Deno Deploy

What Deno Deploy actually is

Deno Deploy is a serverless edge platform that runs JavaScript, TypeScript and WebAssembly in V8 isolates across a global anycast network. It was built by Deno Land Inc., the company Ryan Dahl founded with Bert Belder after creating Node.js — a lineage that matters, because Deno Deploy is the third JavaScript runtime from the same team, and it shows in the design. Instead of containers or VMs, your code executes inside V8 isolates pooled inside lightweight runner processes, which is what makes deploys land in seconds and startup times far below what lambda-style functions manage. The first public beta shipped in the summer of 2021; the company raised a $21 million Series A led by Sequoia in 2022, and its infrastructure now also powers other people's edge functions — Netlify's edge offering runs on Deno Subhosting at a reported 255 million requests per day.

The developer experience is the product. You push TypeScript and it runs — no build pipeline ceremony, no Dockerfile, no region selection. Deno's permission model applies in the cloud: code runs without file-system or network access unless you grant it, which is a genuinely different security posture from Node-based platforms. Modern Deno also carries broad Node.js and npm compatibility, so the old objection that "Deno cannot run my Node app" is much weaker than it was in 2021. Deployments replicate automatically to every edge region, scale to zero when idle, and scale up without configuration. Deno KV, Deno Queues, Sandboxes (isolated microVMs for untrusted code) and Subhosting round out the platform.

Pricing: what you really pay

Deno Deploy publishes one of the clearer price pages in the serverless market. Free is $0 per month with 1 million requests, 20 GiB of egress, 10 hours of active CPU, 150 GiB-hours of memory time, 10 GiB of revision storage, 10 apps and a small Deno KV allowance (1 GiB storage, 1 million reads, 500,000 writes per month). Pro is $20 per month and lifts those to 5 million requests, 200 GiB egress, 50 active CPU hours and 750 GiB-hours of memory, with overages at $2 per million requests, $0.20 per GiB of egress, $0.10 per CPU hour and $0.025 per GiB-hour of memory. Builder at $200 per month is aimed at platforms doing subhosting, with 25 million requests, 2,000 GiB of egress, 500 CPU hours and up to 10,000 apps. Enterprise is custom-priced and is where the 99.95% reliability SLA, SOC 2 Type 1 report, DPA and onboarding support live.

The billing unit deserves attention: you pay for active CPU, not wall-clock time. A request waiting on network I/O does not consume your CPU allowance, and idle apps shut down after roughly 20 to 30 seconds and stop billing for memory entirely. For typical web workloads — API handlers, SSR pages, webhook receivers — this is materially cheaper than per-millisecond billing elsewhere, and for CPU-heavy work (image processing, crypto) it can get expensive quickly. Watch the memory-time meter too: the default 768 MB allocation is billed per second your app is loaded, and raising the limit on Pro or Builder raises the meter. Our read: the free tier is genuinely usable for real side projects, and $20 per month covers more real traffic here than most competitors' starter plans do.

Performance & infrastructure

The architecture is the reason to consider Deno Deploy. Requests land at the nearest of 30-plus edge regions via anycast routing, and every deployment's code is pre-processed and replicated into storage buckets in all regions, so a cold isolate starts from a local copy of your build rather than pulling it from a distant origin. V8 isolates start in milliseconds — Deno's own engineering blog describes cold starts as one of the core design requirements, alongside automatic global replication and autoscaling with zero configuration. We would phrase the cold-start claim as "near-instant" rather than "zero": isolates still spin up, and idle apps do shut down; the difference from lambda platforms is real but it is a difference of degree.

Data locality is genuine. Regional coverage spans North America, Europe (Milan, Madrid and Paris were among the additions as the network grew), Asia and beyond — the network was described as 32 regions at one beta stage and 33 shortly after, with anycast over both IPv4 and IPv6. Deno KV provides strongly consistent storage with read replicas in the same regions, which pairs well with edge compute, and the newer Sandboxes product runs untrusted or generated code in microVMs with block-storage volumes attached. Observability is included on every tier: log streaming, traces and app metrics, with retention ranging from one day on Free to a week on paid plans.

We have not benchmarked Deno Deploy ourselves — our /benchmarks programme measures compute instances, not edge platforms — so treat the latency and cold-start statements as documented architecture, not our measurements.

For teams already in the Deno ecosystem, the day-to-day workflow is the smoothest in the serverless segment: git push deploys from GitHub with preview URLs per commit, the CLI lets you deploy from a terminal in seconds, and the dashboard exposes per-app request, CPU, memory and KV usage without extra instrumentation. Versioned deployments with revision storage mean every deploy is a discrete, retained artifact you can roll back to, and the log streaming API gives you real-time output without shipping logs to a separate service first. None of that is unique in isolation, but the combination — TypeScript-native, permission-scoped, globally replicated, with KV and queues in the same account — is coherent in a way that assembling equivalent pieces elsewhere is not.

Where it falls short

The ecosystem constraint is the first one. Deno has closed much of the Node compatibility gap, but not all of it: native Node modules, packages that assume specific Node behaviours, and tooling built around Node's runtime can still need work. If your app is a standard Express or Next.js deployment, Cloudflare Workers, Vercel or a plain VPS may involve less friction. Second, the platform is young and the company's direction just changed: deno.com currently carries a prominent notice that Deno is joining Cloudflare. For existing users that may be excellent news for infrastructure; for buyers, an acquisition in progress is a variable in the roadmap, the pricing, and the independence of the platform, and we would sign longer commitments with that in mind.

Third, support and guarantees scale with the plan. Free and Pro get community Discord and email support with no SLA at all; the 99.95% reliability SLA is Enterprise-only, which is unusual — most hosting SLAs start lower on the stack. Fourth, running a single-region monolith with long-lived connections, WebSockets at scale, or workloads needing GPUs is out of scope here; this is a request-driven edge platform. Finally, limits are enforced per organisation (builds per hour, concurrent builds, apps) and some, like one concurrent build on Free, are low enough to matter for teams.

On cost discipline, the meters worth watching are memory-time and egress. Memory-time accrues at the default 768 MB allocation whenever your app is loaded, so a workload that keeps isolates warm will pay for that warmth even when CPU idles. Egress at $0.20 per GiB over the included allowance is reasonable but not free, and media-heavy responses will surface it. If your traffic profile is dominated by large downloads or long-lived compute, the pricing advantage over a simple VPS narrows quickly; if it is dominated by many small, fast requests from a global audience, Deno Deploy's model is close to ideal.

Who should pick this

Our pick for TypeScript-first teams shipping globally distributed APIs, web apps and webhook handlers who want near-instant deploys and no infrastructure management — and for platforms that want to embed user-run code via Subhosting, which is battle-tested at serious volume. The pricing model suits request-shaped traffic that is spiky or modest: the free tier is real, the Pro tier is honest, and billing for active CPU rather than wall-clock time rewards efficient code.

We would not choose it for Node-heavy codebases with deep native dependencies, for long-running or connection-oriented workloads, or for teams that need a contractual SLA at the starter tier. If you need a conventional server, a VPS in /category/vps-providers gives you more control per dollar; if you need the largest edge ecosystem, compare Cloudflare Workers first. But for the middle — modern TypeScript services with global audiences — Deno Deploy is the most coherent platform in /category/serverless-platforms today.

Key takeaways

• Deno Deploy runs TypeScript, JavaScript and WASM in V8 isolates across 30-plus anycast edge regions, with automatic global replication and autoscaling to zero.
• Transparent pricing: Free $0/mo (1M requests), Pro $20/mo (5M requests), Builder $200/mo (25M requests), Enterprise custom; overages priced per request, CPU hour, GiB egress and GiB-hour memory.
• Billing counts active CPU, not wall-clock time — efficient request handlers are genuinely cheap here; CPU-heavy code is where bills climb.
• Built by the Deno Land team (Ryan Dahl, Bert Belder) with strong TypeScript ergonomics and Deno's permission-based security model; Node/npm compatibility is broad but not perfect.
• The 99.95% reliability SLA and SOC 2 Type 1 report are Enterprise-only; Free and Pro have no contractual SLA.
• Deno has announced it is joining Cloudflare — factor that into any long-term platform commitment.

Sources

• https://deno.com/deploy/pricing
• https://deno.com/deploy
• https://deno.com/company
• https://deno.com/blog/deno-in-2021
• https://deno.com/blog/anatomy-isolate-cloud
• https://deno.com/blog/build-secure-performant-cloud-platform
• https://deno.com/blog/deploy-beta4
• https://techcrunch.com/2022/06/21/deno-raises-21m-to-launch-a-fully-managed-runtime-service