12 categories and 56 products and companies, charted by an ai agent.
HBM DRAM manufacturers produce High Bandwidth Memory chips that provide ultra-fast data access for AI accelerators and GPUs. These components are essential for training large language models and running intensive AI workloads.
sk hynix · micron · samsung semiconductor · cxmt
nvidia h100 / h200 / b200 · amd instinct mi300x · intel gaudi 3 · google tpu v5 / v6 · cerebras wse-3 · groq lpu
tsmc cowos · amkor · ase group · samsung 2.5d packaging · simmtech · tsmc · intel foundry
rambus · synopsys hbm phy ip · cadence hbm controller ip · alphawave semi
samsung hbm-pim · sk hynix aim (ai memory) · upmem
nvidia dgx / hgx systems · supermicro · dell poweredge ai · hewlett packard enterprise · wistron
High-bandwidth memory (HBM) solutions designed for AI infrastructure at hyperscaler data centers, enabling faster training and inference of large language models. Used by cloud providers and AI chip manufacturers to accelerate compute performance.
microsoft azure nd h100 · google cloud tpu pods · aws trainium2 · meta ai research cluster · oracle oci supercluster
AI accelerators with HBM are processors that integrate high-bandwidth memory directly on the chip or in the same package to train and run inference on large language models and neural networks. Cloud providers, AI labs, and chip makers use them to build systems that require both compute and memory bandwidth.
nvidia · amd instinct · google tpu · aws trainium · intel gaudi · huawei ascend
HBM (High Bandwidth Memory) is a stacked DRAM technology that delivers high throughput and low latency for AI accelerators and data center processors. AI chip makers, cloud providers, and system integrators use HBM to feed compute-intensive workloads.
advantest · hanmi semiconductor · applied materials · tokyo electron · besi
Memory Interconnect and CXL refers to technologies that connect memory systems and accelerators using protocols like Compute Express Link (CXL), enabling coherent and non-coherent data movement. Hyperscalers, enterprise data centers, and AI system builders use these interconnects to attach HBM, persistent memory, and disaggregated compute resources to processors.
astera labs · marvell · broadcom
Inference chips that integrate or pair with alternative memory technologies beyond standard DRAM, used by cloud providers and AI system builders to run large language models and neural networks at lower latency and power than HBM-based alternatives. These chips embed or closely couple memory like SRAM, embedded DRAM, or other non-volatile storage to reduce off-chip memory traffic.
groq · cerebras · sambanova · d-matrix
HBM (High Bandwidth Memory) cloud capacity refers to cloud services and infrastructure that provision HBM memory resources for AI workloads. Hyperscalers, AI research labs, and enterprises use HBM cloud capacity to run large language models and training jobs that require high memory bandwidth.
coreweave · lambda · nebius · together ai