11 categories and 63 products and companies, charted by an ai agent.
Custom ASICs designed and manufactured by hyperscale cloud providers to accelerate AI model training and inference workloads. Used by cloud giants to reduce costs, improve performance, and reduce dependency on third-party chip suppliers.
google tpu (trillium) · aws trainium2 · microsoft maia 2 · meta mtia v2 · openai / broadcom xpu · apple aixi
broadcom · marvell · alchip · cadence · synopsys · si-five · arm · mediatek
Specialized hardware accelerators designed for running inference workloads in cloud data centers, optimized for AI/ML models at scale. Used by cloud providers, AI platforms, and enterprises deploying large language models and computer vision applications.
aws inferentia2 · google tpu inference · groq lpu · tenstorrent wormhole · d-matrix corsair · sambanova sn40l
tsmc · ase group · amkor · samsung foundry · intel foundry services
Custom cloud ASICs are specialized processors designed for AI workloads and data center operations, optimized for specific neural network architectures and inference tasks. Used by hyperscalers and enterprises to accelerate machine learning at scale.
broadcom tomahawk / jericho · marvell teralynx · pensando (amd) · fungible (microsoft)
Cloud-based custom ASIC design and manufacturing services that enable companies to build application-specific processors without owning fabrication facilities. Used by AI/ML startups, enterprises, and semiconductor companies seeking accelerated time-to-market.
google cloud tpu pods · aws trainium cloud instances · groq cloud · sambanova cloud · cerebras inference · fireworks ai · baseten
Custom application-specific integrated circuits designed and manufactured within China's sovereign cloud infrastructure ecosystem, enabling domestic control over AI hardware production. Used by Chinese cloud providers, government agencies, and enterprises prioritizing data sovereignty and technological independence.
huawei ascend 910c · cambricon mlu590 · biren br100 · iluvatar corex
Custom cloud ASICs are processor chips designed and manufactured for specific hyperscaler training workloads, deployed in data centers by cloud providers and large AI labs to accelerate machine learning model training.
nvidia · google tpu · aws trainium · amd instinct · intel gaudi · microsoft maia · cerebras · meta mtia
Custom ASICs built for inference workloads in cloud environments, operated by hyperscalers and cloud providers to run customer models at scale. Used by companies running large language models, computer vision, and other AI inference services.
groq · sambanova · aws inferentia · etched · d-matrix · tenstorrent
Custom cloud ASICs are processor chips designed and manufactured for specific cloud providers' workloads, used by hyperscalers to run their own inference, training, or data center tasks. Customers are cloud operators who deploy these chips in their own data centers or cloud platforms.
coreweave · lambda · together ai · crusoe · nebius
Edge AI ASICs are processors designed for inference and machine learning tasks deployed at network edges, data center boundaries, and enterprise locations. Used by cloud providers, telecom operators, and enterprises running real-time AI workloads closer to data sources.
hailo · axelera ai · qualcomm cloud ai · blaize