RAPIDS
RAPIDS is an NVIDIA-maintained open-source ecosystem of GPU-accelerated Python libraries (cuDF, cuML, cuGraph, cuVS, dask-cudf) that provide drop-in acceleration for pandas, scikit-learn, NetworkX, Polars, and Dask, serving data scientists, ML engineers, BI users, and AI researchers globally.
- Company typePrivate
- Founded2023
- HeadquartersColumbia, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What RAPIDS does
RAPIDS is an open-source GPU-accelerated data science ecosystem owned, maintained, and primarily developed by NVIDIA Corporation. The project provides a suite of Python libraries that run on NVIDIA CUDA GPUs and use the Apache Arrow columnar memory format to deliver drop-in acceleration for the dominant PyData tools. Core libraries include cuDF (pandas-compatible DataFrames, including the cudf.pandas zero-code-change accelerator), cuML (scikit-learn-compatible machine learning), cuGraph (graph analytics, with nx-cugraph as the NetworkX backend), cuVS (vector search), dask-cudf (multi-GPU distributed DataFrames), cuSpatial (geospatial), cucim (image processing), and libcudf (C++ foundation). Distributed under the Apache 2.0 license, RAPIDS is installable via conda, pip, Docker, or cloud notebooks (Google Colab, AWS SageMaker Studio Lab, Paperspace, NVIDIA Launchpad) and deployable on all major hyperscalers.
The business model is freemium-to-enterprise: the open-source software is free, with monetization routed indirectly through NVIDIA hardware sales and directly via NVIDIA AI Enterprise subscriptions that bundle RAPIDS with optimization, certified hardware profiles, and IT support. The GTM motion is fully product-led and community-driven — no direct sales force at the RAPIDS layer; adoption flows through GitHub, Slack, Stack Overflow, technical blogs, and pre-installed cloud notebook environments. Target users span data scientists, ML engineers, graph analytics practitioners, BI users, and AI researchers, served via a horizontal (cross-industry) segmentation approach.
Note on entity structure: while one input source describes RAPIDS as a 2023-vintage private company in Columbia, US, this is contradicted by RAPIDS's own materials, NVIDIA legal disclosures, and the absence of any independent revenue or funding events. RAPIDS functions as an NVIDIA open-source project rather than an independent revenue-generating entity, and revenue is not separately disclosed.
RAPIDS firmographics
Firmographics- Name
- RAPIDS
- Legal name
- NVIDIA Corporation
- Website
- https://rapids.ai
- Company type
- Private
- Founded year
- 2023
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- RAPIDS is an NVIDIA-maintained open-source ecosystem of GPU-accelerated Python libraries (cuDF, cuML, cuGraph, cuVS, dask-cudf) that provide drop-in acceleration for pandas, scikit-learn, NetworkX, Polars, and Dask, serving data scientists, ML engineers, BI users, and AI researchers globally.
- Ownership category
- akta.pro rank
RAPIDS industry classification
Industry- Product category
- GPU-Accelerated Data Science Libraries
- NAICS
- Computer Systems Design and Related Services (54151), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (51821), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Programming, Data Processing, Etc. (7370), Services-Computer Processing & Data Preparation (7374)
- akta.pro primary industry
- Data Lake Platforms (HDAEABAC)
- akta.pro secondary industry
- Lakehouse Platforms (HDAEABAB)
Keywords
Where RAPIDS is headquartered
LocationHeadquarters
- HQ city
- Columbia
- HQ country
- United States
- HQ region
- North America
Markets served
RAPIDS business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Infrastructure, Operations, Marketing or Sales
Revenue model
- Open Source Distribution: RAPIDS libraries are freely available as open-source software under Apache 2.0 license. Revenue is generated indirectly through NVIDIA hardware sales and enterprise support offerings.
- NVIDIA AI Enterprise Support: Enterprise customers can access RAPIDS through NVIDIA AI Enterprise, which provides optimization, certified hardware profiles, and direct IT support.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Pay-as-you-go | Open Source (Free tier) |
Go-to-market motion1 record
Distribution channels9 records
Marketing channels8 records
RAPIDS product offering
Product offeringCore offering
RAPIDS is an open-source suite of GPU-accelerated data science libraries built on NVIDIA CUDA and Apache Arrow, providing drop-in replacements for popular PyData libraries. The ecosystem includes cuDF for DataFrame operations, cuML for machine learning, cuGraph for graph analytics, cuVS for vector search, dask-cudf for distributed computing, cuSpatial for geospatial data, and cucim for image processing. The libraries enable zero-code-change acceleration of existing pandas, scikit-learn, NetworkX, and Polars workflows on NVIDIA GPUs.
Product overview
RAPIDS is an open-source GPU-accelerated data science ecosystem built on NVIDIA CUDA and Apache Arrow. The suite provides libraries for end-to-end GPU-accelerated data science including: cuDF (pandas-compatible DataFrames with cudf.pandas accelerator mode), cuML (scikit-learn-compatible machine learning with cuML accelerator mode), cuGraph (graph analytics with nx-cugraph NetworkX backend), cuVS (vector search), dask-cudf (distributed GPU computing), cuSpatial (geospatial processing), and cucim (image processing). RAPIDS enables zero-code-change GPU acceleration for existing Python data science code through ecosystem compatibility with pandas, Polars, scikit-learn, NetworkX, and Dask.
Differentiator
Problem solved
Functional benefit
Brands
- cuDF: GPU-accelerated DataFrame library providing pandas-like API for tabular data processing
- cuML
- cuGraph
- cuVS
- cudf.pandas
- nx-cugraph
- Polars GPU Engine
- dask-cudf
Products and services
- cuDF GPU-accelerated DataFrame library providing a pandas-like API for loading, joining, aggregating, and manipulating tabular data on NVIDIA GPUs. Includes the cudf.pandas accelerator mode for zero-code-change acceleration of existing pandas workflows.
- cuML GPU-accelerated machine learning library providing a scikit-learn-compatible API with algorithms for classification, regression, clustering, dimensionality reduction, and hyperparameter optimization. Includes the cuML accelerator mode for zero-code-change acceleration of existing scikit-learn code.
- cuGraph GPU-accelerated graph analytics library providing algorithms for graph analysis including PageRank, community detection, centrality measures, and shortest path computations on NVIDIA GPUs.
- cuVS GPU-accelerated vector search library providing efficient similarity search for embeddings and high-dimensional data on NVIDIA GPUs.
- dask-cudf GPU backend for Dask DataFrames enabling distributed multi-GPU and multi-node computing. Automatically registered as the cudf dataframe backend for Dask using Dask-CUDA and LocalCUDACluster.
- nx-cugraph NetworkX backend providing GPU acceleration for NetworkX graph algorithms with zero code change, enabling 50x-500x speedups. Enabled via environment variable NX_CUGRAPH_AUTOCONFIG=True.
- libcudf CUDA C++ library with Apache Arrow-compliant data structures and fundamental algorithms for tabular data processing. Foundation layer for all RAPIDS Python libraries.
- cuSpatial GPU-accelerated geospatial data processing library for spatial and trajectory data operations.
- cucim GPU-accelerated image processing library providing APIs for loading, manipulating, and analyzing medical and scientific images on NVIDIA GPUs.
Quantifiable outcome
- Up to 150x faster pandas operations with cudf.pandas accelerator mode
- +2 more outcomes
Companies that use RAPIDS
Customer profileSegments4 records
Ideal customer profiles3 records
RAPIDS technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
Integration4 records
AI capability6 records
Feature7 records
RAPIDS partnerships and signals
Strategic signalPartnerships
Twelve partnerships are on record, tiered core and major.
- Apache ArrowcoreApache Arrow is a foundational open-source project providing the columnar memory format that RAPIDS libcudf is built upon. This integration enables zero-copy data interchange with other Arrow-native tools.
- DaskcoreDask provides the distributed computing framework that RAPIDS extends via dask-cudf for multi-GPU and multi-node operations. Deep integration enables scaling RAPIDS workflows across clusters.
- NetworkXcorenx-cugraph provides a GPU-accelerated backend for NetworkX, enabling 50x-500x speedups for graph algorithms with zero code changes. This is a first-class backend officially supported by the NetworkX community.
- scikit-learncorecuML provides GPU-accelerated ML algorithms with API compatibility to scikit-learn. cuML accelerator mode enables zero-code-change acceleration of existing scikit-learn code.
- Polarscorecudf-polars provides a GPU execution engine for Polars lazyframe operations. Enables GPU acceleration of Polars queries through a custom executor.
- Google Cloud PlatformmajorRAPIDS available on Google Cloud with integration for Compute Engine, Vertex AI, Google Kubernetes Engine, and Dataproc. Google Colab provides free RAPIDS access.
- Amazon Web ServicesmajorRAPIDS deployable on AWS EC2, EKS, ECS, and SageMaker. AWS provides RAPIDS-optimized GPU instances and SageMaker Studio Lab for free evaluation.
- Microsoft AzuremajorRAPIDS available on Azure with support for Azure VMs, Azure Kubernetes Service, Azure ML, and Dask-based clusters via Azure Portal.
- IBM CloudmajorRAPIDS deployable on IBM Cloud Virtual Servers for VPC with V100 GPU instances. IBM provides bare metal and virtual server options with NVIDIA Tesla GPUs.
- NVIDIAcoreNVIDIA is the primary developer and maintainer of RAPIDS. Provides core funding, engineering resources, hardware access, and integration with NVIDIA AI Enterprise for commercial support.
- XGBoostcoreXGBoost integrates with cuML for GPU-accelerated gradient boosting. The HPO workflow examples demonstrate XGBoost GPU training with dask-ml and cuML.
- PyTorchcorePyTorch is listed as an ecosystem partner, enabling integration between RAPIDS data preprocessing and PyTorch deep learning workflows.
Scale indicators1 record
Recent moves6 records
Expansion highlights5 records
RAPIDS competitors and assessment
Company assessmentEmerging players
- Modin: Drop-in pandas accelerator that parallelizes pandas operations across CPUs and (in some configurations) GPUs. Directly competes with cudf.pandas by offering an alternative zero-code-change acceleration story, though Modin historically emphasizes CPU scaling while RAPIDS emphasizes GPU acceleration.
- Anyscale (Ray): Distributed compute platform built on Ray, with native GPU support and growing integrations with the Python ML ecosystem. Listed as a RAPIDS partner but increasingly overlaps as a competitor for distributed, GPU-heavy data science and ML workloads where teams must choose between Dask (RAPIDS' backend) and Ray.
- Polars: Fast DataFrame library written in Rust that has become a popular pandas alternative. RAPIDS offers the Polars GPU Engine as a backend, but Polars itself is an emerging alternative stack; adoption of native Polars could partially divert workloads from the cuDF/Polars-GPU acceleration path.
Broad incumbents
- Databricks: Unified data analytics platform with built-in GPU acceleration via Spark RAPIDS and Photon. Databricks both partners with RAPIDS ecosystem (cited as a 40+ integration partner) and competes with it by offering an integrated lakehouse with native GPU support, serving the same data science and ML personas.
- NVIDIA AI Enterprise: NVIDIA's commercial software layer that bundles RAPIDS with enterprise support, certified hardware, and optimization services. Functions both as RAPIDS' commercialization vehicle and as a competing broader AI/ML platform versus third-party data science tooling.
- AWS SageMaker: Amazon's managed ML platform providing notebook environments, training, and deployment, with built-in GPU instance support and native RAPIDS integration via optimized containers and Studio Lab. Competes as a one-stop cloud ML platform versus RAPIDS' library-only model.
- Google Vertex AI: Google Cloud's unified ML platform offering training, deployment, and feature store capabilities, with RAPIDS pre-installed in Colab and integrations across Vertex AI. Competes with RAPIDS for enterprise ML workloads, particularly when customers prefer a fully managed environment.
- Snowflake: Cloud data platform with growing support for Python user-defined functions, Snowpark, and ML workflows. While not directly GPU-focused, Snowflake is increasingly enabling GPU-accelerated ML within its platform, competing for the same data scientist persona that RAPIDS targets.
Direct peers
- Dask: Open-source parallel computing library for Python that RAPIDS extends via dask-cudf for GPU-backed distributed DataFrames. Dask is a core partner but is also a competitive option when users want CPU-only distributed computing without the GPU hardware dependency.
- Apache Spark (with Spark RAPIDS accelerator): Distributed data processing engine with the Spark RAPIDS accelerator plugin that uses NVIDIA CUDA to accelerate DataFrames, SQL, and MLlib operations. Targets the same ETL and large-scale data engineering workloads as cuDF/dask-cudf, often considered alongside RAPIDS rather than instead of it.
Market position
Weaknesses4 records
Competitive moat5 records
Key risks5 records
Key highlights6 records
Customer concentration
RAPIDS social profiles
Digital presenceRAPIDS financial estimates
Financial estimateRevenue estimate
Valuation estimate
RAPIDS leadership team
Management profileNumber of profiles
RAPIDS funding detail
Funding detailFunding overview
Funding rounds1 record
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
RAPIDS M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about RAPIDS
What does RAPIDS do?
RAPIDS is an open-source suite of GPU-accelerated data science libraries built on NVIDIA CUDA and Apache Arrow, providing drop-in replacements for popular PyData libraries. The ecosystem includes cuDF for DataFrame operations, cuML for machine learning, cuGraph for graph analytics, cuVS for vector search, dask-cudf for distributed computing, cuSpatial for geospatial data, and cucim for image processing. The libraries enable zero-code-change acceleration of existing pandas, scikit-learn, NetworkX, and Polars workflows on NVIDIA GPUs.
Is RAPIDS a public or private company?
RAPIDS is a private company. It is classified as corporate owned and is currently operating.
When was RAPIDS founded?
RAPIDS was founded in 2023. It employs 1 to 10 people.
Where is RAPIDS based?
RAPIDS is headquartered in Columbia, United States, in the North America region.
How does RAPIDS make money?
Two revenue lines are on record. Open Source Distribution is the primary driver. The others are NVIDIA AI Enterprise Support.
Who are RAPIDS's main competitors?
Emerging players on record are Modin, Anyscale (Ray) and Polars. Broad incumbents are Databricks, NVIDIA AI Enterprise, AWS SageMaker, Google Vertex AI and Snowflake. Direct peers are Dask and Apache Spark (with Spark RAPIDS accelerator).
Does RAPIDS have an API?
No public API is recorded for RAPIDS.
What industry is RAPIDS in?
RAPIDS's product category is GPU-Accelerated Data Science Libraries. Its primary akta.pro industry code is HDAEABAC, Data Lake Platforms, with a secondary code of HDAEABAB, Lakehouse Platforms. Its NAICS code is 54151 and its SIC code is 7372.