GPU nodes at curtailed electricity

By connecting portable GPU nodes to solar and wind assets which suffer from frequent curtailment, compute can be unlocked at a much faster rate than mega-campus datacenters.

Ambra Node

A self-contained AI compute cluster engineered for behind-the-meter deployment at renewable generation sites. Built with enterprise grade AI servers, the system delivers data-center-grade inference and fine-tuning capacity in a footprint sized for direct connection to a service transformer — without the evaporative cooling typical of larger facilities.

Compute

  • 64× NVIDIA RTX PRO 6000 Blackwell Server Edition — 6,144 GB aggregate GPU memory (96 GB GDDR7 each), Gen5 x16 per card

  • 16× AMD EPYC 9375F — 512 cores / 1,024 threads at up to 4.8 GHz boost, optimized for GPU feeding

  • 6 TB DDR5 ECC @ 6400 MT/s (768 GB per node, 96 DIMMs total)

  • 61 TB Gen5 NVMe local scratch for dataset staging (16× 3.84 TB U.2)

  • 7.68 TB NVMe boot/OS (8× 960 GB M.2)7

Power & Footprint

  • ~100 kW sustained

  • 230V input

  • 1.2 m × 1.2 m floor footprint

  • 2 m tall

  • 800 kg

sunsetsolar.jpg

Run your AI inference compute from otherwise wasted renewable energy.

  • We're a energy-first team dedicated to solving AI demand compute responsibly and securely.

  • We deploy Ambra Compute Nodes at renewable energy sites subject to frequent curtailment.

  • We work with leading asset managers and IPPs to rapidly deploy and scale solutions across geographies.

Contact Us


Submit the inquiry form to learn more about Ambra