18 categories and 83 products and companies, charted by an ai agent.
Specialized GPU processors optimized for training large AI models at scale in cloud and enterprise datacenters. Used by AI researchers, cloud providers, and organizations building foundation models.
nvidia h100 · nvidia blackwell b200 · amd instinct mi300x · amd instinct mi325x · intel gaudi 3 · cerebras wse-3 · graphcore bow ipu · sambanova sn40l
Custom silicon chips designed by hyperscalers to optimize AI workloads and reduce dependency on third-party processors. Used internally by cloud providers and enterprises running large-scale AI models.
google tpu v5e · google tpu v5p · aws trainium2 · aws trainium3 · aws inferentia2 · meta mtia v2 · microsoft maia 100
Specialized semiconductor chips designed to efficiently run inference workloads for AI models with optimized power and latency. Used by cloud providers, edge device manufacturers, and enterprises deploying AI applications at scale.
groq lpu · etched sohu · ibm spyre · tenstorrent wormhole · d-matrix corsair
Edge AI chips are specialized processors that run machine learning models directly on devices rather than in the cloud, enabling faster inference and offline capability. Used by smartphone makers, IoT device manufacturers, robotics companies, and automotive firms.
qualcomm snapdragon 8 elite · apple m4 · mediatek dimensity 9400 · hailo-8l · arm ethos-u85 · kneron kl730
Silicon chips optimized for high-speed data interconnection and networking within AI accelerator systems, enabling efficient communication between GPUs, TPUs, and other compute elements. Used by hyperscalers, cloud providers, and AI infrastructure builders.
nvidia connectx-8 · broadcom tomahawk 5 · marvell teralynx 10 · intel habana gaudi 3 nic · fungible dpu
Neuromorphic silicon designs brain-inspired processors that mimic biological neural networks for ultra-efficient AI inference. Used by roboticists, edge AI developers, and autonomous systems engineers who need low-power, real-time decision-making.
intel hala point · brainscales-2
Specialized processors designed for AI inference and training on personal computers and laptops, enabling on-device machine learning without cloud dependency. Used by developers, data scientists, creative professionals, and enterprises requiring local AI computation.
amd ryzen ai max · intel core ultra 200v · apple m4 pro · qualcomm snapdragon x elite · nvidia geforce rtx 50 series
Specialized silicon designed for AI inference and training that minimizes data movement between memory and compute units, maximizing throughput and efficiency. Used by cloud providers, edge device manufacturers, and enterprises running AI workloads.
sk hynix aim · samsung hbm-pim · micron crux · untether ai boqueria
AI chip design EDA and IP tools enable the creation of specialized AI accelerator silicon through automated design, verification, and reusable intellectual property blocks. Used by semiconductor companies, fabless design firms, and chip startups building custom AI processors.
synopsys dsp group ip · cadence cerebrus · arm cortex-x925 · siemens eda calibre
Chinese AI accelerators are specialized semiconductor chips and processing units designed to accelerate artificial intelligence workloads, used by cloud providers, research institutions, and enterprises for training and inference tasks.
huawei ascend 910c · cambricon mlu590 · biren br100 · iluvatar corex bi-v150 · moffett s30
Data center GPUs are general-purpose graphics processors deployed in hyperscaler data centers for AI training and inference workloads. They are used by cloud providers, AI labs, and enterprises running large-scale machine learning operations.
nvidia · amd instinct · intel gaudi · huawei ascend · moore threads · biren technology
Cloud custom accelerators are specialized silicon chips designed and deployed by cloud providers for their own data centers to optimize specific AI workloads. Hyperscalers use them to reduce latency and cost for inference, training, and domain-specific tasks across their platforms.
google tpu · aws trainium · microsoft maia · meta mtia · broadcom · marvell
AI accelerator silicon refers to specialized processors designed to speed up artificial intelligence workloads, used by cloud providers, enterprises, and research labs to run machine learning inference and training at scale.
cerebras · groq · sambanova · graphcore · tenstorrent · etched · d-matrix
Processors and accelerators designed for AI inference and machine learning on edge devices and mobile phones, used by device manufacturers and app developers to run models locally without cloud connectivity.
qualcomm · apple neural engine · mediatek · hailo · google coral · nvidia jetson
Foundry and memory refers to semiconductor manufacturing facilities and memory production that supply AI accelerator silicon. Used by fabless chip designers, hyperscalers, and chip startups who outsource production of their AI processors and memory components.
tsmc · sk hynix · samsung · micron
Photonic AI chips use light instead of electricity to perform neural network computations, reducing latency and power consumption. Data centers, cloud providers, and research institutions evaluate these for matrix multiplication and tensor operations in AI workloads.
lightmatter · celestial ai · ayar labs
Custom ASICs designed and deployed by hyperscalers for AI inference and training workloads in their own cloud infrastructure. Used by cloud providers like Google, Meta, Amazon, and Microsoft to run proprietary AI services.
Custom silicon designed and manufactured by hyperscalers or their partners specifically for running inference workloads at scale in data centers. Deployed by cloud providers, enterprise AI platforms, and large language model serving infrastructure to execute trained models on proprietary hardware.