CUDA adoption is surging, and the Anaconda and NVIDIA partnership has matured into a complete, supported distribution path that makes accelerated computing easier to get, trust, and run in production.
Over the past year, more than 192,000 organizations downloaded hundreds of millions of CUDA packages from Anaconda, and annual CUDA download volume grew 75% year over year. CUDA has become foundational to AI-native development, and a widening base of enterprise teams is reaching for it through Anaconda, from AI start-ups to the world’s largest enterprises.
Why this partnership matters now #
Numbers like these are the result of deliberate, joint investment. Anaconda is a licensed CUDA redistribution partner (and the one serving the enterprise), and over the past year NVIDIA and Anaconda have built out the main
channel into a complete, supported home for CUDA: comprehensive package coverage, curation and security scanning, signed metadata, and tight integration with the rest of the Anaconda Platform.
That maturity is why NVIDIA’s confidence in Anaconda has grown. The partnership has moved from simply packaging CUDA to co-engineering the experience around it, collaborating across releases and go-to-market so that customers get the newest accelerated computing with the assurance that production demands. For organizations adopting AI, the value is direct: developers keep their existing tools without forced migrations, every package and model arrives with security curation and full provenance, and what works in development runs the same way in production.
Adoption that keeps building #
The growth is broad and accelerating. Through the second quarter, customer demand for GPU-enabled frameworks rose more than 40%, GPU’s share of installs climbed to roughly 30%, and runtime usage of CUDA components grew sharply as teams moved workloads into production. The most-downloaded CUDA packages span both build and runtime needs, a sign that customers are not just experimenting but operating real systems. Demand is broadening across industries, from financial services and autonomous mobility to pharmaceutical research, national labs, and global manufacturing, and across markets worldwide.
A simpler path to CUDA updates #
Anaconda’s main
channel is the recommended path for CUDA going forward. The Anaconda-NVIDIA partnership has made main
a complete, enterprise-grade home for CUDA: new toolkit versions arrive promptly after upstream release, packages are security-scanned and curated end to end, and every install is backed by commercial support.
What is changing: after August 7, 2026, a defined set of NVIDIA CUDA products will stop receiving new updates on NVIDIA’s standalone nvidia
channel. Updates for those products continue on main
.
What it means for you: existing packages stay available and current environments keep working. If you already use main
or conda-forge
, nothing changes. If you install these products from the nvidia
channel, switching is a one-line configuration change for most setups; move before August 7, 2026 to keep updates flowing. If you prefer a community-governed solution without a commercial vendor dependency, conda-forge
is also a supported path.
Ready to build with GPU-accelerated AI? Check out the migration guide for step-by-step instructions.