Powering AI innovation
Transforming our infrastructure
for the next generation of AI
To meet the demands of an AI-driven future, we’re building a new, world-class
infrastructure for Meta’s family of apps and our long-term vision of the metaverse.
Event | July 31, 2024
AI Infra @Scale 2024
Meta’s Engineering and Infrastructure teams are excited to return for the second year in a row to host AI Infra @Scale on July 31st. This year’s event will be live streamed, as well as open to a limited number of in-person attendees. Registration is open now for both viewing options.
AI Infra @ScaleWhy we're building
Investing in innovation as we reimagine the future
We’re working to optimize the value and performance of our AI workloads today and beyond through the development of a next-generation AI infrastructure. These efforts are integrating specialty hardware, software, and networking components to support the creation, deployment, scaling and reliability of current and future AI applications.
We’re already making progress on our next-generation AI infrastructure. Read our blog posts for more on the initiatives underway.
An overviewMeta's large-scale clusters
Supporting our current and next generation AI models, including Llama 3
Learn moreMTIA
Meta Training Inference Accelerator:
The next generation of MTIA for Meta’s unique AI workloads
Meta's large-scale clusters
Supporting our current and next generation AI models, including Llama 3
Learn moreMTIA
Meta Training Inference Accelerator:
The next generation of MTIA for Meta’s unique AI workloads
Q&A
In this Q&A excerpt, Santosh Janardham, a key Infrastructure leader at Meta, talks about his role in shaping Meta’s infrastructure for the future.
Santosh Janardhan
Head of Infrastructure, Meta
Santosh is responsible for developing and operating the hardware, network, software, and data centers for all of Meta’s services. Read the Q&A for his take on the challenge of laying the groundwork for the future of the company’s infrastructure and more.
Read the Q&AFeatured tool
PyTorch
An open source deep learning framework built to be flexible and modular for research, with the stability and support needed for production deployment. It enables fast, flexible experimentation through a tape-based autograd system designed for immediate and python-like execution.
Explore toolFoundational models
Latest news
Foundational models