11 categories and 65 products and companies, charted by an ai agent.
GPU silicon vendors that design and manufacture processors optimized for datacenter AI model training workloads. Used by cloud providers, research institutions, and enterprises building large-scale language models and neural networks.
nvidia · amd instinct · intel gaudi · huawei ascend · tenstorrent · moore threads · biren technology
google tpu · aws trainium · microsoft maia · meta mtia · groq lpu · sambanova sn40l · broadcom · marvell
High-performance GPU platforms optimized for training large-scale machine learning models in datacenter environments. Used by AI researchers, enterprises, and cloud providers.
nvidia h100 / h200 · nvidia blackwell b200 · nvidia vera rubin nvl72 · amd mi300x · amd mi350x · nvidia dgx b200
Hyperscale cloud GPU providers offer on-demand access to high-performance GPUs hosted in massive datacenters for AI model training at scale. Used by ML researchers, enterprises, and startups needing distributed training without owning hardware.
aws · google cloud · microsoft azure · oracle cloud infrastructure · alibaba cloud · tencent cloud
GPU cloud specialists provide on-demand access to high-performance graphics processors for machine learning training and computational workloads at scale. Used by ML engineers, researchers, and enterprises building and training large AI models.
coreweave · lambda labs · vast.ai · together ai · crusoe energy · voltage park · lambda · crusoe · runpod
nvidia nvlink / nvswitch · nvidia infiniband (mellanox) · broadcom ethernet asics · cisco gpu fabric · ultra ethernet consortium
GPU cluster orchestration manages distributed training workloads across datacenter GPU infrastructure, enabling efficient resource allocation and job scheduling. Used by ML engineers, researchers, and enterprises running large-scale deep learning training.
slurm · kubernetes / kubeflow · run:ai · nvidia base command · determined ai
nebius · scaleway · naver cloud · australia nci gadi
Wafer-scale systems integrate compute, memory, and networking on a single silicon die or tightly-coupled chiplet array for datacenter training workloads. Hyperscalers and AI labs use these to reduce latency and increase throughput in large-scale model training.
cerebras · groq · sambanova · graphcore
Cluster networking and interconnect products connect multiple GPUs across nodes in datacenter training clusters, enabling distributed training workloads. Used by hyperscalers, research labs, and enterprise AI teams running large model training.
arista · astera labs · cisco · celestial ai · ayar labs
Software stacks and frameworks that orchestrate training runs across multiple datacenter GPUs, managing distributed computation, gradient synchronization, and resource allocation. Used by ML engineers and researchers at labs, cloud providers, and enterprises to train large language models and other deep learning systems.
cuda · rocm · pytorch · jax · deepspeed · modular