# DigitalOcean Droplets

Official pricing: https://www.digitalocean.com/pricing/droplets  
Category: hyperscaler · Isolation: vm

## Pricing regimes (raw)

- **Basic Droplet, Regular CPU (shared vCPU), per-second with 672 h monthly cap** (sizes): s-1vcpu-512mb-10gb (Basic Regular, shared) 1 vCPU/0.5 GiB $0.00595/h; s-1vcpu-1gb (Basic Regular, shared) 1 vCPU/1 GiB $0.00893/h; s-1vcpu-2gb (Basic Regular, shared) 1 vCPU/2 GiB $0.01786/h; s-2vcpu-2gb (Basic Regular, shared) 2 vCPU/2 GiB $0.02679/h; s-2vcpu-4gb (Basic Regular, shared) 2 vCPU/4 GiB $0.03571/h; s-4vcpu-8gb (Basic Regular, shared) 4 vCPU/8 GiB $0.07143/h
- **Basic Droplet, Premium AMD / Intel (shared vCPU, NVMe), per-second with 672 h cap** (sizes): s-1vcpu-1gb-amd (Basic Premium AMD, shared) 1 vCPU/1 GiB $0.01042/h; s-1vcpu-2gb-amd (Basic Premium AMD, shared) 1 vCPU/2 GiB $0.02083/h; s-2vcpu-2gb-amd (Basic Premium AMD, shared) 2 vCPU/2 GiB $0.03125/h; s-2vcpu-4gb-amd (Basic Premium AMD, shared) 2 vCPU/4 GiB $0.04167/h; s-2vcpu-8gb-amd (Basic Premium AMD, shared) 2 vCPU/8 GiB $0.0625/h; s-4vcpu-8gb-amd (Basic Premium AMD, shared) 4 vCPU/8 GiB $0.08333/h
- **Dedicated vCPU (General Purpose / CPU-Optimized / Memory-Optimized / Storage-Optimized, Regular), per-second with 672 h cap** (sizes): g-2vcpu-8gb (General Purpose Regular, dedicated) 2 vCPU/8 GiB $0.09375/h; g-4vcpu-16gb (General Purpose Regular, dedicated) 4 vCPU/16 GiB $0.1875/h; g-8vcpu-32gb (General Purpose Regular, dedicated) 8 vCPU/32 GiB $0.375/h; g-16vcpu-64gb (General Purpose Regular, dedicated) 16 vCPU/64 GiB $0.75/h; g-32vcpu-128gb (General Purpose Regular, dedicated) 32 vCPU/128 GiB $1.5/h; g-40vcpu-160gb (General Purpose Regular, dedicated) 40 vCPU/160 GiB $1.875/h
- **Dedicated vCPU Premium Intel (General Purpose / CPU-Optimized / Memory-Optimized / Storage-Optimized), per-second with 672 h cap** (sizes): g-2vcpu-8gb-intel (General Purpose Premium Intel, dedicated) 2 vCPU/8 GiB $0.1131/h; g-4vcpu-16gb-intel (General Purpose Premium Intel, dedicated) 4 vCPU/16 GiB $0.2247/h; g-8vcpu-32gb-intel (General Purpose Premium Intel, dedicated) 8 vCPU/32 GiB $0.4494/h; g-16vcpu-64gb-intel (General Purpose Premium Intel, dedicated) 16 vCPU/64 GiB $0.9003/h; g-32vcpu-128gb-intel (General Purpose Premium Intel, dedicated) 32 vCPU/128 GiB $1.8006/h; g-48vcpu-192gb-intel (General Purpose Premium Intel, dedicated) 48 vCPU/192 GiB $2.6994/h
- **v5 Shared (5th-gen AMD EPYC, shared vCPU, per-resource, no monthly cap), US only** (resource): $0.0158219/vCPU-h, $0.0040411/GiB-h
- **v5 General Purpose (dedicated vCPU, per-resource, no monthly cap), US only** (resource): $0.028/vCPU-h, $0.0040411/GiB-h
- **Managed Agents / Harness Runtime microVM sandbox (public preview, RIC1 only), per-second, CPU on actual use + allocated RAM** (resource): $0.044/vCPU-h, $0.0095/GiB-h [beta]
- **MicroVMs (public preview, MKC1 only), per-second while running, auto-pause on idle** (resource): $0.022/vCPU-h, $0.0085/GiB-h [beta]

## Features

Yes: snapshots, fork/clone, persistent disk, volumes, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, full VM (own kernel), Docker inside, nested virtualization, root, systemd, SSH, public IPv4, custom domains, raw TCP inbound, egress allowlist, open internet, static egress IP, private networking, SOC 2, HIPAA, SSO, GPU, live fork (no pause), EU data residency, MCP server, usage per sandbox (API), snapshots on demand, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables), extra volumes, shared volumes, FUSE mounts, kernel features (eBPF, mounts, cgroups)

No: memory snapshots, pause/resume, idle auto-stop, your own Docker/OCI image, browser, desktop GUI, computer-use API, code interpreter, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, HTTPS preview URLs, self-hosting, BYOC, open source, arm64, Windows, macOS, Python, Node.js, GPU desktop, credential injection, HTTP method/path egress rules, wake on request, live resize, memory fork, webhooks, spend limits, automatic snapshots, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/digitalocean.json → feature_evidence

## Company

- Ownership: public
- Stage: public [source](https://en.wikipedia.org/wiki/DigitalOcean)
- Accelerators: Techstars 2012 [source](https://www.techstars.com/portfolio)
- Total raised: $223.2M (estimated: Sum of disclosed equity rounds per Wikipedia: $3.2M seed (2013) + $37.2M series A (2014) + $83M series B (2015) + $50M additional raise (2020) = $173.4M equity; plus $50M debt (2014) not typically counted as funding. Figure here ($223.2M) is equity+the 2014 debt facility combined and is only a rough aggregate since not confirmed by a single primary source as a cumulative total; excludes the 2016 $130M credit facility and 2026 $800M stock offering (post-IPO, not private funding).) [source](https://en.wikipedia.org/wiki/DigitalOcean)
- Last round: public (secondary/follow-on stock offering), $800M, 2026-03 (confirmed) [source](https://en.wikipedia.org/wiki/DigitalOcean)
- Investors: Andreessen Horowitz, Access Industries, IA Ventures, Fortress Investment Group, BoxGroup
- Revenue: $692.9M (confirmed: Company-reported figure per Wikipedia infobox citing DigitalOcean's FY2023 earnings/10-K.) [source](https://en.wikipedia.org/wiki/DigitalOcean) annual revenue, 2023

DigitalOcean is a publicly traded company (NYSE: DOCN since March 2021); pricing card products modeled here (Droplets VMs and the Managed Agents/Harness Runtime sandbox preview) are offerings of this single public company, not a separate subsidiary. Revenue figure is company-wide FY2023 ($692.9M); more recent (FY2024/FY2025) figures were not retrieved within the research window because ir.digitalocean.com was behind a Cloudflare bot-challenge.

## Caveats

- Per-second billing (60 s or $0.01 minimum) for CPU Droplets since 2026-01-01; bundled plans are capped at 672 h (28 days)/month, so hour = monthly/672 exactly. Billing runs from create to destroy: powered-off Droplets bill in full. Destroy + recreate from snapshot to stop paying.
- Snapshots $0.06/GB-month (Droplet snapshots are full, not incremental; billed on snapshot size). Recreating a Droplet from a snapshot is the only 'resume' path (fs clone, no memory state).
- Egress: each Droplet's allowance (e.g. 5,000 GiB for s-4vcpu-8gb, 4,000-12,000 GiB across plans) accrues per second (1/2,419,200 of the allowance per second) into a team-wide pool; overage $0.01/GiB. egress_free_gib here is the per-Droplet monthly allowance of the 4 vCPU/8 GiB plans, not an account constant: real pool = sum over Droplets of allowance x hours/672.
- v5 Droplets (US: ATL1, RIC1, MKC1, MEM1 only) have NO monthly cap and bill the boot disk separately ($0.000137/GiB-h); their public IPv4 is a separate line item priced $0.00 'introductory, subject to change'. Bundled plans include IPv4.
- storage.disk_gib_month $0.10 is the Volumes price (and equals the v5 boot-disk rate); bundled sizes include local SSD (disk_gib) so extra disk is only billed when the workload asks for more than the plan's disk.
- No public commitment / reserved discounts for Droplets; volume pricing only via sales. No free tier beyond a $5 signup credit (expires after 90 days).
- Default Droplet limit is published: Tier 1 = 3 Droplets (Basic up to $48 only). 50 concurrent Droplets needs Tier 5 (100 Droplets, $4,000 one-time prepayment for immediate unlock; Tier 4 = 25 for $1,000).
- API provisioning time not published by DigitalOcean ('deploy in seconds'); no measured benchmark captured here.
- Nested virtualization: supported (DigitalOcean's live-migration doc tells users running nested-virtualization workloads to restart nested VMs after a migration; third-party reports of Firecracker/KVM on Droplets). Not a formally documented feature with guarantees.
- GPU mode: vCPU/RAM are fixed per GPU model; two representative shapes are listed with hour=0 and the per-GPU price in the gpu map. RTX 4000/6000 Ada, MI300X/MI325X and spot GPUs have no engine key.
- Prices exclude taxes/VAT (added per country). 2022-07-01 was the last broad Droplet price increase.
- Engine does not implement the $0.01 minimum monetary charge; short sessions can be understated. See validation/engine-edge-cases.json.
- Monthly caps apply per surviving instance ID; destroying and replacing VMs cannot pool caps across replacements.
- 2026-09-28: GPU Droplet mode removed from this card: it duplicated the digitalocean-gpu card (same $4.41 H100 rate) under a wrong RTX/L40S size label. GPU pricing lives only on digitalocean-gpu.
- Boot time not published (docs/droplets/details/limits has only the 10-simultaneous-create cap) and not measured in benchmarks.json; boot_overhead_s left unset. DIY VM: no sandbox API, you install and operate the guest.
- Harness Runtime (Managed Agents, preview since 2026-09-21) is DigitalOcean's agent sandbox product: microVM per session, pause/resume/fork with memory state, prepaid balances only, still RIC1 only (availability table 2026-10-14). Since the 2026-10-13 pricing update CPU bills on actual CPU consumed ($0.044/vCPU-h) and memory on allocation. Consider splitting it into its own card (digitalocean-managed-agents) since Droplet card-level features/storage/egress (snapshot $0.06, 5,000 GiB free egress, fs snapshots, no pause) do not apply to it.
- Bandwidth doc says Droplet seconds are 'rounded to the nearest minute' (while the pricing docs say per-second with a 60 s / $0.01 minimum); effective granularity may be 60 s. Negligible for bills.
- MicroVMs (public preview since 2026-10-09, MKC1 only) are modelled as mode microvms-preview, but card-level storage and network fields belong to Droplets: the engine applies $0.10/GiB-month disk and the 5,000 GiB free egress allowance to it, while MicroVMs bill $0.05/GiB-month on used disk and checkpoints and $0.01/GiB egress with no allowance. Compute is exact; heavy egress or disk is understated / misstated. Deserves its own card (digitalocean-microvms). See .work/engine-needs/digitalocean.md.
- Agent Droplets plans (Pro $50 = 15% off, Team $200 = 20% off, public preview) are modelled as included_usd = fee / (1 - discount) scoped to the Harness Runtime mode. Third-party models are not discounted but still draw from the subscription balance at standard rates; unused balance expires monthly. The plans list concurrency 3 to mirror the team tier; the engine cannot restrict a plan to one mode, so they are never chosen for Droplet workloads only because they add no benefit there.
- GPU types MI350X and HGX B300 are sales-reserved ('Contact sales to reserve capacity'); MI355X is Spot-only in MEM1. See the digitalocean-gpu card for GPU pricing.

## How this provider charges


Digital Ocean sells fixed-size VMs ("Droplets") in bundled plans, plus a new per-resource "v 5" family.
Since **2026-01-01 CPU Droplets are billed per second** (minimum 60 s or $0.01, whichever is higher).
Bundled plans are capped at **672 h (28 days) per month**, so the hourly price is exactly monthly/672
(s-4vcpu-8gb: $48/672 = $0.07143/h). v 5 configurations (US-only, launched 2026-08-25) are per-resource,
per-second and have **no monthly cap**. A Droplet bills from create to destroy, including while powered off.
All prices USD list, excl. tax/VAT. There are no regional price differences for Droplets or bandwidth.

Reference shape 4 vCPU / 8 GiB: Basic Regular `s-4vcpu-8gb` **$0.07143/h, cap $48/mo** (160 GiB SSD, 5,000 GiB transfer, IPv 4 included).

## Regime table

| Regime | When it applies | How billed | Numbers (4 vCPU / 8 GiB unless noted) | Source |
|---|---|---|---|---|
| Basic Regular (shared vCPU) | Default cheapest line, all regions except MEM1 | Per second (60 s / $0.01 min) at monthly/672, capped at the monthly price after 672 h | s-4vcpu-8gb $0.07143/h, cap $48; 2/4 $0.03571/h cap $24; 8/16 $0.14286/h cap $96 (Basic max 8 vCPU) | https://www.digitalocean.com/pricing/droplets, https://docs.digitalocean.com/products/droplets/details/pricing/ |
| Basic Premium AMD / Intel (shared, NVMe) | Newer CPUs, NVMe, up to 10 Gbps | Same as above | s-4vcpu-8gb-amd/-intel $0.08333/h, cap $56; up to 8 vCPU/32 GiB ($0.25-$0.28571/h) | same |
| CPU-Optimized / General Purpose / Memory- / Storage-Optimized (dedicated vCPU, Regular) | Need a dedicated hyperthread | Same per-second + 672 h cap | CPU-Opt c-4 (4/8) $0.125/h cap $84; GP g-4vcpu-16gb $0.1875/h cap $126; CPU-Opt up to 48 vCPU, GP up to 40 (Regular) / 60 (Premium Intel) | same |
| Dedicated Premium Intel | NVMe + 10 Gbps dedicated | Same | c-4-intel $0.1622/h cap $109; g-4vcpu-16gb-intel $0.2247/h cap $151 | same |
| v 5 Shared (per-resource) | US datacenters ATL1, RIC1, MKC1, MEM1 only | Per second, 60 s min, **no monthly cap**; vCPU + RAM + boot disk priced separately | $0.0158219/vCPU-h + $0.0040411/GiB-h + $0.000137/GiB-h disk: 4/8/50 GiB = $0.0956 + $0.00685 = **$0.1025/h**; RAM 1x-3x per vCPU, max 8 vCPU / 24 GB | https://docs.digitalocean.com/products/droplets/details/pricing/, https://docs.digitalocean.com/products/droplets/concepts/choosing-a-plan/ |
| v 5 General Purpose (per-resource, dedicated) | US v 5 datacenters | Same, no cap | $0.028/vCPU-h + $0.0040411/GiB-h + disk; 4/8 = $0.1443/h + disk; 4/16/30 = $0.181/h; 2-64 vCPU, RAM 2x-8x | same |
| Monthly cap (automatic) | Bundled Droplet exists >= 672 h in the month | Charge stops accruing at 672 h | always-on s-4vcpu-8gb = $48/mo (vs $51.43 at 720 h) = ~7% off | https://docs.digitalocean.com/products/droplets/details/pricing/ |
| Powered off | Droplet shut down but not destroyed | **Billed in full** (resources stay reserved) | same as running | same |
| Snapshots | Keep state after destroying | $0.06/GB-month on snapshot size (full images, not incremental; min $0.01); free multi-region | 50 GiB = $3.00/mo | https://docs.digitalocean.com/products/snapshots/details/pricing/ |
| Backups | Automatic, per Droplet | Weekly +20% / daily +30% of Droplet cost, or usage-based $0.04 (weekly) ... $0.01 (4-hourly)/GiB-month | s-4vcpu-8gb weekly = $9.60/mo if on all month | https://docs.digitalocean.com/products/backups/details/pricing/ |
| Volumes | Extra block storage | $0.10/GiB-month while the volume exists (attached or not) | 1 GiB-16 TiB | https://docs.digitalocean.com/products/volumes/details/pricing/ |
| Egress | Outbound public traffic | Per-Droplet allowance accrues per second (allowance/2,419,200 s) into a team pool; overage $0.01/GiB; inbound free | s-4vcpu-8gb contributes 5,000 GiB per 672 h | https://docs.digitalocean.com/platform/billing/bandwidth/ |
| IPv 4 | Public IP | Included in bundled plans; Reserved IPv 4 free while assigned, $5/mo when unassigned; v 5 IPv 4 separate line item at $0.00 intro price | $0 | https://docs.digitalocean.com/products/networking/reserved-ips/details/pricing/ |
| GPU Droplets | GPU workloads | Per second, 60 s min, per-GPU hourly, no cap, billed while off | L40S $1.57, H100 $4.41 (8x $35.28), H200 $4.47, RTX 4000 Ada $0.76, RTX 6000 Ada $1.57, MI300X $2.59, MI325X $3.80; spot preview (MI350X/MI355X/B300) varies daily | https://docs.digitalocean.com/products/droplets/details/pricing/ |
| Commit / reserved | - | **None published** for Droplets; contract GPUs via sales | n/a | pricing page |
| Signup credit | New team | One-time | $5, expires after 90 days | https://docs.digitalocean.com/platform/billing/signup-credit/ |

Dated changes: 2022-07-01 price increase for Droplets, snapshots, reserved IPs and custom images (first in 10 years; $4 512 MiB
Droplet introduced). 2026-01-01 per-second billing with 60 s minimum replaced hourly billing for CPU Droplets.
2026-08-25 v 5 per-resource configurations (no monthly cap) in ATL1/RIC1/MKC1/MEM1. Source: https://docs.digitalocean.com/release-notes/archive/

## Gotchas

1. **Power-off is not a pause.** A stopped Droplet keeps billing at the full rate. For bursty agent work you must destroy and recreate from a snapshot (no memory state; boot + restore time).
2. **672 h cap only helps always-on machines.** A Droplet reaches the cap after 28 days; part-time machines pay hourly.
3. **v 5 has no cap and a separate disk charge.** For always-on use a v 5 Shared 4/8 costs ~$69-$76/mo (720-744 h, + disk) vs $48 for Basic Regular. v 5 IPv 4 is $0 "introductory, subject to change".
4. **Transfer pool is prorated per second.** Short-lived Droplets contribute only a fraction of their allowance; a fleet that exists 176 h/month contributes 26% of the monthly allowance each.
5. **Default Droplet limit is unpublished** and has to be raised via a request for 50 concurrent machines; creates are limited to 10 per request.
6. **Snapshots are full images** (not incremental), so per-Droplet snapshots add up; a 160 GiB disk with 20 GiB used snapshots at roughly used size.
7. **No ARM, no Windows.** Custom images are Linux/Unix only.
8. **Nested virtualization works** (Digital Ocean's own live-migration doc tells users to restart nested VMs after migration) but it is not a formally guaranteed feature; live migration can disrupt nested VMs.

## Worked example

Workload: 4 vCPU / 8 GiB, 50 concurrent x 8 h/day x 22 days = **8,800 instance-hours**, 30% CPU (irrelevant: allocation billing),
50 GiB snapshots retained, 100 GiB egress, 50 GiB disk. Droplets destroyed after each shift and recreated from snapshot.

| Regime | Feasible? | Compute | Snapshots | Egress | Monthly total |
|---|---|---|---|---|---|
| **Basic Regular s-4vcpu-8gb, hourly (per-second)** | yes; 176 h/Droplet < 672 h cap | 8,800 x $0.07143 = $628.58 | $3.00 | $0 (pool ~65,000 GiB) | **$631.58** (cheapest) |
| Basic Premium AMD/Intel | yes | 8,800 x $0.08333 = $733.30 | $3.00 | $0 | $736.30 |
| Dedicated CPU-Optimized c-4 | yes | 8,800 x $0.125 = $1,100.00 | $3.00 | $0 | $1,103.00 |
| Dedicated Premium Intel c-4-intel | yes | 8,800 x $0.1622 = $1,427.36 | $3.00 | $0 | $1,430.36 |
| v 5 Shared 4/8 + 50 GiB boot disk (US only) | yes | 8,800 x $0.0956164 = $841.42 + disk 50 x 8,800/730 x $0.10 = $60.27 | $3.00 | $0 | $904.70 |
| v 5 General Purpose 4/8 (US only) | yes | 8,800 x $0.1443288 = $1,270.09 + disk $60.27 | $3.00 | $0 | $1,333.37 |
| Always-on 50 x s-4vcpu-8gb (kept, not destroyed) | yes | 50 x $48 cap = $2,400.00 | $3.00 | $0 | $2,403.00 |
| Powered off between shifts (not destroyed) | same as always-on | 50 x $48 | | | $2,400+ |
| Commit / reserved | not offered | - | - | - | n/a |

Cheapest: **Basic Regular per-second, destroy between shifts: ~$632/month** ($0.0718 per instance-hour all-in).
Engine check (`PM.priceCard`, cards/digitalocean.json): $631.58 via mode `shared-basic-regular`; always-on 50: $2,403.