Green GPU hosting
Compute, powered by the sun.
A different kind of compute: Renewable and Sustainable. Mother nature approves.
See offeringsSustainable compute
Panels charge during the day. Surplus energy is stored in batteries to carry the workload through the night.
Offerings
Current and upcoming GPU hardware.
NVIDIA RTX 5090
Blackwell fleet for inference, fine-tuning, rendering and batch compute.
- Memory
- 32 GB GDDR7
- Bandwidth
- ~1.8 TB/s
- CUDA cores
- 21,760
- Board power
- 575 W
- CPU
- Intel Core Ultra 9 or AMD Ryzen 9
- RAM
- 64 GB DDR5-6000
- Reliability
- 99.8%
NVIDIA H200 nodes
Hopper nodes for large-model training and high-throughput inference.
- Memory
- 141 GB HBM3e
- Bandwidth
- 4.8 TB/s
- Board power
- up to 700 W
Built for your workloads
From everyday inference to heavy training, on hardware that runs clean.
AI inference
Serve models on always-on GPUs.
Fine-tuning
Adapt models to your own data.
LLM training
Large-model training on upcoming H200 nodes.
Rendering
3D and video rendering workloads.
Simulation & HPC
Memory-bandwidth-hungry scientific compute.
Batch compute
Video processing and other long-running jobs.
About us
- Independent, innovative GPU hosting
- Modern NVIDIA hardware paired with solar generation and battery storage
- Your workloads run on clean power at any hour
- A small team you deal with directly
- Based in South San Francisco, in the heart of the Bay Area and Silicon Valley
- California business entity no. 202253216927, established in 2022
Why Bay Greens
Renewable by design
Powered by solar, not by offsets.
Always on
Battery storage keeps capacity available around the clock.
Multi-gig fiber connected
High-bandwidth fiber connectivity for fast data transfer to and from your GPUs.
Direct access
A small, focused operation. You talk to the people running it.
Get in touch
Questions about capacity, pricing or availability? Send us a note, or email info@bay-greens.com.