cluster/

Shaping the future of AI development

Free compute infrastructure for communities, educators, and developers building AI that matters

Apply for a Grant

Powering the Mission

Peak AI Compute1.5+ PFLOPS

Total VRAM Pool215 GB

Status

Active / Optimized

01/

The Cluster

The ECO-Foundry is an 9-node hybrid compute environment spanning four NVIDIA architecture generations. Energy-conscious and always-on, it provides free inference, fine-tuning, and deployment capacity for community partners across 279 GB of combined GPU memory.

Node	GPU	VRAM	Architecture
FeynmanFlagship	GB10 Grace Blackwell Superchip	128 GB unified LPDDR5x	Blackwell + Grace
Laplace	RTX 3080 Ti	12 GB GDDR6X	Ampere
Euler	RTX 3080 Ti	12 GB GDDR6X	Ampere
Riemann	RTX 3080 Ti	12 GB GDDR6X	Ampere
Noether	2x TITAN Xp	2x 12 GB GDDR5X (24 GB total)	Pascal
Dirac	RTX 2080 Ti	11 GB GDDR6	Turing
Gauss	GTX 1080	8 GB GDDR5X	Pascal
Planck	GTX 1080	8 GB GDDR5X	Pascal
Turing	M4 Max 40-core GPU	64 GB unified LPDDR5x	Apple Silicon

02/

Cloud Tier

Beyond the local cluster, Epoch operates Exo-Scale: elastic cloud GPU capacity for community projects that need more scale than on-premise VRAM can provide.

Cloud Tier

On-demand access to H100, H200, A100, B200, and GB300 NVL72 instances via the cloud. Elastic GPU capacity for community projects that need distributed training or more memory than the local cluster can supply. No egress fees. NVIDIA InfiniBand on clusters.

View full specifications→

Epoch Inference Lab

Los Angeles