Menu

DigitalOcean GPU prices: L40S, H100, H200, MI300X per hour

We may earn commissions when you shop through the links below. Learn more

Developer cloud with self-serve per-second GPU Droplets (H100, H200, MI300X, L40S), $4 VPS plans and 1-Click AI agent images.

GPU cloudAI VPSAI agent hostingInference hostingManaged AI hostingPrivate AI hosting

Cloud platform · Headquarters: Broomfield, United States

From published sources, not hands-on tested

From
$1.57≈ €1.40 (estimated conversion from $1.57)/GPU-hrL40S 48GB · On-demandChecked digitalocean.com
GPU families
5
MI300X · H200 · H100 · +2 more · 5 models
Regions
16
2 in the EU
Billing
Per second
Hourly, monthly, spot, reserved, serverless, usage-based

Prices checked 48 sourcesSources (48)How we check

In short

DigitalOcean (founded 2012, headquartered in Broomfield, Colorado) is a developer-focused cloud with 16 datacenters in North America, Europe (Amsterdam, Frankfurt, London), Asia and Australia.

Read full summary

For AI it sells self-serve GPU Droplets (NVIDIA H100, H200, L40S, RTX 4000 and RTX 6000 Ada, AMD MI300X; 1 or 8 GPUs) billed per second, 12-month reserved and spot GPU capacity, sales-only Bare Metal GPUs, an Inference Engine with serverless per-token and dedicated per-GPU-hour endpoints, Knowledge Bases and Managed Agents (public preview). CPU Droplets (VPS) start at $4/month, and the Marketplace has 1-Click images for OpenClaw, Hermes Agent, Ollama, n8n and vLLM. GPU capacity is concentrated in North America: in the EU only H100 GPU Droplets in Amsterdam are listed.

From

$1.57≈ €1.40 (estimated conversion from $1.57)/GPU-hrL40S 48GB · On-demandChecked digitalocean.com
Visit DigitalOcean Opens digitalocean.com in a new tab · Check the final plan, price and terms there before you order.Compare with alternatives to DigitalOcean
Setup guides for DigitalOcean
Billing increment
Per second
Egress
$0.0093≈ €0.0083 (estimated conversion from $0.0093) per GBSource
Persistent storage
Yes
Managed service
Yes

Pricing

Prices checked 31 pricesHow we check

Billing: per second, hourly, monthly, spot, reserved, serverless, usage-based. Lowest pay-as-you-go GPU price: $1.57≈ €1.40 (estimated conversion from $1.57) per GPU-hour (L40S 48GB, on-demand). Lowest plan price: $4.00≈ €3.57 (estimated conversion from $4.00) per month (VPS).

Prices in USD, as listed, excluding VAT.≈ Estimated conversion from the listed USD price at the ECB reference rate of 23 Oct 2026 (1 EUR = 1.1193 USD). Billing currency not verified: providers may bill in USD, and your bank's rate and fees differ. Prices exclude VAT.

GPU offers11

DigitalOcean GPU offers: 11 offers grouped by GPU; prices per GPU-hour, excl. VAT.

AMD Instinct MI300X 192GB192 GB VRAM · 3 offers · from $1.91≈ €1.71 (estimated conversion from $1.91) per GPU-hour

  • Reserved20 vCPU · 240 GB RAM · 720 GB disk$1.91≈ €1.71 (estimated conversion from $1.91)/GPU-hrUS (atl1) · per second · Not verifiedChecked digitalocean.com
  • On-demand20 vCPU · 240 GB RAM · 720 GB disk$2.59≈ €2.31 (estimated conversion from $2.59)/GPU-hrUS (atl1) · per second · AvailableChecked digitalocean.com
  • On-demand8-GPU node · 160 vCPU · 1920 GB RAM · 2046 GB disk$2.59≈ €2.31 (estimated conversion from $2.59)/GPU-hrUS (atl1) · per second · AvailableChecked docs.digitalocean.com

NVIDIA H200 SXM 141GB141 GB VRAM · 3 offers · from $3.40≈ €3.04 (estimated conversion from $3.40) per GPU-hour

  • Reserved8-GPU node · 192 vCPU · 1920 GB RAM · 2046 GB disk$3.40≈ €3.04 (estimated conversion from $3.40)/GPU-hrVaries · per second · Not verifiedChecked digitalocean.com
  • On-demand24 vCPU · 240 GB RAM · 720 GB disk$4.47≈ €3.99 (estimated conversion from $4.47)/GPU-hrUS (nyc2, atl1) · per second · AvailableChecked digitalocean.com
  • On-demand8-GPU node · 192 vCPU · 1920 GB RAM · 2046 GB disk$4.47≈ €3.99 (estimated conversion from $4.47)/GPU-hrUS (nyc2, atl1) · per second · AvailableChecked docs.digitalocean.com
Show all 11 offers (from $1.57≈ €1.40 (estimated conversion from $1.57) per GPU-hour)

NVIDIA H100 SXM 80GB80 GB VRAM · 3 offers · from $3.26≈ €2.91 (estimated conversion from $3.26) per GPU-hour

  • Reserved8-GPU node · 160 vCPU · 1920 GB RAM · 2046 GB disk$3.26≈ €2.91 (estimated conversion from $3.26)/GPU-hrVaries · per second · Not verifiedChecked digitalocean.com
  • On-demand20 vCPU · 240 GB RAM · 720 GB disk$4.41≈ €3.94 (estimated conversion from $4.41)/GPU-hrVaries (nyc2, ams3, tor1) · per second · AvailableChecked digitalocean.com
  • On-demand8-GPU node · 160 vCPU · 1920 GB RAM · 2046 GB disk$4.41≈ €3.94 (estimated conversion from $4.41)/GPU-hrVaries (nyc2, ams3, tor1) · per second · AvailableChecked docs.digitalocean.com

NVIDIA L40S 48GB48 GB VRAM · 1 offer · from $1.57≈ €1.40 (estimated conversion from $1.57) per GPU-hour

Checked digitalocean.com

  • On-demand8 vCPU · 64 GB RAM · 500 GB disk$1.57≈ €1.40 (estimated conversion from $1.57)/GPU-hrVaries (tor1) · per second · Available

NVIDIA RTX 6000 Ada Generation 48GB48 GB VRAM · 1 offer · from $1.57≈ €1.40 (estimated conversion from $1.57) per GPU-hour

Checked digitalocean.com

  • On-demand8 vCPU · 64 GB RAM · 500 GB disk$1.57≈ €1.40 (estimated conversion from $1.57)/GPU-hrVaries (tor1) · per second · Available

Prices are per single GPU-hour, excluding VAT. Multi-GPU nodes are divided by their GPU count; the node size is shown with the price type. Each row shows when the price was observed and links to the page it was read from.

Research notes for these offers11
AMD Instinct MI300X 192GB · Reserved
12-month reserved price $1.91 per GPU-hour (contact sales). Contract slugs exist for 1x (gpu-mi300x1-192gb-contracted), 8x and an 8x multi-node fabric variant; specs shown are the 1x configuration. Contract MI300X is listed in ATL1. Capacity depends on the sales agreement, so availability is shown as Not verified.
AMD Instinct MI300X 192GB · On-demand
gpu-mi300x1-192gb: AMD Instinct MI300X 192 GB at $2.59 per GPU-hour. 20 vCPU, 240 GiB RAM, 720 GiB boot disk plus 5 TiB scratch disk, 15,000 GiB transfer allowance. AMD AI/ML-ready image ships ROCm 7.14. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Atlanta (ATL1) only.
AMD Instinct MI300X 192GB · On-demand · 8-GPU node
gpu-mi300x8-1536gb: 8x MI300X at $20.72 per Droplet-hour ($2.59 per GPU). 160 vCPU, 1,920 GiB RAM, 2,046 GiB boot disk plus 40 TiB scratch disk, 60,000 GiB transfer allowance. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Atlanta (ATL1) only.
NVIDIA H200 SXM 141GB · Reserved · 8-GPU node
12-month reserved price $3.40 per GPU-hour (contact sales). Contract H200 configurations are 8-GPU (gpu-h200x8-1128gb-contracted and a multi-node fabric variant), about $27.20 per 8-GPU Droplet-hour (derived). H200 is not in the contract availability matrix, so datacenters are not verified.
NVIDIA H200 SXM 141GB · On-demand
gpu-h200x1-141gb, sold as 'NVIDIA HGX H200' (141 GB) and listed as the H200 SXM 141GB. $4.47 per GPU-hour. 24 vCPU, 240 GiB RAM, 720 GiB boot disk plus 5 TiB scratch disk, 15 TB transfer allowance. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. NYC2 and ATL1 only.
NVIDIA H200 SXM 141GB · On-demand · 8-GPU node
gpu-h200x8-1128gb: 8x H200 141 GB at $35.76 per Droplet-hour ($4.47 per GPU). 192 vCPU, 1,920 GiB RAM, 2,046 GiB boot disk plus 40 TiB scratch disk, 60 TB transfer allowance. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. NYC2 and ATL1 (availability is published per model).
NVIDIA H100 SXM 80GB · Reserved · 8-GPU node
12-month reserved price $3.26 per GPU-hour on the GPU pricing page ('Reserve via a contractual commitment of 12 months', contact sales, prices subject to change). Contract H100 configurations are 8-GPU only (gpu-h100x8-640gb-contracted; the features page lists the same slug again in its multi-node table, whereas H200 and MI300X have separate '-fabric-contracted' slugs), about $26.08 per 8-GPU Droplet-hour (derived). The docs say contract GPU Droplets are billed per second, but the 12-month commitment applies. The contract availability matrix does not list H100, so datacenters are not verified. Existing contracts keep their rate; renewals move to the 1 Aug 2026 reserved rates.
NVIDIA H100 SXM 80GB · On-demand
Self-serve GPU Droplet gpu-h100x1-80gb, sold as 'NVIDIA HGX H100' (80 GB). HGX is NVIDIA's SXM baseboard and DigitalOcean's bare-metal docs specify HGX H100 SXM5, so it is listed as the H100 SXM 80GB. $4.41 per GPU-hour since the 1 Aug 2026 price change. 20 vCPU, 240 GiB RAM, 720 GiB NVMe boot disk (persistent) plus 5 TiB NVMe scratch disk (non-persistent); transfer allowance 15,000 GiB on the pricing page (docs write 15 TB); 10 Gbps public / 25 Gbps VPC. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Datacenters NYC2, AMS3 (Amsterdam, EU) and TOR1 per the regional availability docs (last verified 24 Sep 2026); the status page's GPU Droplets component does not list AMS3.
NVIDIA H100 SXM 80GB · On-demand · 8-GPU node
gpu-h100x8-640gb: 8x H100 80 GB at $35.28 per Droplet-hour ($4.41 per GPU). 160 vCPU, 1,920 GiB RAM, 2,046 GiB NVMe boot disk plus 40 TiB NVMe scratch disk, 60 TB transfer allowance. The 8-GPU AI/ML-ready image adds NVIDIA Fabric Manager, NVLink and DOCA RoCE packages. Multi-node GPU Droplets must be enabled by support. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Availability is published per GPU model, not per size, so the 8x datacenters are assumed to match the 1x list (NYC2, AMS3, TOR1).
NVIDIA L40S 48GB · On-demand
gpu-l40sx1-48gb: single NVIDIA L40S 48 GB at $1.57 per GPU-hour (single-GPU only). 8 vCPU, 64 GiB RAM, 500 GiB NVMe boot disk, no scratch disk, 10,000 GiB transfer allowance on the pricing page (docs: 10 TB). No NVLink/Fabric Manager support. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Toronto (TOR1, Canada) only.
NVIDIA RTX 6000 Ada Generation 48GB · On-demand
NVIDIA RTX 6000 Ada Generation 48 GB, docs slug gpu-6000adax1-48gb (the pricing page's link uses gpu-rtx6000x1-48gb). $1.57 per GPU-hour, single GPU. 8 vCPU, 64 GiB RAM, 500 GiB boot disk, no scratch disk, 10,000 GiB transfer allowance. Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Toronto (TOR1, Canada) only.

Compare these prices with other providers’: H100 prices comparedH200 prices comparedL40S prices compared.

Plans21

VPS11 plans · from $4.00≈ €3.57 (estimated conversion from $4.00) per month

  • Basic Droplet 1 vCPU / 512 MiB (s-1vcpu-512mb-10gb)1 vCPU · 0.5 GB RAM · 10 GB storage · 0.488 TB traffic · GlobalChecked digitalocean.com$4.00≈ €3.57 (estimated conversion from $4.00)/monthor $0.00595≈ €0.0053 (estimated conversion from $0.00595)/hr
  • Basic Droplet 1 vCPU / 1 GiB (s-1vcpu-1gb)1 vCPU · 1 GB RAM · 25 GB storage · 0.977 TB traffic · GlobalChecked digitalocean.com$6.00≈ €5.36 (estimated conversion from $6.00)/monthor $0.00893≈ €0.008 (estimated conversion from $0.00893)/hr
  • Basic Droplet 1 vCPU / 2 GiB (s-1vcpu-2gb)1 vCPU · 2 GB RAM · 50 GB storage · 1.953 TB traffic · GlobalChecked digitalocean.com$12.00≈ €10.72 (estimated conversion from $12.00)/monthor $0.01786≈ €0.016 (estimated conversion from $0.01786)/hr
  • Basic Droplet 2 vCPU / 2 GiB (s-2vcpu-2gb)2 vCPU · 2 GB RAM · 60 GB storage · 2.93 TB traffic · GlobalChecked digitalocean.com$18.00≈ €16.08 (estimated conversion from $18.00)/monthor $0.02679≈ €0.024 (estimated conversion from $0.02679)/hr
  • Basic Droplet 2 vCPU / 4 GiB (s-2vcpu-4gb)2 vCPU · 4 GB RAM · 80 GB storage · 3.906 TB traffic · GlobalChecked digitalocean.com$24.00≈ €21.44 (estimated conversion from $24.00)/monthor $0.03571≈ €0.032 (estimated conversion from $0.03571)/hr
  • CPU-Optimized Droplet 2 vCPU / 4 GiB (c-2)2 vCPU · 4 GB RAM · 25 GB storage · 3.906 TB traffic · GlobalChecked digitalocean.com$42.00≈ €37.52 (estimated conversion from $42.00)/monthor $0.0625≈ €0.056 (estimated conversion from $0.0625)/hr
  • Basic Droplet 4 vCPU / 8 GiB (s-4vcpu-8gb)4 vCPU · 8 GB RAM · 160 GB storage · 4.883 TB traffic · GlobalChecked digitalocean.com$48.00≈ €42.88 (estimated conversion from $48.00)/monthor $0.07143≈ €0.064 (estimated conversion from $0.07143)/hr
  • General Purpose Droplet 2 vCPU / 8 GiB (g-2vcpu-8gb)2 vCPU · 8 GB RAM · 25 GB storage · 3.906 TB traffic · GlobalChecked digitalocean.com$63.00≈ €56.29 (estimated conversion from $63.00)/monthor $0.09375≈ €0.084 (estimated conversion from $0.09375)/hr
  • Memory-Optimized Droplet 2 vCPU / 16 GiB (m-2vcpu-16gb)2 vCPU · 16 GB RAM · 50 GB storage · 3.906 TB traffic · GlobalChecked digitalocean.com$84.00≈ €75.05 (estimated conversion from $84.00)/monthor $0.125≈ €0.112 (estimated conversion from $0.125)/hr
  • Basic Droplet 8 vCPU / 16 GiB (s-8vcpu-16gb)8 vCPU · 16 GB RAM · 320 GB storage · 5.859 TB traffic · GlobalChecked digitalocean.com$96.00≈ €85.77 (estimated conversion from $96.00)/monthor $0.14286≈ €0.128 (estimated conversion from $0.14286)/hr
  • v5 Droplet (shared) 1 vCPU / 1 GiB / 30 GiB (s5-1vcpu-1gb-30gb)1 vCPU · 1 GB RAM · 30 GB storage · USChecked digitalocean.com$0.024≈ €0.021 (estimated conversion from $0.024)/hr
Show all 21 plans

Managed agent hosting2 plans · from $0.0535≈ €0.048 (estimated conversion from $0.0535) per hour

  • Managed Agents Harness Runtime sandbox XSmall (mars-1vcpu-1gb)1 vCPU · 1 GB RAM · US · ManagedChecked docs.digitalocean.com$0.0535≈ €0.048 (estimated conversion from $0.0535)/hr
  • Managed Agents Harness Runtime sandbox Medium, default (mars-2vcpu-4gb)2 vCPU · 4 GB RAM · US · ManagedChecked docs.digitalocean.com$0.126≈ €0.113 (estimated conversion from $0.126)/hr

Dedicated server6 plans · from $2.59≈ €2.31 (estimated conversion from $2.59) per hour

  • Dedicated Inference endpoint, 1x AMD MI300XVaries · Managed · AMD Instinct MI300X 192GBChecked docs.digitalocean.com$2.59≈ €2.31 (estimated conversion from $2.59)/hr
  • Dedicated Inference endpoint, 1x NVIDIA H100Varies · Managed · NVIDIA H100 SXM 80GBChecked docs.digitalocean.com$4.41≈ €3.94 (estimated conversion from $4.41)/hr
  • Dedicated Inference endpoint, 1x NVIDIA H200Varies · Managed · NVIDIA H200 SXM 141GBChecked docs.digitalocean.com$4.47≈ €3.99 (estimated conversion from $4.47)/hr
  • Dedicated Inference endpoint, 8x AMD MI300XVaries · Managed · AMD Instinct MI300X 192GBChecked docs.digitalocean.com$20.70≈ €18.49 (estimated conversion from $20.70)/hr
  • Dedicated Inference endpoint, 8x NVIDIA H100Varies · Managed · NVIDIA H100 SXM 80GBChecked docs.digitalocean.com$30.32≈ €27.09 (estimated conversion from $30.32)/hr
  • Dedicated Inference endpoint, 8x NVIDIA H200Varies · Managed · NVIDIA H200 SXM 141GBChecked docs.digitalocean.com$35.78≈ €31.97 (estimated conversion from $35.78)/hr

Serverless endpoint1 plan · price not verified

GPU instance1 plan · from $0.76≈ €0.679 (estimated conversion from $0.76) per hour

  • GPU Droplet NVIDIA RTX 4000 Ada Generation 20 GB (gpu-4000adax1-20gb)8 vCPU · 32 GB RAM · 500 GB storage · 9.766 TB traffic · VariesChecked digitalocean.com$0.76≈ €0.679 (estimated conversion from $0.76)/hr
Research notes for these plans21
Basic Droplet 1 vCPU / 512 MiB (s-1vcpu-512mb-10gb)
512 MiB RAM, 10 GiB SSD, 500 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
Basic Droplet 1 vCPU / 1 GiB (s-1vcpu-1gb)
1 GiB RAM, 25 GiB SSD, 1,000 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
Basic Droplet 1 vCPU / 2 GiB (s-1vcpu-2gb)
2 GiB RAM, 50 GiB SSD, 2,000 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
Basic Droplet 2 vCPU / 2 GiB (s-2vcpu-2gb)
2 GiB RAM, 60 GiB SSD, 3,000 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
Basic Droplet 2 vCPU / 4 GiB (s-2vcpu-4gb)
4 GiB RAM, 80 GiB SSD, 4,000 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
CPU-Optimized Droplet 2 vCPU / 4 GiB (c-2)
4 GiB RAM, 25 GiB SSD, 4,000 GiB outbound transfer. Dedicated vCPUs, 'Regular' CPU option (Premium Intel variant priced separately, not listed here). Billed per second (min 60 s or $0.01), capped at 672 hours per month. RAM, SSD and transfer published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Full availability in 12 datacenters, limited in ATL1, RIC1 and MKC1, not offered in MEM1.
Basic Droplet 4 vCPU / 8 GiB (s-4vcpu-8gb)
8 GiB RAM, 160 GiB SSD, 5,000 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
General Purpose Droplet 2 vCPU / 8 GiB (g-2vcpu-8gb)
8 GiB RAM, 25 GiB SSD, 4,000 GiB outbound transfer. Dedicated vCPUs, 'Regular' CPU option (Premium Intel variant priced separately, not listed here). Billed per second (min 60 s or $0.01), capped at 672 hours per month. RAM, SSD and transfer published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Full availability in 12 datacenters, limited in ATL1, RIC1 and MKC1, not offered in MEM1.
Memory-Optimized Droplet 2 vCPU / 16 GiB (m-2vcpu-16gb)
16 GiB RAM, 50 GiB SSD, 4,000 GiB outbound transfer. Dedicated vCPUs, 'Regular' CPU option (Premium Intel variant priced separately, not listed here). Billed per second (min 60 s or $0.01), capped at 672 hours per month. RAM, SSD and transfer published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Full availability in 12 datacenters, limited in ATL1, RIC1 and MKC1, not offered in MEM1.
Basic Droplet 8 vCPU / 16 GiB (s-8vcpu-16gb)
16 GiB RAM, 320 GiB SSD, 6,000 GiB outbound transfer. Shared vCPU, 'Regular' CPU option (the tab shown by default; Premium Intel/AMD variants are priced separately and are not listed here). Billed per second (minimum 60 s or $0.01) and capped at 672 hours per month, so the monthly price is the maximum. Public IPv4 included. RAM, SSD and transfer are published in GiB; the transfer allowance is shown in TiB (GiB divided by 1,024). Basic plans have full availability in 11 datacenters including AMS3, FRA1 and LON1, limited availability in SFO2, ATL1, RIC1 and MKC1, and are not offered in MEM1.
v5 Droplet (shared) 1 vCPU / 1 GiB / 30 GiB (s5-1vcpu-1gb-30gb)
v5 configuration on 5th Gen AMD EPYC. The pricing page shows it as 'Starting at $0.024/hour' for 'Agentic Platforms & AI/ML tooling'. Docs per-resource rates: shared vCPU $0.0158219/h, memory $0.0040411 per GiB-hour, boot disk $0.000137 per GiB-hour (sum about $0.02397/h). No monthly cap: billed per second for actual hours, so a 730-hour month is about $17.50 (derived, not published). Public IPv4 is a separate line item at an introductory $0.00. v5 is offered only in ATL1, RIC1, MKC1 and MEM1. The v5 transfer allowance was not verified, so it is shown as Not verified.
Managed Agents Harness Runtime sandbox XSmall (mars-1vcpu-1gb)
Managed Agents (public preview since 21 Sep 2026): Harness Runtime microVM sandbox for agent harnesses such as Claude Code, Codex, OpenCode, Hermes or LangGraph agents, or your own OCI image. Price shown is per hour at full CPU allocation and peak memory. Billed per second at $0.044 per vCPU-hour (billed on the CPU actually consumed: an agent that is waiting and uses no CPU has no CPU charge; checked 14 Oct 2026) plus $0.0095 per GB-hour of memory; paused sessions incur no compute charge. Session storage $0.05/GiB-month, internet egress $0.01/GiB, Action Gateway calls and model inference are billed separately (pricing overview, DigitalOcean pricing overview page (in Sources): Action Gateway 'Starting at $0.10/1,000 calls', tool search free, code execution and paid tools extra). Prepaid balance required. Harness Runtime runs only in RIC1 (Richmond, US).
Managed Agents Harness Runtime sandbox Medium, default (mars-2vcpu-4gb)
Managed Agents (public preview since 21 Sep 2026): Harness Runtime microVM sandbox for agent harnesses such as Claude Code, Codex, OpenCode, Hermes or LangGraph agents, or your own OCI image. Price shown is per hour at full CPU allocation and peak memory. Billed per second at $0.044 per vCPU-hour (billed on the CPU actually consumed: an agent that is waiting and uses no CPU has no CPU charge; checked 14 Oct 2026) plus $0.0095 per GB-hour of memory; paused sessions incur no compute charge. Session storage $0.05/GiB-month, internet egress $0.01/GiB, Action Gateway calls and model inference are billed separately (pricing overview, DigitalOcean pricing overview page (in Sources): Action Gateway 'Starting at $0.10/1,000 calls', tool search free, code execution and paid tools extra). Prepaid balance required. Harness Runtime runs only in RIC1 (Richmond, US). Other sizes: Small 2 vCPU/2 GB $0.107/h, Large 4 vCPU/8 GB $0.252/h, XLarge 16 vCPU/32 GB $1.008/h.
Dedicated Inference endpoint, 1x AMD MI300X
Managed Dedicated Inference deployment (Inference Engine), billed per GPU-hour; not a VM with root access. Available in NYC2, TOR1, ATL1 and RIC1 (no EU datacenter). Same hourly rate as the 1x MI300X GPU Droplet.
Dedicated Inference endpoint, 1x NVIDIA H100
Managed Dedicated Inference deployment (Inference Engine), billed per GPU-hour; not a VM with root access. Available in NYC2, TOR1, ATL1 and RIC1 (no EU datacenter). The pricing page does not state the GPU variant; shown as the SXM version because DigitalOcean's H100/H200 capacity is sold as HGX (SXM) elsewhere. Same hourly rate as the 1x H100 GPU Droplet. Other Dedicated Inference GPUs not in the catalog: MI325X $2.98/h (8x $23.82/h), MI350X $6.89/h, B300 $10.39/h (8x $83.10/h).
Dedicated Inference endpoint, 1x NVIDIA H200
Managed Dedicated Inference deployment (Inference Engine), billed per GPU-hour; not a VM with root access. Available in NYC2, TOR1, ATL1 and RIC1 (no EU datacenter). The pricing page does not state the GPU variant; shown as the SXM version because DigitalOcean's H100/H200 capacity is sold as HGX (SXM) elsewhere. Same hourly rate as the 1x H200 GPU Droplet.
Dedicated Inference endpoint, 8x AMD MI300X
Managed Dedicated Inference deployment (Inference Engine), billed per GPU-hour; not a VM with root access. Available in NYC2, TOR1, ATL1 and RIC1 (no EU datacenter). $20.70/hour ($2.5875 per GPU) versus $20.72/hour for the 8x MI300X GPU Droplet.
Dedicated Inference endpoint, 8x NVIDIA H100
Managed Dedicated Inference deployment (Inference Engine), billed per GPU-hour; not a VM with root access. Available in NYC2, TOR1, ATL1 and RIC1 (no EU datacenter). The pricing page does not state the GPU variant; shown as the SXM version because DigitalOcean's H100/H200 capacity is sold as HGX (SXM) elsewhere. $30.32/hour is $3.79 per GPU, below the $35.28/hour 8x H100 GPU Droplet rate; DigitalOcean does not explain the difference.
Dedicated Inference endpoint, 8x NVIDIA H200
Managed Dedicated Inference deployment (Inference Engine), billed per GPU-hour; not a VM with root access. Available in NYC2, TOR1, ATL1 and RIC1 (no EU datacenter). The pricing page does not state the GPU variant; shown as the SXM version because DigitalOcean's H100/H200 capacity is sold as HGX (SXM) elsewhere. $35.78/hour ($4.4725 per GPU) versus $35.76/hour for the 8x H200 GPU Droplet.
Serverless Inference (pay per token, prepaid)
OpenAI-compatible serverless inference for 70+ models: DigitalOcean-hosted open models and commercial OpenAI/Anthropic models at provider-aligned rates. Prepaid balance required; access is suspended at $0. Examples, USD per 1M input/output tokens: gpt-oss-120b $0.10/$0.70, gpt-oss-20b $0.05/$0.45, Llama 4 Maverick $0.25/$0.87, DeepSeek V4 Flash $0.14/$0.28, Qwen 3.5 397B A17B $0.55/$3.50, an Anthropic model $3.00/$15.00. Batch inference is up to 50% cheaper on OpenAI and Anthropic models. Web search tool $10 per 1,000 requests and web fetch $3 per 1,000 (not charged with Anthropic models). Bring-your-own model weights: $5.00/month storage. Data handling (DigitalOcean AI Data Privacy (in Sources), updated 1 Sep 2026): DigitalOcean stores no inputs or outputs; DigitalOcean-hosted models run entirely on DigitalOcean infrastructure and prompts are not sent to model creators; OpenAI and Anthropic models are passed through under those providers' zero-retention policies (two Anthropic model versions require 30-day retention). No processing region or datacenter is stated, so data residency remains not verified.
GPU Droplet NVIDIA RTX 4000 Ada Generation 20 GB (gpu-4000adax1-20gb)
The GPU is not in the AIHostCompare catalog (NVIDIA RTX 4000 Ada Generation, 20 GB), so this plan names no catalog GPU and is not listed among the GPU offers. On-demand at $0.76 per GPU-hour, DigitalOcean's lowest GPU Droplet rate. 8 vCPU, 32 GiB RAM, 500 GiB NVMe boot disk, no scratch disk, 10,000 GiB transfer allowance (shown in TiB). Billed per second with a minimum of 60 seconds or $0.01 (docs); the GPU pricing page FAQ says a 5-minute minimum instead. No monthly cap, and a powered-off GPU Droplet is billed until destroyed. Toronto (TOR1) only. The inference-ready image offers only FP8 model variants on this GPU.

Storage and data transfer

Volumes Block Storage (network-attached SSD)
$0.093≈ €0.083 (estimated conversion from $0.093) per GB-month
Block storage · Persistent$0.10 per GiB-month (about $0.0931 per decimal GB), 1 GiB to 16 TiB per volume, accrues hourly while the volume exists, attached or not. Docs: Volume Pricing (docs) (opens in a new tab)
MoreAvailable in all 16 datacenters; usable with Droplets and DOKS. Attachment to GPU Droplets was not separately verified. Published as USD 0.1 per GiB-month; stored per GB (1 GiB = 1.073741824 GB).
Spaces Standard Storage (S3-compatible)
$0.019≈ €0.017 (estimated conversion from $0.019) per GB-month
Object storage · PersistentSubscription $5.00/month includes 250 GiB of storage and 1,024 GiB of outbound transfer shared by all buckets; extra storage $0.02 per GiB-month, extra transfer $0.01/GiB, built-in CDN at no extra cost, inbound free. Docs: Spaces Pricing (docs) (opens in a new tab)
MoreOutbound transfer from Spaces to Droplets is free only for the listed pairs (NYC3 to NYC1/NYC2/NYC3; SFO2 or SFO3 to SFO1-SFO3; AMS3 to AMS2/AMS3; same-datacenter SGP1, FRA1, SYD1, BLR1, ATL1, LON1, TOR1). RIC1 and MKC1 are not listed. Droplets using the VPC-local DNS resolver reach Spaces privately without drawing on the Spaces transfer allowance. Not offered in NYC1, NYC2 or MEM1. Published as USD 0.02 per GiB-month; stored per GB (1 GiB = 1.073741824 GB).
Spaces Cold Storage
$0.0065≈ €0.0058 (estimated conversion from $0.0065) per GB-month
Object storage · Persistent$0.007 per GiB-month plus $0.01 per GiB retrieved (retrieval fee waived up to the average daily Cold Storage usage); 30-day minimum retention with early deletion/update billed at $0.007 per GiB (first 250 GiB per month free); objects under 128 KiB billed as 128 KiB. Docs: Spaces Pricing (docs) (opens in a new tab)
MorePublished as USD 0.007 per GiB-month; stored per GB (1 GiB = 1.073741824 GB).
Network File Storage, Standard tier
$0.14≈ €0.125 (estimated conversion from $0.14) per GB-month
File storage · Persistent$0.15 per provisioned GiB-month, minimum share 50 GiB, maximum 32 TiB; no charge for operations or mounted clients; VPC transfer free; snapshots $0.06/GiB-month. Docs: Network File Storage Pricing (docs) (opens in a new tab)
MoreDatacenters NYC2, AMS3, ATL1, RIC1, MKC1 and MEM1 per the regional availability table (the status page lists NFS components only for ATL1, MKC1 and RIC1). Published as USD 0.15 per GiB-month; stored per GB (1 GiB = 1.073741824 GB).
Network File Storage, High Performance tier
$0.279≈ €0.25 (estimated conversion from $0.279) per GB-month
File storage · Persistent$0.30 per provisioned GiB-month, designed for AI/ML, GPU training and multi-node parallel access; minimum share 500 GiB (so at least $150/month, derived), maximum 32 TiB. Docs: Network File Storage Pricing (docs) (opens in a new tab)
MoreSame datacenters as the Standard tier. Published as USD 0.3 per GiB-month; stored per GB (1 GiB = 1.073741824 GB).
GPU Droplet local NVMe boot disk (included)
Not verified
Included disk · PersistentLocal persistent NVMe boot disk included in the GPU Droplet price: 500 GiB on L40S and RTX 4000/6000 Ada, 720 GiB on 1x H100/H200/MI300X, 2,046 GiB on 8-GPU Droplets. Docs: Droplet Features (docs, updated 2026-09-03) (opens in a new tab)
GPU Droplet local NVMe scratch disk (included, non-persistent)
Not verified
Local NVMe · Not persistentNon-persistent local NVMe scratch disk for staging inference and training data: 5 TiB on 1x H100/H200/MI300X, 40 TiB on 8-GPU Droplets, none on L40S and RTX Ada Droplets. Docs: Droplet Features (docs, updated 2026-09-03) (opens in a new tab)
MoreIncluded in the Droplet price.
Egress (data out)
$0.0093≈ €0.0083 (estimated conversion from $0.0093) per GB
$0.01 per GiB (about $0.0093 per decimal GB) of outbound Droplet transfer beyond the team's pooled allowance; inbound free; no regional variation. Docs: Bandwidth Billing (docs, last verified 14 Sep 2026) (opens in a new tab)
MoreEach Droplet accrues its plan's allowance pro rata (e.g. 500 GiB on the $4 Basic plan, 15,000 GiB on 1x H100/H200/MI300X, 60 TB on 8-GPU Droplets); the pool is team-wide and does not roll over. Traffic within and across VPCs in the same datacenter is free; private VPC egress between datacenters (inter-datacenter VPC peering) costs $0.01/GiB (DigitalOcean pricing overview page (in Sources)). Spaces includes 1,024 GiB per subscription, then $0.01/GiB; App Platform overage is $0.02/GiB; Harness Runtime egress is $0.01/GiB. The control panel does not display network transfer for GPU Droplets. Published as USD 0.01 per GiB; stored per GB (1 GiB = 1.073741824 GB).

Before you compare

From DigitalOcean’s record: its first stated uses and limitation, how its prices are set, and the providers compared below.

Suited to
  • Self-serve 1x or 8x H100, H200 and MI300X VMs billed per second, without a sales call
  • Hosting AI agents (OpenClaw, Hermes Agent, n8n) on low-cost Droplets with 1-Click images
All 4 uses it suits
Main limitation
GPU Droplets are concentrated in North America: the only EU GPU location is H100 in Amsterdam (AMS3), which the status page does not list under GPU Droplets; Frankfurt and London have no GPUs
All 9 limitations
Price basis
GPU prices per GPU-hour, billed per second; egress priced per GB; 7 storage options listed
All prices, storage and egress
Alternatives
OVHcloud provider pageScaleway provider pageTogether AI provider pageFly.io provider page
Compare side by side with DigitalOceanHow they differ

Who it fits

Based on DigitalOcean’s public documentation, checked . We don’t score providers. How we build records

Suited to

  • Self-serve 1x or 8x H100, H200 and MI300X VMs billed per second, without a sales call
  • Hosting AI agents (OpenClaw, Hermes Agent, n8n) on low-cost Droplets with 1-Click images
  • Teams that want serverless LLM APIs, dedicated inference endpoints, managed RAG and VMs on one bill
  • North American inference workloads that need an OpenAI-compatible API or vLLM-ready GPU images

Good fit for

  • Real-time inferenceStrong fitServerless Inference (70+ models, OpenAI-compatible), Dedicated Inference per GPU-hour, vLLM inference-ready GPU images and Hugging Face 1-Click Models. Managed inference endpoints run only in NYC2, TOR1, ATL1 and RIC1, and serverless inference requires a prepaid balance.
  • AI agentsStrong fitOpenClaw, Hermes Agent and n8n Droplet 1-Clicks, Basic Droplets from $4/month with per-second billing in 15 datacenters, and Managed Agents microVM sandboxes (public preview, RIC1 only, prepaid).
  • RAG applicationsStrong fitManaged Knowledge Bases (embedding tokens from $0.009 per 1M), vector database options (Weaviate, OpenSearch, pgvector per the docs index), Spaces object storage and serverless LLMs in one account. Knowledge Base embeddings are stored in a separately billed OpenSearch cluster.
  • TrainingSuitableSelf-serve 8-GPU H100, H200 and MI300X Droplets (NVLink/Fabric Manager on 8-GPU NVIDIA images) plus 12-month reserved pricing and 8-GPU Bare Metal via sales. Multi-node clusters need support enablement or a contract, B200/GB200 are not offered, B300 is spot or contract only, and GPU Droplets carry a 99% SLA.
Show 3 more
  • Fine-tuningSuitableSingle H100 (80 GB), H200 (141 GB) or MI300X (192 GB) Droplets billed per second suit parameter-efficient fine-tuning; the AI/ML-ready images ship the GPU drivers and toolkit (CUDA 13.1 on NVIDIA, ROCm 7.14 on AMD), and frameworks such as PyTorch are installed by the user. Most GPU models sit in one to three North American datacenters, and powered-off GPU Droplets are still billed.
  • Batch inferenceSuitableBatch inference API at up to 50% off OpenAI and Anthropic model rates, per-second GPU Droplets and spot B300/MI350X/MI355X capacity. No spot pricing for H100, H200 or L40S.
  • Private LLM hostingSuitableSelf-hosted models on GPU Droplets (Ollama, vLLM and 1-Click Models), VPC-only Private Droplets, single-tenant Bare Metal GPUs, SOC 2 Type II and a GDPR DPA. EU GPU capacity is limited to H100 in Amsterdam, and serverless inference states no processing region (DigitalOcean-hosted models run on DigitalOcean infrastructure without storing prompts or outputs).

Look elsewhere for

  • RenderingPartial fitRTX 4000 Ada, RTX 6000 Ada and L40S GPU Droplets exist but only in Toronto, and Droplets are Linux-only VMs. DigitalOcean does not market GPU Droplets for rendering.

Not sure DigitalOcean fits what you want to build?Help me choose

Strengths and limitations

Summarised from the research record for DigitalOcean. We don’t score providers.

Strengths

  • Self-serve on-demand GPU Droplets: H100 $4.41, H200 $4.47, MI300X $2.59, L40S and RTX 6000 Ada $1.57, RTX 4000 Ada $0.76 per GPU-hour; H100, H200 and MI300X also as 8-GPU Droplets at the same per-GPU rate
  • Per-second billing (60-second or $0.01 minimum) for CPU and GPU Droplets; bundled CPU plans are capped at 672 hours a month, and Basic Droplets start at $4/month
  • Official Marketplace 1-Clicks for OpenClaw, Hermes Agent, Ollama with Open WebUI, n8n and vLLM inference-ready images
  • Managed AI stack in the same account: Serverless Inference (70+ models), Dedicated Inference, batch inference, Knowledge Bases and Managed Agents (public preview)
Show 3 more
  • Large included transfer (15,000 GiB per 1x H100/H200/MI300X Droplet, 60 TB per 8-GPU Droplet) with $0.01/GiB overage, free inbound traffic and free VPC traffic within a datacenter
  • SOC 2 Type II and SOC 3 reports, a GDPR DPA built into the Terms of Service, and a 99.99% per-Droplet SLA for CPU Droplets
  • Free 24/7 ticket support for all customers; paid plans add faster responses and live chat

Limitations

  • GPU Droplets are concentrated in North America: the only EU GPU location is H100 in Amsterdam (AMS3), which the status page does not list under GPU Droplets; Frankfurt and London have no GPUs
  • The GPU Droplet SLA is 99% per node, versus 99.99% for CPU Droplets
  • Powered-off GPU Droplets are billed until destroyed, GPU Droplets cannot be resized, and multi-node GPU Droplets need support enablement
  • On-demand and reserved GPU prices changed on 1 Aug 2026; reserved rates need a 12-month contract via sales, and Bare Metal GPU pricing is contact-sales only
Show 5 more
  • Spot capacity (public preview) covers only B300, MI350X and MI355X; B200, GB200 and A100 are not offered as GPU Droplets (A100 only via the separate Paperspace console)
  • Dedicated Inference, the Agent Platform and Managed Agents run only in US or Canadian datacenters; Serverless Inference and Managed Agents require a prepaid balance
  • Official pages disagree on details: the GPU pricing FAQ says a 5-minute minimum while the docs say 60 seconds, and MI325X is priced on-demand but listed as contract-only in the docs
  • Some 1-Clicks are vendor-maintained (n8n by automaze.me), and the Ollama image bundles Anaconda under a 'Non-Commercial Use Only' license note
  • No Dify template or DigitalOcean-specific Dify guide was found; Dify needs a manual install

Infrastructure

Regions

16 regions across 8 countries; 2 are recorded as EU regions.

Europe (3)
Amsterdam (AMS3) · Netherlands (EU); Frankfurt (FRA1) · Germany (EU); London (LON1) · United Kingdom
North America (10)
Atlanta (ATL1) · United States; Kansas City (MKC1) · United States; Memphis (MEM1) · United States; New York City (NYC1) · United States; New York City (NYC2) · United States; New York City (NYC3) · United States; Richmond (RIC1) · United States; San Francisco (SFO2) · United States; San Francisco (SFO3) · United States; Toronto (TOR1) · Canada
Asia (2)
Bangalore (BLR1) · India; Singapore (SGP1)
Oceania (1)
Sydney (SYD1) · Australia

Source: Docs: Regional Availability (docs, last verified 24 Sep 2026) (opens in a new tab)

Region notes14
New York City (NYC1)
CPU Droplets, Volumes and managed databases; no GPU Droplets and no Spaces.
New York City (NYC2)
On-demand H100 and H200 GPU Droplets, contract MI325X, DOKS GPU worker nodes, Dedicated Inference and BYOM, Network File Storage. No Spaces.
New York City (NYC3)
CPU Droplets, Spaces, Bare Metal GPUs (via sales); no GPU Droplets.
Amsterdam (AMS3)
Only EU datacenter with GPUs: on-demand H100 GPU Droplets (regional availability docs, last verified 24 Sep 2026, and GPU availability page), Bare Metal GPUs (sales), DOKS GPU worker nodes, Network File Storage. The status page's GPU Droplets component does not list AMS3. No Dedicated Inference, Agent Platform or Harness Runtime. Legacy AMS2 accepts new resources only from existing users.
San Francisco (SFO2)
CPU Droplets (Basic plans limited availability), Spaces; no GPU.
San Francisco (SFO3)
CPU Droplets, Spaces, contract MI325X GPU Droplets (via sales).
Singapore (SGP1), Bangalore (BLR1), Sydney (SYD1)
CPU Droplets, Volumes, Spaces; no GPU.
London (LON1)
Europe but outside the EU (UK GDPR applies). CPU Droplets, Volumes, Spaces; no GPU.
Frankfurt (FRA1)
CPU Droplets, Volumes, Spaces and managed databases; no GPU Droplets and no Network File Storage.
Toronto (TOR1)
On-demand H100, L40S, RTX 4000 Ada and RTX 6000 Ada GPU Droplets, contract MI325X, DOKS GPU worker nodes, Dedicated Inference, BYOM and Agent Platform.
Atlanta (ATL1)
On-demand H200 and MI300X GPU Droplets, spot MI350X, contract MI300X/MI325X/MI350X, Bare Metal GPUs, DOKS GPU worker nodes, Dedicated Inference, BYOM, Agent Platform, v5 Droplets, Network File Storage.
Richmond (RIC1)
Spot B300 (air-cooled) and MI350X, contract MI350X and B300, DOKS GPU worker nodes, Dedicated Inference, BYOM, Agent Platform, Managed Agents Harness Runtime (only datacenter), v5 Droplets, Network File Storage.
Kansas City (MKC1)
Spot B300 (liquid-cooled), contract B300, DOKS GPU worker nodes, v5 Droplets, Network File Storage.
Memphis (MEM1)
Spot MI355X, v5 Droplets only (no bundled Droplet plans), Network File Storage. Not listed on the status page.

Frameworks

How DigitalOcean works with common AI serving, agent and training frameworks. Support levels, strongest first: an official template, documentation, a community guide, or manual setup on a general-purpose machine.

Hosting

Support for running the framework itself on DigitalOcean.

Hermes Agent hosting providers
Official template
Hermes Agent Droplet 1-Click (2026.9.14, Ubuntu 24.04) preconfigured for DigitalOcean Serverless Inference (default model minimax-m2.5). Docs: Hermes Agent Marketplace 1-Click (opens in a new tab)
MoreManaged Agents' Harness Runtime also lists Hermes as a supported harness.
n8n
Official template
n8n Droplet 1-Click on the official Marketplace, built and maintained by third-party vendor automaze.me (Docker Compose + Caddy, latest stable n8n). Docs: n8n Marketplace 1-Click (by automaze.me) (opens in a new tab)
Ollama hosting providers
Official template
'Ollama with Open WebUI' Droplet 1-Click (Ollama 0.10.2, Open WebUI 0.31.2), whose guide calls it DigitalOcean's Ollama Droplet 1-Click. Docs: Ollama with Open WebUI Marketplace 1-Click (opens in a new tab)
MoreIt is a CPU Droplet image; larger models need larger Droplets. It bundles Anaconda 2024.06-1 listed as 'Non-Commercial Use Only'.
OpenClaw hosting providers
Official template
OpenClaw Droplet 1-Click (Staff Pick, listed version 2026.9.4, software table 2026.9.3) described as DigitalOcean's hardened deployment: gateway token, firewall rate limiting, non-root user, Docker sandbox, DM pairing. Docs: OpenClaw Marketplace 1-Click (opens in a new tab)
MoreRecommended 2 CPU/4 GB for 1-5 users; the listing says one-click Droplets start at $12/month.
Text Generation Inference
Official template
Hugging Face 1-Click Models on GPU Droplets expose an OpenAI-compatible endpoint; the docs link the TGI API reference. Docs: 1-Click Models (docs, updated 2024-10-25) (opens in a new tab)
MoreThe docs page dates from 25 Oct 2024 and the model list is older (Llama 3.1/3.2, Mixtral, Gemma 2).
vLLM hosting providers
Official template
'1-Click Inference Ready' single- and multi-GPU images by DigitalOcean AI (v1.1.1: vLLM 0.19.1, CUDA 12.9) for H100/H200 GPU Droplets, also documented as the inference-optimized image (docs list vLLM 0.9.0). Docs: 1-Click Inference Ready - Single GPU (by DigitalOcean AI) (opens in a new tab)
Flowise
Documented setup
Flowise's own docs include a DigitalOcean guide (Docker Compose on a $6/month Basic Droplet, Nginx and Certbot; page last updated about a year ago). No Marketplace 1-Click. Secondary source: Flowise docs: Deploy on Digital Ocean (opens in a new tab)
LangGraph
Documented setup
Managed Agents docs list agents built with LangGraph among supported Harness Runtime workloads (public preview, RIC1 only). No LangGraph 1-Click image. Docs: Managed Agents (public preview, docs) (opens in a new tab)
Dify
Manual setup
No Dify 1-Click in the Marketplace sitemap and no DigitalOcean-specific Dify guide found; self-install with Docker Compose on a Droplet (Docker 1-Click available). Docs: DigitalOcean Marketplace sitemap (opens in a new tab)
PyTorch
Manual setup
DigitalOcean's PyTorch 1-Click for AMD ROCm (PyTorch 2.10.0) is no longer on the Marketplace as of 20 Oct 2026. Docs: Recommended Drivers and Software for GPU Droplets (docs, updated 2026-08-24) (opens in a new tab)
MoreThe AI/ML-ready images ship ROCm 7.14 (AMD) or CUDA toolkit 13.1 and the NVIDIA Container Toolkit (NVIDIA) but no PyTorch, so users install it themselves.
SGLang
Manual setup
DigitalOcean's SGLang 1-Click for AMD ROCm (SGLang 0.5.14) is no longer on the Marketplace as of 20 Oct 2026, and the GPU setup docs name no SGLang image. Docs: Recommended Drivers and Software for GPU Droplets (docs, updated 2026-08-24) (opens in a new tab)
MoreSGLang is installed by the user on a GPU Droplet built from the AI/ML-ready image (ROCm 7.14 on AMD, CUDA toolkit 13.1 on NVIDIA).

Operations

Managed service: yes. Each feature below states what the provider documents, with the source it was checked against. “Not verified” means no reliable source was found.

Features

Compute and access

Custom containers (Docker)
Yes
Docker 1-Click (docker-engine + compose); the inference-optimized GPU image includes Docker, and the NVIDIA AI/ML-ready image ships the NVIDIA Container Toolkit. Docs: Docker Marketplace 1-Click (opens in a new tab)
Root access
Yes
Droplets are unmanaged IaaS VMs; docs and 1-Click guides log in as root over SSH. Not applicable to managed Inference or Managed Agents. Docs: Recommended Drivers and Software for GPU Droplets (docs, updated 2026-08-24) (opens in a new tab)
SSH access
Yes
SSH with root password or SSH key on all Droplets, including GPU Droplets. Pricing page: Droplet (CPU) pricing page (opens in a new tab)
API access
Yes
REST API v2, doctl CLI, official Terraform provider and MCP servers. Docs: DigitalOcean documentation index (opens in a new tab)
Jupyter notebooks
Partial
The Marketplace lists a JupyterLab 1-Click for AMD Instinct MI300X GPU Droplets and a Jupyter Notebook 1-Click; no managed notebook service in the DigitalOcean console (Paperspace Notebooks is a separate product). Docs: DigitalOcean Marketplace sitemap (opens in a new tab)
Templates
Yes
Marketplace 1-Clicks incl. OpenClaw, Hermes Agent, Ollama with Open WebUI, n8n and vLLM inference-ready images, plus Hugging Face 1-Click Models. Docs: DigitalOcean Marketplace sitemap (opens in a new tab)
Multi-GPU instances
Yes
8-GPU Droplets for H100, H200 and MI300X; multi-node (fabric) configurations need support enablement or a contract; Bare Metal GPUs have 8 GPUs. Docs: Droplet Features (docs, updated 2026-09-03) (opens in a new tab)
Bare metal
Yes
Single-tenant 8x H100, H200 or MI300X Bare Metal GPUs in NYC3, AMS3 and ATL1; pricing only via sales. Docs: Bare Metal GPUs Details (docs) (opens in a new tab)
Kubernetes
Yes
DOKS managed Kubernetes with free control plane (HA $40/month); GPU worker nodes billed per second at GPU Droplet prices, available in NYC2, AMS3, TOR1, ATL1, RIC1 and MKC1. Docs: Kubernetes (DOKS) Pricing (docs, updated 2026-09-14) (opens in a new tab)
Show all 21 features

Management

Managed service
Yes
Managed Serverless and Dedicated Inference, Knowledge Bases, managed Kubernetes and databases, and Managed Agents (public preview). Droplets themselves are unmanaged. Docs: Inference Pricing (docs, updated 2026-09-25) (opens in a new tab)
Serverless
Partial
Serverless Inference (per-token, OpenAI-compatible, prepaid) and DigitalOcean Functions (CPU). No serverless GPU containers for custom code; Dedicated Inference is billed per GPU-hour. Docs: Inference Pricing (docs, updated 2026-09-25) (opens in a new tab)

Storage and data

Persistent storage
Yes
Volumes block storage ($0.10/GiB-month), Network File Storage, Spaces object storage and persistent local boot disks; GPU scratch disks are non-persistent. Docs: Volume Pricing (docs) (opens in a new tab)
Snapshots
Yes
Droplet snapshots $0.06/GB-month; docs recommend snapshotting a configured GPU Droplet as a base image. Volume and NFS snapshots available. Pricing page: Droplet (CPU) pricing page (opens in a new tab)
Managed backups
Yes
Droplet backups at 20% (weekly) or 30% (daily) of the Droplet price, or usage-based from $0.01/GiB-month with custom frequency and retention; available in all 16 datacenters. Pricing page: Droplet (CPU) pricing page (opens in a new tab)
MoreBackup support for GPU Droplets specifically was not verified.

Billing

Per-second billing
Yes
CPU and GPU Droplets billed per second with a minimum of 60 s or $0.01 (the GPU pricing page FAQ says a 5-minute minimum). Managed Agents are also billed per second. Docs: Droplet Pricing (docs, last verified 25 Aug 2026) (opens in a new tab)
Spot pricing
Yes
Spot GPU Droplets in public preview for B300, MI350X and MI355X only (none in the GPU catalog); price locked at creation, reclaim with at least two hours' notice aimed for. Docs: Droplet Pricing (docs, last verified 25 Aug 2026) (opens in a new tab)
MoreSpot preview terms (Public Preview Terms: Spot GPU Droplets and Node Pools (in Sources), updated 9 Sep 2026): the 2-hour notice is a best-effort target and 'not a service commitment', emergency reclamation can be immediate, no SLA or service credits apply, and boot/scratch disk data is lost on reclamation. The terms say the Spot rate is 'not a discount applied to' any on-demand rate, while the pricing docs say it is always equal to or less than on-demand.
Reserved pricing
Yes
12-month reserved GPU prices via sales (e.g. H100 $3.26, H200 $3.40, MI300X $1.91 per GPU-hour; 'Reserve via a contractual commitment of 12 months', contact sales). Pricing page: DigitalOcean GPU Droplets pricing page (opens in a new tab)
MoreThe pricing overview (DigitalOcean pricing overview page (in Sources)) advertises GPU Droplets 'Starting at $1.91/GPU/hour' 'based on multi-month contractual commitment'. No savings-plan or other commitment discount is published; prepayment is by sales contact only.

Regions and compliance

EU region
Yes
Amsterdam (AMS3) and Frankfurt (FRA1); London (LON1) is outside the EU. GPU Droplets in the EU: H100 in AMS3 only. Docs: Regional Availability (docs, last verified 24 Sep 2026) (opens in a new tab)
SOC 2 report
Yes
SOC 2 Type II and SOC 3 Type II reports from Schellman & Company; SOC 2 report available via the Security Reports center. Docs: Trust Platform: Certification Reports and Resources (opens in a new tab)
GDPR data processing agreement
Yes
DPA (updated 6 Feb 2026) is an addendum to the Terms of Service covering GDPR, with EU-U.S. DPF and SCCs/UK Addendum for transfers. Docs: Data Processing Agreement (updated 6 Feb 2026) (opens in a new tab)
Customer-managed keys (BYOK)
Partial
Spaces object storage supports SSE-C (server-side encryption with customer-provided keys) per object; bucket-level encryption settings are not supported. Docs: Spaces S3 Compatibility (docs) (opens in a new tab)
MoreWhether Volumes, Droplets, GPU Droplets or NFS offer a KMS-style customer-managed key option is not verified. API keys for model providers are a separate feature: Inference accepts your own OpenAI/Anthropic keys (billed by those providers) and Harness Runtime accepts external provider keys (Inference Pricing (in Sources), Harness Runtime Pricing (in Sources)).

SLA and support

Uptime SLA
99% uptime · GPU Droplets only
GPU Droplet SLA (updated 21 Oct 2025): 99% monthly uptime per server node, measured as 5-minute intervals without external connectivity; credits of 10% (below 99%), 25% (below 95%) and 50% (below 90%) of the node's monthly bill as future GPU Droplet credit; downtime must be reported within 24 hours and credits requested within 3 months. Docs: DigitalOcean GPU Droplet SLA (updated 21 Oct 2025) (opens in a new tab)
MoreCPU Droplets have a separate 99.99% per-Droplet SLA (updated 3 Jun 2025) with a 100% credit for the affected Droplet (source DigitalOcean CPU Droplet SLA (in Sources)). Other products have their own SLAs (not verified).
Support channels
6 listed
  • support tickets (24/7, all customers)
  • email
  • live chat (Standard and Premium plans)
  • dedicated Slack channel (Premium)
  • video calls (Premium)
  • community tutorials and Q&A
Free Starter support (<24 h response); Developer $24/month (<8 h); Standard $99/month (<2 h, 24/7 live chat, senior staff); Premium $999/month (<30 min, Slack with a Solution Architect in business hours, video calls). Docs: Support Plans (docs, updated 2026-07-13) (opens in a new tab)
MoreOfficial support language is English. Vendor 1-Clicks are supported by their vendors (e.g. AMD Developer Cloud, automaze.me).

Sources and verification

48 sources, accessed between 26 Sep 2026 and 20 Oct 2026. Official provider pages are listed first; the 1 secondary source is labelled as such.

How we verify

Prices, regions, storage, features, framework support, SLA and support channels link to the page they were read from, and every price shows when it was checked. Fit labels, strengths and limitations summarise the same research record. The DigitalOcean record was last checked on 26 Sep 2026. A check date older than 7 days is marked “Stale” with an amber dot, and one older than 30 days “Needs refresh”.

Last verified
26 Sep 2026, 19:24 UTC
Checked · Stale
Method
Source-by-source review by AI agents, independently re-checked
Confidence
High
Checked by
AI research agent; every value re-checked against its source by a separate AI agent

Verification notes

On 26 September 2026 we checked every price and spec here against DigitalOcean's live pricing pages, documentation and status page. Its pages disagree on the GPU billing minimum (60 seconds in the docs, 5 minutes in the pricing FAQ) and on H100 in Amsterdam (on the availability pages, not the status page); the listing follows the docs. Not verified: live GPU stock, reserved H100 and H200 locations and the serverless inference region. Later corrections are listed on the methodology page.

Commercial relationship

The Visit provider link is an affiliate link: AIHostCompare may earn a commission if you sign up with DigitalOcean through it. Commissions never change ordering, alternatives, fit labels or conclusions. Read the affiliate disclosure.

Pricing

  1. 1.DigitalOcean GPU Droplets pricing page (opens in a new tab) · Official pricing page · accessed
  2. 2.DigitalOcean pricing overview page (opens in a new tab) · Official pricing page · accessed
  3. 3.Droplet (CPU) pricing page (opens in a new tab) · Official pricing page · accessed

Documentation

  1. 4.1-Click Inference Ready - Single GPU (by DigitalOcean AI) (opens in a new tab) · Official docs · accessed
  2. 5.1-Click Models (docs, updated 2024-10-25) (opens in a new tab) · Official docs · accessed
  3. 6.1-Click Models product page (opens in a new tab) · Official docs · accessed
  4. 7.About DigitalOcean (opens in a new tab) · Official docs · accessed
  5. 8.Bandwidth Billing (docs, last verified 14 Sep 2026) (opens in a new tab) · Official docs · accessed
  6. 9.Bare Metal GPU Features (docs) (opens in a new tab) · Official docs · accessed
  7. 10.Bare Metal GPUs Details (docs) (opens in a new tab) · Official docs · accessed
Show all 48 sources

Documentation

  1. 11.DigitalOcean Affiliate Program (opens in a new tab) · Official docs · accessed
  2. 12.DigitalOcean documentation index (opens in a new tab) · Official docs · accessed
  3. 13.DigitalOcean Marketplace sitemap (opens in a new tab) · Official docs · accessed
  4. 14.DigitalOcean Referral Program (opens in a new tab) · Official docs · accessed
  5. 15.Docker Marketplace 1-Click (opens in a new tab) · Official docs · accessed
  6. 16.Droplet Features (docs, updated 2026-09-03) (opens in a new tab) · Official docs · accessed
  7. 17.Droplet Limits (docs, updated 2026-08-19) (opens in a new tab) · Official docs · accessed
  8. 18.Droplet Pricing (docs, last verified 25 Aug 2026) (opens in a new tab) · Official docs · accessed
  9. 19.GPU Droplet Availability by Region (docs, updated 2026-08-10) (opens in a new tab) · Official docs · accessed
  10. 20.Harness Runtime Availability (docs) (opens in a new tab) · Official docs · accessed
  11. 21.Harness Runtime Pricing (docs, updated 2026-09-22) (opens in a new tab) · Official docs · accessed
  12. 22.Hermes Agent Marketplace 1-Click (opens in a new tab) · Official docs · accessed
  13. 23.Inference Availability (docs) (opens in a new tab) · Official docs · accessed
  14. 24.Inference Details (docs) (opens in a new tab) · Official docs · accessed
  15. 25.Inference Pricing (docs, updated 2026-09-25) (opens in a new tab) · Official docs · accessed
  16. 26.Kubernetes (DOKS) Pricing (docs, updated 2026-09-14) (opens in a new tab) · Official docs · accessed
  17. 27.Managed Agents (public preview, docs) (opens in a new tab) · Official docs · accessed
  18. 28.n8n Marketplace 1-Click (by automaze.me) (opens in a new tab) · Official docs · accessed
  19. 29.Network File Storage Pricing (docs) (opens in a new tab) · Official docs · accessed
  20. 30.Ollama with Open WebUI Marketplace 1-Click (opens in a new tab) · Official docs · accessed
  21. 31.OpenClaw Marketplace 1-Click (opens in a new tab) · Official docs · accessed
  22. 32.Paperspace Pricing (docs) (opens in a new tab) · Official docs · accessed
  23. 33.Recommended Drivers and Software for GPU Droplets (docs, updated 2026-08-24) (opens in a new tab) · Official docs · accessed
  24. 34.Regional Availability (docs, last verified 24 Sep 2026) (opens in a new tab) · Official docs · accessed
  25. 35.Spaces Pricing (docs) (opens in a new tab) · Official docs · accessed
  26. 36.Spaces S3 Compatibility (docs) (opens in a new tab) · Official docs · accessed
  27. 37.Support Plans (docs, updated 2026-07-13) (opens in a new tab) · Official docs · accessed
  28. 38.Trust Platform: Certification Reports and Resources (opens in a new tab) · Official docs · accessed
  29. 39.Upcoming GPU Pricing Updates (blog, 21 Jul 2026) (opens in a new tab) · Official docs · accessed
  30. 40.Volume Pricing (docs) (opens in a new tab) · Official docs · accessed

Terms & policies

  1. 41.Affiliate Program Agreement (last updated 31 Dec 2024) (opens in a new tab) · Official docs · accessed
  2. 42.Data Processing Agreement (updated 6 Feb 2026) (opens in a new tab) · Official docs · accessed
  3. 43.DigitalOcean AI Data Privacy (docs, updated 2026-09-01) (opens in a new tab) · Official docs · accessed
  4. 44.DigitalOcean CPU Droplet SLA (updated 3 Jun 2025) (opens in a new tab) · Official docs · accessed
  5. 45.DigitalOcean GPU Droplet SLA (updated 21 Oct 2025) (opens in a new tab) · Official docs · accessed
  6. 46.Public Preview Terms: Spot GPU Droplets and Node Pools (updated 9 Sep 2026) (opens in a new tab) · Official docs · accessed

Status

  1. 47.DigitalOcean Status (opens in a new tab) · Official status page · accessed

Secondary sources

  1. 48.Flowise docs: Deploy on Digital Ocean (opens in a new tab) · Secondary source · accessed
DigitalOcean

Alternatives to DigitalOcean

Providers that share a hosting category with DigitalOcean, the most relevant first, spread so that every provider is listed on several provider pages. Not a ranking.

Also sharing categories: Fly.io profile

Compare all 4 side by side

Related comparisons and guides

Categories that list DigitalOcean