Prometheus is a memory-first AI server that collapses racks of infrastructure into one unit.
Prometheus is a memory-first AI server built to break the memory wall in AI infrastructure. It delivers rack-scale performance in a single unit with up to 128TB of uniform shared high-bandwidth memory per system, running multi-trillion-parameter models at a fraction of the power and cost.
The server runs on Ignite, a processor built around memory with ARM cores plus RISC-V vector and tensor cores operating in one shared memory space. It supports PyTorch, vLLM, and Triton with no rewrites, targeting hyperscalers and large enterprises running agentic systems and large-context workloads.
d-Matrix
Corsair AI Compute Platform
d-matrix.ai
Groq
LPU
groq.com
Cerebras Systems
Wafer-Scale Engine
cerebras.ai
| Attribute | Prometheus | Corsair AI Compute Platform | LPU | Wafer-Scale Engine | |
|---|---|---|---|---|---|
| Provider |  | | | | |
| Founded | 2023 | 2019 | 2016 | 2016 | |
| Sells To | Enterprises | Enterprise | Enterprise | — | |
| Pricing Model | Direct Sales | — | Recurring | — | |
| Ownership | Venture-backed | Private | Venture-backed | — |
Provider
Founded 2023
Sells To Enterprises
Pricing Model Direct Sales
Ownership Venture-backed
Provider
Founded 2019
Sells To Enterprise
Pricing Model —
Ownership Private
Provider
Founded 2016
Sells To Enterprise
Pricing Model Recurring
Ownership Venture-backed
Provider
Founded 2016
Sells To —
Pricing Model —
Ownership —
SambaNova DataScale is a data-center AI computing platform.
Fastest cloud for training/inference of 1B-24T models with no sharding.
RDU is SambaNova's Reconfigurable Dataflow Unit, the company's AI chip family for inference and training at scale.
World's largest AI chip: 4T transistors, 125 petaflops on one wafer.
Analog compute-in-memory processor delivering 25 TOPS for edge AI inference.
Analog in-memory inference platform collocating deep sub-threshold compute inside multi-level non-volatile memory.
Sohu is a purpose-built inference chip designed for high-performance AI model serving.
SambaNova DataScale is a data-center AI computing platform.
Fastest cloud for training/inference of 1B-24T models with no sharding.
RDU is SambaNova's Reconfigurable Dataflow Unit, the company's AI chip family for inference and training at scale.
World's largest AI chip: 4T transistors, 125 petaflops on one wafer.
Purpose-built AI supercomputers integrating GPUs, CPUs, networking, and software for enterprise training and inference.
MatX One is a high-throughput chip for large language model training and inference.
NVIDIA Hopper is a data-center GPU architecture for AI and high-performance computing workloads.
Analog compute-in-memory processor delivering 25 TOPS for edge AI inference.
Analog in-memory inference platform collocating deep sub-threshold compute inside multi-level non-volatile memory.