Governed data. GPU-accelerated AI.
New: Starburst Torin
SQL, natively on NVIDIA GPUs.
Starburst Torin is a GPU-native, Trino-based analytics engine built on NVIDIA cuDF, now in limited general availability for Starburst Enterprise. It runs alongside Starburst’s CPU engine, with the flexibility to run workloads on CPUs or GPUs and no changes to existing SQL queries, applications, or agents.
Starburst Enterprise’s CPU engine also supports the NVIDIA Vera CPU.
Up to7x
Faster TPC-H query performance at production scale¹
¹ Starburst benchmarks ran on AWS in a development environment on single-node configurations, with warmup runs followed by measured runs. Speedups are geometric means across per-query ratios. TPC-H ran on g7e.4xlarge instances and ClickBench on g6.4xlarge instances (NVIDIA L4), compared with an m8g.8xlarge CPU baseline. Cost figures are hardware only and exclude software licensing. Results vary by workload and deployment.
Joint architecture
Governed data to AI inference. One pipeline.
Starburst provides the governed data layer. NVIDIA provides the compute. Together: a complete, enterprise-grade AI data pipeline, without exposing raw data to models.
Your data: any source
Amazon S3, Snowflake, Postgres, Kafka, on-premises, cloud: federated via Starburst, no copying required.
Enterprise Context Layer
Governed data products: versioned, AI-ready metadata, built for consumption by models and agents.
Starburst Agentic Control Plane
Policy enforcement, RBAC/masking, audit log, and rate limits: applied before any data reaches a model.
NVIDIA NIM microservices
GPU-accelerated inference: LLM, embedding, and RAG models consuming governed context via MCP / REST.
What you get
Enterprise AI. Built for production, not just the lab.
NIM-ready data products
Starburst data products expose a governed MCP interface that NVIDIA NIM microservices query directly: structured, policy-enforced context without access to raw tables or PII.
AI governance before the GPU
Every inference request that requires data goes through Starburst's Agentic Control Plane: RBAC, column masking, and row-level filters applied before data reaches any model.
Accelerated RAG pipelines
Build retrieval-augmented generation pipelines where the retrieval layer is Starburst, offering governed, fresh, federated data products. Generation is NVIDIA-accelerated LLMs on NIM.
AIDA on NVIDIA infrastructure
Run AIDA on NVIDIA DGX or GPU-accelerated cloud: the same governed natural-language data assistant, with the inference speed that enterprise workloads demand.
See what changes when intelligence reaches all your data.
Start free in minutes, or talk with a Starburst expert.