14 categories and 54 products and companies, charted by an ai agent.
Specialized hardware accelerators designed for training large AI models in data center environments. Used by cloud providers, research institutions, and AI companies to reduce training time and costs.
cerebras · groq · sambanova · tenstorrent · d-matrix · etched · rain ai · olix computing · graphcore
Specialized semiconductor chips optimized for running pre-trained AI models with minimal latency and power consumption. Used by enterprises, cloud providers, and edge device manufacturers for inference workloads.
groq lpu · sambanova sn40l · untether ai · tenstorrent wormhole · axelera ai · esperanto technologies
Photonic computing uses light instead of electricity to process information, enabling faster computation with lower power consumption. Used by data centers, AI researchers, and telecommunications companies seeking performance advantages.
lightmatter · celestial ai · neurophos · ayar labs · luminous computing · astera labs · xscape photonics
Edge AI chips are specialized processors designed to run machine learning models directly on devices without cloud connectivity, enabling real-time inference with low latency. Used by manufacturers of smartphones, autonomous vehicles, industrial IoT devices, and robotics systems.
hailo · mythic · blumind · recogni · kneron · syntiant
enfabrica · intel habana · xsight labs · fungible (dpu)
Specialized semiconductor processors designed to run autonomous vehicle AI workloads like perception, planning, and decision-making at the edge. Used by automakers, robotaxi companies, and autonomous vehicle platforms.
hailo-8 · aurora innovation · quadric · ambarella
AI-powered tools that automate chip design workflows, from RTL generation to physical layout optimization. Used by semiconductor startups and design teams to accelerate time-to-market and reduce design complexity.
ricursive · synopsys dso.ai · chipflow
In-memory compute accelerators are specialized processors that perform AI computations directly in memory to reduce latency and power consumption, eliminating data movement bottlenecks. Used by enterprises, cloud providers, and edge computing systems for real-time inference and analytics.
mythic · upmem · gyrfalcon technology
Inference-first chips are processors optimized to run trained AI models at low latency and power consumption after training is complete. Data centers, edge devices, and embedded systems operators use them to deploy large language models, computer vision, and other AI applications in production.
positron · rebellions · furiosaai
Edge and embedded AI chips execute inference and machine learning workloads directly on devices at the network edge—phones, IoT sensors, robots, industrial equipment—without sending data to cloud servers. Device makers, roboticists, industrial automation engineers, and mobile app developers use them to run models locally.
blaize · sima.ai · deepx
Cloud-based AI chip design and simulation platforms that allow startups to design, verify, and optimize custom silicon without building full in-house fab infrastructure. Used by AI chip startups, design teams, and semiconductor companies to accelerate chip development cycles.
cerebras inference · groqcloud · sambanova cloud
Analog and in-memory compute systems perform neural network inference and machine learning workloads using analog circuits or processing-in-memory architectures to reduce data movement. Data centers, edge devices, and embedded systems use these chips to run inference at lower power than digital alternatives.
encharge ai · sagence ai
Software tools and platforms that help chip designers build AI processors, from RTL to layout. Used by chip startups, semiconductor companies, and in-house AI teams designing custom silicon.
chipstack