PyTorch Foundation
The PyTorch Foundation is a vendor-neutral, community-driven home for the open-source intelligence layer used to train, optimize, serve, and run models on any chip, in any cloud, for any agent. Its mission is to drive adoption of AI and deep learning through an open ecosystem built around PyTorch.
At a Glance
- AI and machine-learning researchers and developers
- Enterprises building and deploying AI at scale
- Cloud providers, hardware and accelerator companies, and AI infrastructure vendors
- Universities, academic labs, and Open Source Program Offices
- +1 more
AI Tools by PyTorch Foundation
(5)torchtitan
PyTorch Native LLM Training Platform
torchtune
PyTorch LLM Fine Tuning Library
PyTorch/XLA
PyTorch on Cloud TPUs Compiler
TorchAO
PyTorch Quantization and Sparsity Library
PyTorch
Open Source Deep Learning Framework
Discussions
No discussions yet
Be the first to start a discussion about PyTorch Foundation
Latest News
PyTorch 2.15 released, including a rewritten NCCL backend becoming the default and other framework changes.
PyTorch Conference North America 2026 recap: 3,500+ attendees, Foundation portfolio update, and nine new members announced around the San Jose event.
Foundation welcomed nine members, including Platinum members Nebius, Nscale, and Shopify and Silver members LightSeek Foundation and Lablup.
Foundation announced nine new members as agentic-AI demand grows, including Clockwork.io, Emmi AI, NIPA, Nota AI, yasp.ai, CommonAI CIC, Carnegie Mellon, Monash, and Leicester.
Products & Services
The core open-source deep-learning framework for building and deploying machine-learning and AI systems.
A high-throughput, memory-efficient open-source inference and serving engine for large language models, with PagedAttention, OpenAI-compatible serving, offline batch inference, and multi-node support.
A distributed training and inference optimization suite supporting mixed precision, gradient accumulation, checkpointing, ZeRO, 3D parallelism, and inference acceleration.
A unified framework for scaling AI and Python applications, combining a distributed runtime with libraries for data processing, model training, serving, and reinforcement learning.
Market Position
The Foundation positions itself as a vendor-neutral umbrella for the open-source AI stack, spanning training, inference, optimization, serving, distributed compute, and agentic systems. Its differentiation is neutral governance and cross-company stewardship around PyTorch and adjacent projects; competitive alternatives at individual layers include TensorFlow/Keras and JAX for frameworks, and proprietary cloud or vendor-specific AI stacks for deployment.
Leadership
Founders
Meta
Created and incubated PyTorch; Meta researchers led the framework before its September 2022 transition to the Linux Foundation and continued investing in it as a primary AI research and production framework.
The Linux Foundation
Nonprofit open-source technology consortium that became the Foundation's host and steward when PyTorch moved from Meta in September 2022.
Executive Team
Mark Collier
Executive Director, PyTorch Foundation; General Manager of AI and Infrastructure, Linux Foundation
Co-founder of OpenStack and the OpenInfra Foundation; leads the Foundation's operational and business execution, governance committees, community programs, staff, finances, and membership development.
Matt White
Chief Technology Officer, PyTorch Foundation; Global CTO of AI, Linux Foundation
Served as Executive Director of the PyTorch Foundation for nearly two years before moving into the CTO role; leads technical strategy, hosted-project alignment, TAC and working groups, CI and multi-cloud infrastructure, academic/OSPO outreach, and PyTorch Certified.
Board of Directors
Founding Story
PyTorch development began in 2016 as a collective effort by members of the LuaTorch community, with substantial support from Meta and contributions from NVIDIA, Twitter, and others. As PyTorch became a multi-stakeholder project, Meta and the community moved it to the Linux Foundation in September 2022 to provide transparent business governance, vendor neutrality, and long-term stewardship while leaving technical decisions with maintainers.
Business Model
Revenue Model
Membership dues from companies and organizations, plus paid training and certification programs; members also contribute infrastructure, hardware, engineering effort, cash, or cloud credits. The Foundation is hosted by the Linux Foundation rather than operating as a venture-backed product company.
Pricing Tiers
One Governing Board voting seat and one TAC representative, strategic program participation, and 50 training/certification vouchers.
Eligibility for Governing Board and TAC representation and 50 training/certification vouchers.
The scale runs from $5,000 for 1-49 employees to $75,000 for 5,000+ employees; Silver members can be nominated for a Governing Board seat, capped at three seats.
Limited to academic and nonprofit institutions.
Target Markets
- AI and machine-learning researchers and developers
- Enterprises building and deploying AI at scale
- Cloud providers, hardware and accelerator companies, and AI infrastructure vendors
- Universities, academic labs, and Open Source Program Offices
- AI startups, government technology organizations, and open-source projects
- Research and development of deep-learning and foundation models
- Large-scale distributed pretraining and post-training
- LLM inference, serving, and production deployment
- Multimodal data processing and model serving
- Reinforcement learning and agentic AI workloads
- Model optimization, hardware portability, and multi-cloud deployment
- Meta
- Amazon Web Services
- Microsoft Azure
- Google Cloud