AMD Instinct
AMD's data-center accelerator line, paired with the open ROCm software stack. The most credible alternative to CUDA for large-scale training.
Visit AMD Instinct →Every serious place to run an AI workload
Training and inference hardware is the most expensive decision in an AI stack and the hardest to reverse. This is a plain directory of who makes the silicon, who rents it by the hour, and who packages it into machines — grouped so you can compare like with like.
26 listings · 5 categories · updated September 2026Advertisement
Browse AI hardware
Grouped by where the hardware actually sits: in a data centre, at the edge, or in someone else's cloud that you rent.
Cerebras
Builds wafer-scale processors — a single enormous chip instead of many linked ones — to avoid the cost of moving data between devices.
Visit Cerebras →Google Cloud TPU
Google's tensor processing units, available only as cloud capacity. Tightly integrated with JAX and TensorFlow.
Visit Google Cloud TPU →Graphcore
Intelligence Processing Units built around large on-chip memory and fine-grained parallelism.
Visit Graphcore →Groq
Deterministic inference hardware built for very low latency token generation, offered both as silicon and as a hosted API.
Visit Groq →Intel Gaudi
Intel's deep-learning accelerator family, positioned on price-performance for training and inference rather than peak specification.
Visit Intel Gaudi →NVIDIA
The default AI training platform. Its advantage is as much CUDA and the surrounding software ecosystem as the silicon itself.
Visit NVIDIA →SambaNova
Reconfigurable dataflow architecture sold as a complete system rather than as loose accelerator cards.
Visit SambaNova →Tenstorrent
RISC-V based AI processors with an open software stack, sold as cards, systems, and licensable IP.
Visit Tenstorrent →Axelera AI
In-memory computing accelerators for edge vision workloads, sold as modules and evaluation systems.
Visit Axelera AI →Blaize
Graph-native edge processors and a low-code toolchain for building vision pipelines on them.
Visit Blaize →EdgeCortix
Energy-efficient edge inference processors and a compiler stack that targets them from standard frameworks.
Visit EdgeCortix →FuriosaAI
Korean designer of inference accelerators targeting efficient data-center and enterprise deployment.
Visit FuriosaAI →Google Coral
Small Edge TPU modules and boards for on-device inference, aimed at prototyping and low-power products.
Visit Google Coral →Hailo
Edge AI processors for cameras, industrial equipment and vehicles, aimed at high inference throughput inside a small power budget.
Visit Hailo →NVIDIA Jetson
Embedded modules that bring the CUDA ecosystem to robots, drones and industrial devices.
Visit NVIDIA Jetson →Qualcomm AI
On-device AI across phones, PCs and automotive, with a focus on performance per watt in battery-powered products.
Visit Qualcomm AI →Rebellions
AI inference chip company focused on energy efficiency for data-center and enterprise inference.
Visit Rebellions →CoreWeave
Specialised GPU cloud built for AI and rendering workloads rather than general-purpose computing.
Visit CoreWeave →Lambda
GPU cloud and on-premise deep-learning workstations and servers from the same vendor.
Visit Lambda →RunPod
On-demand and spot GPU containers aimed at developers who want to rent capacity by the minute.
Visit RunPod →Together AI
GPU clusters and hosted open-model inference, with training and fine-tuning on the same infrastructure.
Visit Together AI →Vast.ai
Marketplace that matches GPU renters with owners of idle capacity, usually at lower cost and more variable reliability.
Visit Vast.ai →Dell AI Infrastructure
Validated AI server and storage designs for enterprises that want a supported, pre-integrated stack.
Visit Dell AI Infrastructure →Supermicro
Builds the GPU server chassis and rack systems that most accelerator deployments physically live in.
Visit Supermicro →Lightmatter
Photonic interconnect and computing, attacking the data-movement bottleneck rather than the arithmetic.
Visit Lightmatter →No matches. Try a different search.
Advertisement
Questions
Short, direct answers.
What is the difference between a GPU and an AI accelerator?
A GPU is a general parallel processor that happens to be excellent at the matrix maths neural networks need. A dedicated AI accelerator is purpose-built for that maths and usually trades flexibility for throughput or efficiency. GPUs win on ecosystem and software maturity; accelerators often win on performance per watt for a narrower set of models.
Should I buy hardware or rent GPU cloud?
Rent while your workload is still changing shape, because the wrong purchase locks up capital in silicon that depreciates fast. Buy when utilisation is high and steady — sustained usage above roughly half of capacity is the point where owning usually beats renting, though the exact crossover depends on your power and hosting costs.
What does 'inference at the edge' mean?
Running a trained model on a device near where the data is created — a camera, a vehicle, a factory sensor — instead of sending the data to a data centre. It lowers latency, cuts bandwidth cost, and keeps data local, which is often the deciding factor for privacy or regulation.
Why is memory bandwidth talked about more than raw compute?
Large models spend much of their time moving weights rather than multiplying them. When a chip cannot feed its own compute units fast enough, headline throughput numbers are unreachable in practice. For large-model inference, memory capacity and bandwidth usually constrain real performance before the arithmetic units do.
Does this directory take payment for listings?
No. Listings are editorial and unpaid. Advertising on the site is labelled as advertising and has no bearing on what gets listed or where it appears.