Introducing Equinix Inference Exchange: AI Infrastructure for a Distributed World

As AI scales from experimentation to production, organizations face a new challenge: delivering inference where it matters most, close to users, applications, and data. In this session, Bruce Owen is joined by Vipul Ved Prakash, Co-Founder and CEO of Together AI, and Raj Mirpuri, Vice President of Global AI Clouds and Infrastructure Ecosystems at NVIDIA, for a discussion on the future of distributed AI inference and the launch of the Equinix Inference Exchange. Together, they explore how enterprises are moving beyond choosing a single AI model to optimizing multiple models for specific workloads across diverse environments. The conversation examines the growing importance of distributed infrastructure, high-performance compute, cost-efficient inference, and security as AI becomes an increasingly critical part of business operations. Key topics include: - The launch of Equinix Inference Exchange and what it means for enterprise AI deployments - How distributed inference supports multi-cloud, multi-model, and global AI strategies - The role of NVIDIA's validated AI infrastructure and Together AI's inference platform - Why performance, durability, and flexibility are becoming critical requirements for AI infrastructure - Strategies for reducing inference costs and improving token efficiency - Security, compliance, and data sovereignty considerations for enterprise AI Learn how Equinix, NVIDIA, and Together AI are helping organizations bring AI inference closer to their data, applications, and users while maintaining the performance, scalability, and control required for production workloads. Chapters: 0:00 Why distributed inference needs a neutral place to run 0:30 What Equinix Inference Exchange is 1:13 One customer, six models: what enterprise inference needs now 3:31 Reducing inference cost: tokens per GPU and optimized systems 5:09 Security, control and zero data retention FAQs: Q: What is distributed AI inference? A: Distributed inference runs AI models across multiple clouds, models and geographies, close to the users, applications and data they serve, instead of in a single centralized location. Q: How can enterprises reduce AI inference cost? A: By matching each workload to the right model, using optimized hardware and software that produce more tokens per GPU, and using open weight models that give more freedom to optimize than a generic deployment. Q: Why does it matter where AI inference runs? A: Where inference runs affects latency, cost and control over data. Running inference close to the applications and data environments you already trust helps keep performance high and data promises intact. Learn where each workload should run: https://youtu.be/WRILJzj8Olw #AIInference #DistributedAI #Equinix #opensourceai Equinix, Inc. (Nasdaq: EQIX) shortens the path to boundless connectivity anywhere in the world. Its digital infrastructure, data center footprint and interconnected ecosystems empower innovations that enhance our work, life and planet. Equinix connects economies, countries, organizations and communities, delivering seamless digital experiences and cutting-edge AI—quickly, efficiently and everywhere. 6859