# TensorDock

Canonical: https://inetgeek.com/gpu/tensordock/

GPU cloud marketplace where independent hosts set their own hourly prices, selling KVM virtual machines billed per second from a prepaid balance.

## Who it suits

Low-cost GPU servers from a marketplace of independent hosts, billed per second from a $5 deposit with no egress fees; its H100 page lists on-demand H100s from $2.25 an hour and an eight-GPU 3-year rate of $1.50 per GPU-hour.

Consider something else if You need current prices or provider-level compliance before signing up: live prices sit behind sign-in and vary by host, the public price table is dated July 2024, SOC 2 and ISO 27001 belong to the data centres, and servers are deleted when the balance reaches $0.

## iScore

55 / 100 (80% interval 49–61), rank 6 of 13 in gpu — interval overlaps the next rank, so this is an ordering, not a finding.

inetGeek's own reading, computed from sourced facts against category peers by disclosed rules. Not a provider claim. Intervals are 80%; a provider with few documented facts is pulled toward the category middle rather than scored on what little is known.
Method: https://inetgeek.com/iscore/. Full output with every criterion's basis: https://inetgeek.com/scores.json. Every gpu peer ranked: https://inetgeek.com/gpu/tensordock/alternatives/

- Price: 46 (36–56), weight 50%, 2 of 10 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: budget_input_price, cheapest_gpu_price, flagship_input_price, flagship_output_price, free_card_required, free_offer_type, free_tier, storage_price
- Capacity and limits: 50 (36–64), weight 24%, 0 of 3 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: context_window, max_vram, regions
- Capabilities: 100 (94–100), weight 3%, 0 of 1 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: prompt_caching
- Trust and compliance: 73 (67–79), weight 13%, 5 of 12 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: customer_managed_keys, encryption_at_rest, gdpr_data_residency, hipaa_eligible, private_networking, support_included, uptime_sla
- Evidence: 79 (79–79), weight 11%, 4 of 4 criteria scored

## What the documentation says

### Pricing

- Entry paid plan: No plan tiers: deposit funds from as little as $5, drawn down as servers run. ([source](https://www.tensordock.com/cloud-gpus.html), read 2026-09-27)
- Pricing model: A marketplace of independent hosts who set their own prices, so one GPU model can carry several prices. Pay as you go from a deposited balance deducted continuously while a server runs; servers are deleted when the balance reaches $0. ([source](https://www.tensordock.com/), read 2026-09-27)
- Egress over allowance: Free. TensorDock states no ingress or egress fees on its GPU cloud. ([source](https://www.tensordock.com/), read 2026-09-27)

### Infrastructure

- Regions: 100+ locations in over 20 countries, supplied by a marketplace of independent hosts. The public site gives counts rather than a named region list. ([source](https://www.tensordock.com/), read 2026-09-27)
- Uptime SLA: Credit compensation rather than a percentage SLA: downtime is credited at 10x the hourly rate on Top Hosts and 5x on Standard Hosts, minimum $5, capped at 100% of the previous month's spend. Stated as a guide that may not be honoured in special circumstances. ([source](https://docs.tensordock.com/legal-information/downtime-compensation), read 2026-09-27)

### GPU

- H100 SXM, per GPU-hour: From $2.25 per hour for an on-demand H100 SXM5 VM (1 to 8 GPUs), depending on the CPU and RAM allocated; $1.30 as a spot instance. Quoted on TensorDock's undated H100 page; live marketplace prices are shown only after sign-in. ([source](https://www.tensordock.com/gpu-h100.html), read 2026-09-27)
- Billing granularity: Per second, with no minimum; prices are quoted per hour. ([source](https://dashboard.tensordock.com/deploy), read 2026-09-27)
- Reserved pricing: Published only for H100 SXM5 at an 8-GPU minimum: $2.00 an hour per GPU monthly, $1.90 annual and $1.50 on a 3-year term, against $2.25 hourly on demand (about 33% off at three years). ([source](https://www.tensordock.com/gpu-h100.html), read 2026-09-27)

### Compliance

- SOC 2: Data-centre level only: TensorDock says it aggregates GPUs from Tier 3/4 data centres holding certifications including SOC 2. No SOC 2 report for TensorDock itself is claimed on the page. ([source](https://www.tensordock.com/security.html), read 2026-09-27)
- ISO 27001: Data-centre level only: named among the certifications held by the Tier 3/4 data centres TensorDock aggregates GPUs from, not a certification of TensorDock itself. ([source](https://www.tensordock.com/security.html), read 2026-09-27)
- PCI DSS: Data-centre level only: named among the certifications held by the Tier 3/4 data centres TensorDock aggregates GPUs from, not a certification of TensorDock itself. ([source](https://www.tensordock.com/security.html), read 2026-09-27)
- Role-based access control: An IAM system controls who in an organization can access what and when, through the dashboard and API. ([source](https://www.tensordock.com/security.html), read 2026-09-27)
- Dedicated infrastructure: TensorDock's own bare-metal clusters with dedicated routers are offered to customers who need more security and control. ([source](https://www.tensordock.com/security.html), read 2026-09-27)