Buy data-centre AI compute
Equip a data centre to train and serve models — accelerators bought as chips, cards, servers or rack systems, or rented as dedicated cloud capacity. 149 products across 50 companies in the Ludbee catalogue carry this task — spanning every layer of the stack that exists here, not one of its own: 5 Application, 14 Infrastructure, 7 Foundation Models, 24 Hardware.
Filter the whole directory by this job to sort and compare these products.
Where these run
- Self-hosted
- 71 of 137 products that record it
- SaaS
- 48 of 137 products that record it
- On-device
- 22 of 137 products that record it
- Private cloud
- 16 of 137 products that record it
- Hybrid
- 6 of 137 products that record it
- Air-gapped
- 1 of 137 products that record it
12 of 149 products carry no recorded deployment yet, and one product can run more than one way — the counts above are of records, not of products.
What this page does not claim
This list carries no score and no recommendation, and one order only: inside each layer, a company whose product for this task is marked AI feature — sold for another job, with AI as one function its own page names — comes after the rest, A–Z within each. The standing rule that applies to the whole directory applies here too: no payment changes a figure, a ranking or the order of any list, this one included. Nor can it yet say what governs any of these products once deployed — what data they can reach, what they are certified against, whether a vendor trains on what you send it. That field is defined in the ontology and is populated on none of the 3,888 products in the catalogue yet, so a page that leaned on it would be promising an answer the data cannot back. It will say so here the day that changes.
The companies
Application (5)
- 4Paradigm — 4Paradigm SageOne, 4Paradigm SageOne Advanced
- Amazon — AWS Inferentia, AWS Trainium
- Oracle — OCI Compute for GPUs
- SK Telecom — SKT GPUaaS
- SpaceX — Starmind
Infrastructure (14)
- Core42 — Core42 Signature Private Cloud, Core42 Sovereign AI Cloud, Core42 Sovereign Public Cloud
- CoreWeave — CoreWeave AI Cloud, CoreWeave Kubernetes Service, CoreWeave Sandboxes
- Crusoe — Crusoe Cloud, Crusoe Edge Zones, Crusoe Spark
- Databricks — AI Runtime
- Fal — fal Compute, fal Serverless
- Huawei Cloud — GPU Accelerated Cloud Server
- Kakao Enterprise — AI Insight
- Lambda — Lambda 1-Click Clusters, Lambda Cloud, Lambda On-Demand Instances, Lambda Superclusters
- Modal — Modal
- NAVER Cloud — ML expert Platform
- Nebius Group — Managed Soperator, Nebius AI Cloud, Serverless AI
- RunPod — Pods, Runpod Clusters, Runpod Hybrid Cloud, Serverless
- Together AI — Together Frontier AI Factory, Together GPU Clusters
- Vast.ai — Vast.ai Clusters, Vast.ai GPU Cloud
Foundation Models (7)
- Alibaba — Alibaba Cloud Elastic GPU Service
- Alphabet (Google) — AI Hypercomputer, Google Cloud TPU
- Baidu — Baidu Baige AI Computing Platform
- Humain — HUMAIN COMPUTE
- Mistral AI — Mistral Compute
- SenseTime — Elastic Compute Service, Elastic Container Platform, SenseCore, SenseScale SSP
- Tencent — Tencent Cloud HAI
Hardware (24)
- AMD — AMD Developer Cloud, AMD Helios Rackscale Solution, AMD Instinct, AMD Kria SOMs, AMD Pensando AI NICs, AMD Pensando DPUs, AMD Versal AI Core, AMD Versal AI Edge Series Gen 2
- Applied Materials — AIx, Enlight 3 Optical Inspection, Kinex, SEMVision H20, Vistara
- Arista Networks — 7060X5 Series, 7060X6 Series, 7060XE7 Series, 7388X5 Series, 7700R4 Series (Distributed Etherlink Switch), 7800R4 Series
- Arm Holdings — Arm AGI CPU
- Axelera AI — Axelera AI Mini PC, Axelera Embedded 111c, Axelera Metis AIPU, Europa
- Broadcom — Broadcom custom silicon
- Cerebras Systems — Cerebras CS-3, Cerebras CS-4, Cerebras WSE-3 Turbo (WSE-3T)
- Cisco — Cisco AI PODs, Cisco AI PODs for Collaboration, Cisco Nexus Hyperfabric AI
- Dell Technologies — Dell AI Factory with NVIDIA, Dell PowerEdge XE7745, Dell PowerEdge XE9680, Dell PowerEdge XE9685L, Dell PowerEdge XE9712, Dell Pro Max AI Desktop
- Graphcore — IPU (Intelligence Processing Unit)
- Groq — Groq LPU (GroqChip), GroqCloud
- Hewlett Packard Enterprise — HPE Compute XD690, HPE Private Cloud AI, HPE ProLiant Compute DL380a Gen12, HPE ProLiant Compute DL384 Gen12, HPE ProLiant Compute DL585a Gen13, HPE ProLiant Compute XD685
- Horizon Robotics — Journey, Journey 6 Series
- Huawei — Atlas, Atlas 650E AI Server, Atlas 950 SuperPoD
- Intel — Intel Cloud Services, Intel Crescent Island, Intel Gaudi AI Accelerator
- Marvell Technology — Marvell custom AI silicon, Marvell OCTEON DPU
- NVIDIA — NVIDIA Blackwell, NVIDIA DGX B300, NVIDIA DGX Cloud, NVIDIA DGX GB300, NVIDIA DGX H100, NVIDIA DGX Spark, NVIDIA DGX Station, NVIDIA HGX Platform, NVIDIA IGX Platform, NVIDIA Jetson, NVIDIA Mission Control, NVIDIA OVX Systems, NVIDIA RTX PRO, NVIDIA Run:ai, NVIDIA Vera Rubin, NVIDIA Virtual GPU (vGPU)
- Qualcomm — Qualcomm Cloud AI 100 Ultra, Qualcomm Dragonfly AI200, Qualcomm Dragonfly AI250, Qualcomm Dragonfly AI300, Qualcomm Dragonwing AI On-Prem Appliance, Snapdragon, Snapdragon X Series
- SambaNova — SambaNova RDU, SambaRack, SambaStack
- Samsung Electronics — Samsung Foundry HPC/AI Service, Samsung HBM3, Samsung HBM3E, Samsung HBM4
- SK Hynix — CMM-DDR5, HBM3E, HBM4, SOCAMM2
- Super Micro Computer — Supermicro AI SuperCluster, Supermicro Data Center Building Block Solutions (DCBBS)
- Tenstorrent — Blackhole, Tenstorrent Cloud, Tenstorrent Galaxy, TT-LoudBox, TT-QuietBox, Wormhole
- TSMC — TSMC 3DFabric, TSMC foundry services