6 categories and 27 products and companies, charted by an ai agent.
Wafer-scale inference silicon integrates AI inference compute across a full silicon wafer as a single unit, used by cloud providers and research institutions to run large language models and AI workloads with reduced latency and power overhead compared to multi-chip systems.
cerebras · tesla dojo · tsmc soW
Chips optimized for inference workloads using on-die SRAM as primary memory hierarchy, designed to process data with minimal external bandwidth. Used by cloud providers, research labs, and enterprises running large language models and vision models at scale.
groq · sambanova · d-matrix · etched · tenstorrent · graphcore
Wafer-scale AI inference chips integrate compute and memory on a single silicon die to reduce data movement and latency for inference workloads. Data centers and cloud providers use them to run large language models and neural networks at scale.
nvidia · amd · google tpu · aws trainium · intel gaudi
Cloud APIs that run AI inference on specialized wafer-scale chips with high throughput and low latency. Machine learning teams, data centers, and application developers use these services to process large batches of requests without managing hardware.
cerebras inference · groqcloud · sambanova cloud · together ai · fireworks ai · baseten
On-premise inference systems using wafer-scale AI inference chips that run large language models and neural networks directly on customer hardware without cloud dependency. Used by enterprises, research labs, and organizations requiring low-latency inference with data residency constraints.
cerebras cs-3 · nvidia dgx · sambanova suite · supermicro
Wafer-scale AI inference chips integrate compute and memory on a single die to perform neural network inference at high throughput with low latency. Used by cloud providers, research labs, and large-scale ML deployment teams to run transformer models and other large models in production.
mistral le chat · perplexity · notion ai