers. Deno's docs describe opening a managed KV database from a local Deno CLI app with the database ID and an access token in DENO_KV_ACCESS_TOKEN. A short script that lists every key into a file you control beats a migration ticket filed in March. The same page says KV data is stored in the US, with its primary region in Northern Virginia (us-east4), which is worth knowing before a client asks.

Set your own date, earlier than theirs. End of March 2027 for Deploy. For the runtime, decide by spring whether you move to Node, to Bun or to a community fork. If Deno's permission flags were the reason you chose it, Node's permission model was added in v20.0.0 and declared stable in v23.5.0 and v22.13.0; Simon Willison notes that it cannot yet allow-list specific network hosts.

Choose the destination by where the data goes, not by the bill. Workers with Cloudflare's migration help, a self-hosted fleet on celld or workerd, or a generic host running a standard runtime. If you pick Workers, keep celld in mind as the way back out. Varda's own argument is that the open runtime is the escape hatch, and he writes that Cloudflare has former customers who left that way.

Add one question to vendor due diligence. As a CTO, this is the one I would now ask about every runtime and host under a product: who pays the people who maintain it, and do they ship their own product on it? A license tells you the code will survive. The answer to that question tells you whether anyone will be paid to keep it working.

Deno will outlive its company's interest in it. Plan for the year in between.

Sources

A project like this one?

I design and deploy products like this. Let's talk.

Let's talk