lsh/metal-GPU
Unlock unprecedented AI performance
Whatever you're building, deploy the GPUs for it
Serve models on bare metal with low, predictable latency and no noisy neighbours.
Start now
Performance your models will be proud of
Features
./gpu-portfolio
A GPU instance for every need
NVIDIA RTX PRO 6000
NVIDIA HGX B300
NVIDIA H100 80GB
GPUs are one piece. We give you the whole .
Storage
Object, File and Block storage. NVMe-backed and high-throughput to keep GPUs fed.
Explore Storage
Networking
High-speed, low-latency networking between nodes for distributed training.
Explore Networking
Kubernetes (LKS)
Orchestrate GPU workloads on managed Kubernetes, on bare metal.
Explore LKS
Bare Metal
Pair GPUs with CPU bare metal for pre- and post-processing in the same environment.
Explore Metal
Build
Custom AI infrastructure designed around your workload.
Latitude for Enterprise
APIs & Terraform
Provision every layer programmatically. REST API, Terraform, SDKs and the lsh CLI.
APIsML-ready images
Deep-learning instances pre-installed with PyTorch, TensorFlow, Keras and CUDA.
Deploy GPUs close to your models and your users
./products
Purpose-built for AI workloads
Choose your hardware
Deploy in seconds
Manage your fleet
Automate everything
Full control of your infra. And your costs.
On-demand
Pay by the hour for GPUs. Ideal for experiments, fine-tunes and bursty inference.
Self-service
Deploy instantly from the dashboard. Pay by card. Or crypto, if you prefer.
Reserved capacity
Lower rates for always-on servers, billed monthly or yearly.
Enterprise
Custom agreements, volume pricing, SSO/SAML and dedicated account support.
Trusted by teams shipping AI in production
“The overall stability, performance, and cost-effectiveness of the platform make it a reliable choice for our infrastructure needs.”
Syed Ghazanfer Anwar
CTO
Frequently asked questions
