# Verda (formerly DataCrunch)

Official pricing: https://verda.com/pricing  
Category: gpu-cloud · Isolation: vm

## Pricing regimes (raw)

- **1x H100 SXM5 80GB on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **1x H100 SXM5 80GB spot** (resource): $0/vCPU-h, $0/GiB-h [spot]
- **1x H200 SXM5 141GB on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **1x H200 SXM5 141GB spot** (resource): $0/vCPU-h, $0/GiB-h [spot]
- **1x B200 SXM6 180GB on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **1x B200 SXM6 180GB spot** (resource): $0/vCPU-h, $0/GiB-h [spot]
- **1x B300 SXM6 268GB on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **1x B300 SXM6 268GB spot** (resource): $0/vCPU-h, $0/GiB-h [spot]

## Features

Yes: snapshots, fork/clone, persistent disk, volumes, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, full VM (own kernel), Docker inside, root, SSH, public IPv4, HTTPS preview URLs, static egress IP, SOC 2, GPU, Python, wake on request, audit logs, EU data residency, spend limits, MCP server, snapshots on demand, gVisor or VM (no shared kernel), firewall inside (nftables), extra volumes, shared volumes

No: memory snapshots, pause/resume, idle auto-stop, browser, desktop GUI, computer-use API, code interpreter, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, custom domains, raw TCP inbound, egress allowlist, private networking, self-hosting, BYOC, open source, HIPAA, SSO, arm64, Windows, macOS, Node.js, GPU desktop, HTTP method/path egress rules, live resize, live fork (no pause), memory fork, automatic snapshots, agent harness API, hosted agent API (their own agent), inbound access rules, secret proxy, secret proxy for any API

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

## Caveats

- Verda is the renamed DataCrunch; the locations page lists only FIN-01, FIN-02, FIN-03 (all Helsinki, Finland).
- NVMe block storage $0.20/GiB-month; a boot volume must be added (no included disk).
- Reserved discounts (2-25%) not encoded.
- Pay-as-you-go is charged in pre-paid 10-minute increments; if the instance is terminated early the unused part of the block is refunded in the next billing period (effective granularity not published; min_billed_seconds 60 remains an assumption).
- A shut-down (stopped) instance keeps billing compute at the full hourly rate; only Delete stops the instance charge. Volumes not selected for deletion keep billing at $0.20/GiB-month, including OS volumes of evicted spot instances (default keep_detached) and volumes in trash (restore charged at PAYG for the deleted time).
- Confidential CPU instances (AMD SEV-SNP, GA Aug 2026) use the same shapes as CPU nodes at +2% ($0.01224 per vCPU-h incl. 4 GB RAM; CPU.4V.16G.CC $0.04896/h, spot $0.01958/h). Not a separate mode because it is never cheaper than on-demand-cpu.
- Account quotas on GPUs, spot GPUs, total vCPUs, NVMe storage, storage item count and projects; defaults unpublished, raised via console request (~1 business day). API: 500 requests/min per project, 60 requests/min per endpoint (<=60 instance creations/min).

## How this provider charges


GPU cloud. Added by the missing-providers audit (single-H100 SKUs).

| Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|
| On-demand | default | per GPU-hour | 1x H100 SXM5 $3.56 (30 CPU/120 GB), H200 $4.64, A100 80GB $1.80 | https://verda.com/pricing |
| Spot | spot | per GPU-hour | H100 $1.78, H200 $2.32 | https://verda.com/pricing |
| Reserved | 1m..2y | discount on on-demand | 2%..25% | https://verda.com/pricing |
| Storage | always | per GiB-month | $0.20 | https://verda.com/pricing |

## Gotchas
- No included disk: boot volume billed at $0.20/GiB-month.

## Worked example
1x H100: 250 h x $3.56 = $890 (+100 GiB x $0.20 = $20).

Sources: https://verda.com/pricing