4 categories and 22 products and companies, charted by an ai agent.
Datacenter training chips are specialized processors that accelerate the training of large machine learning models in cloud and on-premises compute clusters. Machine learning researchers, cloud providers, and enterprises use these chips to reduce training time for foundation models and other deep learning workloads.
cerebras · graphcore · sambanova · tenstorrent · etched · lightmatter · d-matrix
Inference cloud services provide on-demand API access to run trained AI models without owning hardware; used by application developers, enterprises, and researchers who need to execute inference workloads at scale.
groq · cerebras inference · sambanova cloud · together ai · fireworks ai
Edge AI chips are specialized processors that run machine learning inference on devices at the network edge—smartphones, IoT devices, cameras, industrial equipment—rather than in cloud datacenters. Device makers, robotics manufacturers, and autonomous system builders use them to process data locally with low latency and reduced cloud dependency.
hailo · axelera ai · blaize · sima.ai · deepx · untether ai
AI compiler and software stack products that optimize neural network models for execution on hardware accelerators. Used by ML engineers, platform teams, and hardware manufacturers to bridge models and silicon.
modular · tenstorrent tt-metal · tiny corp · mlir lab