We’re excited to announce the second OPI Summit on DPU/IPUs, happening October 15, 2026 in San Jose, California, alongside the OCP Global Summit!
This gathering is designed for anyone who is building, deploying, or simply curious about the fast-growing world of Data Processing Units (DPUs) and Infrastructure Processing Units (IPUs). Whether you’re a chip designer, systems engineer, software developer, operator, or researcher, this will be your chance to learn, share, and connect with others shaping the future of infrastructure.
What can you expect from OPI Summit 2026?
As workloads demand more power, higher throughput, and lower latency in the era of AI, traditional CPU-centric architectures are hitting their limits. Introducing DPUs and IPUs to handle critical workloads such as security, networking, and telemetry has taken some of the pressure off CPUs, allowing them to focus on what they do best. But as powerful as DPUs and IPUs are, they bring complexity.
That is where this summit comes in: it’s a space for the community to share lessons learned, and discuss what’s working (and what’s not) in real-world deployments.
Have a topic? Submit your proposal to OPI Summit 2026
At this year’s summit, we will focus heavily on ease-of-use, Proof of Concept (PoC) implementations, and real-world experimental results. We want to see how these technologies perform in the wild—across systems of all scales, from small embedded edge devices to massive, multi-tenant hyperscale clouds.
We are seeking original, high-quality submissions that dive deep into the practical aspects of DPU/IPU integration. If you are working on cutting-edge implementations, we want to hear from you!
Please submit your proposal by August 23, 2026:
Submit your proposal to OPI Summit here
Key areas of interest for the 2026 Summit include:
Next-Gen Applications & Storage
- Networking & Security: High-performance networking and security offloads.
- Storage Disaggregation: Innovations in storage protocols, including NVMe-oF (NVMe over Fabrics) and High-Performance Computing (HPC) workloads.
Management, Standards, & Kubernetes
- Fleet Management: Vendor-agnostic solutions, lifecycle management (LCM), and API standardization to prevent vendor lock-in.
- Kubernetes Environments: Real-world use-cases, operators, resource-management, and ensuring DPUs support multi-tenancy for different Independent Software Vendors (ISVs).
- Ecosystem Interoperability: Building unified standards that apply across diverse hardware.
AI Infrastructure & Inference: As AI scales, the data bottleneck is real. We are looking for solutions addressing:
- Managing data movement for AI inference.
- AI-driven storage, including KV cache offloads.
- Secure GPU assignment via device attestation.
- Reducing virtualization compute overhead to free up valuable rack space and power for hungry AI hardware.
Security & Virtualization
- Confidential Computing: Securing the boundaries between the host and the IPU, establishing tenant-specific secure tunnels (host pod isolation), and leveraging DPUs for heavy cryptographic offloads.
- Virtualization Stack Optimization: Supporting legacy enterprise environments (OpenStack/QEMU), optimizing OVS offloads, and utilizing Host-Initiated IPU (Standard NIC/FNIC mode) for frictionless, drop-in adoption.
Sustainability & The Edge
- Green Data Centers: Optimizing power and thermal envelopes in high-density facilities.
- Edge Computing: Leveraging DPUs/IPUs at the network edge for ultra-low latency, localized processing.
Learn more about the event here and make sure to register for October 15, 2026!
