The Complete Verda (ex-DataCrunch) Setup & Cluster Guide (2026)

Sign up on verda.com (formerly datacrunch.io) with developer credentials.

• •Navigate to Verda via GPUy's direct referral link (verda.com).
• •Register with your professional or developer email address. Verda serves enterprise ML labs and fast-moving AI startups.
• •Verify your email address via the one-time activation link.

Authorize a valid credit card to unlock high-end Blackwell B200 and H200 clusters.

• •Add a business or personal credit card (Visa/Mastercard) in the Billing tab.
• •Verda places a temporary $1 authorization hold (released automatically) to verify payment legitimacy and prevent crypto-mining abuse.
• •Once authorized, your account is immediately granted On-Demand spinning privileges.

Choose high-throughput European sovereign AI infrastructure with 100% green energy.

• •Click 'Deploy Instance' in the Cloud console.
• •Select your target architecture: NVIDIA Blackwell B300 (268GB SXM), B200 (180GB SXM), or H200 (141GB SXM5).
• •Choose between On-Demand (guaranteed capacity) and Spot (50% discount with interruption notice).
• •Select your operating system: Ubuntu 24.04 LTS with pre-installed NVIDIA CUDA 12.6 drivers.

Connect over dedicated public IPv4 without third-party proxy relays.

• •Verda provides direct public IPv4 addresses for every instance.
• •Connect using your local terminal with root or ubuntu user privileges.

ssh root@YOUR_VERDA_INSTANCE_IP -i ~/.ssh/id_ed25519

Understand Verda's high-speed local NVMe architecture and billing lifecycle.

• •Verda features ultra-fast local NVMe scratch disks (up to 7.68TB) attached directly to the PCIe bus.
• •When your training job finishes, sync weights to cloud storage and delete the instance to stop per-second charges immediately.