AI Compute Clusters
GPU clusters, training infrastructure, and AI datacenters tracked across all companies
Total Clusters
24
Operational
16
Deploying
5
Announced
3
GPUs Tracked
552K
2.9M targeted
24 clusters
DeepSeek
Operational
DeepSeek training cluster
H800 (export-control variant)
GPUs
2K
DeepSeek R1 technical report, Jan 2025 — R1 trained on ~2048 H800 GPUs, efficiency-first
Anthropic
Operational
AWS training cluster
H100 / Trainium2
Scale not disclosed
Source ↗Amazon-Anthropic partnership announcement
Baidu
Operational
ERNIE Bot cluster
Ascend 910B + Nvidia A100
Scale not disclosed
Source ↗Baidu AI Day 2024
Alibaba
Operational
Qwen training cluster
Nvidia H800 + Ascend 910B
Scale not disclosed
Source ↗Alibaba Cloud WAIC 2024
Tencent
Operational
Hunyuan training cluster
Nvidia H800 + H100
Scale not disclosed
Source ↗Tencent Cloud Summit 2024
Huawei
Operational
CloudMatrix Ascend cluster
Ascend 910B
Scale not disclosed
Source ↗Huawei CloudMatrix announcement 2024
Mistral AI
Operational
Mistral training cluster
H100 (AWS / Azure)
Scale not disclosed
Source ↗Mistral AI blog, 2024
NTT Group
Operational
tsuzumi LLM cluster
H100 (est.)
Scale not disclosed
Source ↗NTT tsuzumi launch, Mar 2024
Naver
Operational
HyperCLOVA X training cluster
H100 (est.)
Scale not disclosed
Source ↗Naver HyperCLOVA X launch, Aug 2023
Aleph Alpha
Operational
Aleph Alpha PHARIA-1 training
H100 (JUWELS Booster, Forschungszentrum Jülich)
Scale not disclosed
Source ↗Aleph Alpha blog, 2024
Reliance Jio
Deploying
Jio AI-Cloud DGX cluster
NVIDIA DGX H100
Scale not disclosed
Source ↗NVIDIA / Reliance Jio partnership, 2024
GPU counts are disclosed or estimated from public sources. “Target” indicates planned/announced capacity. MW figures where available from company announcements.