Hot Interconnects 2026 Program


All times below are Pacific Time (PT). The program is subject to change.

Conference Day 1: Wednesday, August 19

Time Title Presenters
09:00 - 09:15 Introduction & Day 1 Welcome Sayan Ghosh (Pacific Northwest National Laboratory)
09:15 - 10:15 Keynote 1: Lessons from networking Meta’s gigawatt-scale AI fleet
[details]
Omar Baldonado (Meta)
Session chair: Dan Pitt (Palo Alto Innovation Advisors)
10:15 - 10:30 Diamond Sponsor Talk: Lightmatter
BiDi DWDM: The Interconnect Foundation for High-Performance AI Training and Inference
[details]
Nick Harris
Session chair: Sayan Ghosh (Pacific Northwest National Laboratory)
10:30 - 10:40 Break
10:40 - 11:40 Technical Paper Session A: Network Design at Scale Session chair: Kapil Shrikhande (Upscale AI)
Multipass Random Leaf-Spine Networks for Extreme Scale Computing Systems
Scaling Inference Prefill with High-Radix Photonic Interconnect
11:40 - 11:50 Platinum Sponsor Talk: Broadcom
Distributed Computing @ Scale for AI Training & Inference
[details]
Mohan Kalkunte
Session chair: Kapil Shrikhande (Upscale AI)
11:50 - 12:00 Platinum Sponsor Talk: Cisco
Ethernet for Shared AI Fabrics: InfiniBand-Class Performance with Multi-Tenant Scale
[details]
Will Eatherton
Session chair: Kapil Shrikhande (Upscale AI)
12:00 - 12:10 Platinum Sponsor Talk: Meta
Co-Designing MTIA’s Communication Stack
[details]
Kirtesh Patil
Session chair: Kapil Shrikhande (Upscale AI)
12:10 - 13:00 Lunch
13:00 - 14:00 Technical Paper Session B: Hot Topic Presentations Session chair: Darius Bunandar (Lightmatter)
Demonstration of the 12.8T 8×DR8 High-Density Liquid-Cooled 64-Channel Pluggable XPO LRO Module and XPO Emulation Platform
Breaking Through the Network Wall: Petabit-Class Switching for Single-Hop MoE Training at Scale
The Multipath Reliable Connection (MRC) Transport
Diagnosing Hidden Interconnect Bottlenecks with Omnistat and Cassini Telemetry
14:00 - 15:30 Panel: Interconnects at the Edge Data Center: What are the roles of Scale-Up, Scale-Out, and Scale-Across now?
[details]
Panelists: Inder Monga (ESnet), Yosuke Aragane (NTT), Tanner Ryan (Cloudflare), Marc Austin (Hedgehog), Ofer Shapiro (Resolight), Amy Leeland (GEICO)
Session chair: Dan Pitt (Palo Alto Innovation Advisors)
Moderator: Timothy Crawford (KeyBanc Capital Markets)
15:30 - 15:45 The Head Bubba Memorial and HotI Closing Remarks Day 1 Dan Pitt (Palo Alto Innovation Advisors)

Conference Day 2: Thursday, August 20

Time Title Presenters
08:50 - 09:00 Day 2 Welcome Sayan Ghosh (Pacific Northwest National Laboratory)
09:00 - 10:00 Keynote 2: Networking Innovations for Gigascale AI Systems
[details]
Gilad Shainer (NVIDIA)
Session chair: Ron Brightwell (Sandia National Laboratories)
10:00 - 10:15 Diamond Sponsor Talk: Marvell
A Shared Memory Architecture for Token-Efficient AI Infrastructure
[details]
Ravi Mahatme
Session chair: Ron Brightwell (Sandia National Laboratories)
10:15 - 10:30 Diamond Sponsor Talk: Napatech
AI Fabrics in Motion: Programmable Interconnects for Scale-Up, Scale-Out, and Scale-Across
[details]
Alex Omø Agerholm
Session chair: Ron Brightwell (Sandia National Laboratories)
10:30 - 10:40 Break
10:40 - 11:40 MSA Lightning Talk Session
[details]
Session chair: Dan Pitt (Palo Alto Innovation Advisors)
OCI Drew Alduino (Meta)
Open CPX Ryan Yu (TeraHop)
SDM4 MCF Duane Robbins (Corning)
XPO Sunil Priyadarshi (Arista)
11:40 - 11:50 Platinum Sponsor Talk: Qualcomm
Scaling AI with Chiplets & CPO
[details]
Dave Kulansky
Session chair: Dan Pitt (Palo Alto Innovation Advisors)
11:50 - 12:00 Platinum Sponsor Talk: Eliyan
From Long Reach to Smart Reach: The Evolution of SerDes for AI Infrastructure
[details]
Rohan Gandhi
Session chair: Dan Pitt (Palo Alto Innovation Advisors)
12:00 - 13:00 Lunch
13:00 - 14:00 Keynote 3: The Future of AI Interconnects: Open Approaches to High-Performance AI Infrastructure
[details]
Bilal Riaz (Ciena)
Session chair: Pasha Shamis (NVIDIA)
14:00 - 15:30 Technical Paper Session C: Distributed AI Communication Session chair: Shelby Lockhart (AMD)
Diagnosing Overhead in Dispatch Operations: Cross-architecture Observatory
From Datacenter to Wide Area: Scale-Across Federated Learning on Real Long-Haul RDMA
Demystifying NVSHMEM: A System-Level Analysis on Symmetric Memory and Device-Initiated Operations in GPU Communication
15:30 - 15:40 HotI Awards Ryan Grant (Queen’s University)
Session chair: Ryan Grant (Queen’s University)
15:40 - 15:55 HotI Closing Remarks David Ozog (NVIDIA)

Conference Day 3: Friday, August 21

Track 1

Time Title Presenters
Morning Sessions
09:00 - 12:15 GPU Communication Libraries for Accelerating HPC and AI Applications Benjamin Glick, Pouya Kousha (NVIDIA), Andreas Herten (Jülich Supercomputing Centre)
12:15 - 13:30 Break
Afternoon Sessions
13:30 - 16:30 Principles and Practice of Scalable and Distributed AI Training and Inference Dhabaleswar K. (DK) Panda, Nawras Alnaasan (The Ohio State University)

Track 2

Time Title Presenters
Morning Sessions
08:00 - 11:00 Open Interconnect Innovations: A Deep Dive into the UALink Chiplet Specification Tutorial presented by UALink and UCIe Consortia Rob Pelt (AMD), Peter Onufryk (Intel Corporation)
Session chair: Shannon Kinkead (Sandia National Laboratories)
11:00 - 11:30 Break
Afternoon Sessions
11:30 - 15:15 A Practical Guide to Communication Libraries on AMD GPUs Aurelien Bouteiller, Corey Derochie, Edgar Gabriel, Nusrat Islam (AMD)

Track 3

Time Title Presenters
Morning Sessions
09:00 - 12:00 High Performance and Smart Networking Technologies for HPC and AI Dhabaleswar K. (DK) Panda, Benjamin Michalowicz (The Ohio State University)
Session chair: Amanda Bienz (University of New Mexico)
12:00 - 12:30 Break
Afternoon Sessions
12:30 - 17:00 Understanding Communication Performance on Emerging Heterogeneous Architectures Bob Robey, Shelby Lockhart, Giacomo Capodaglio (AMD), Amanda Bienz, Patrick Bridges (University of New Mexico), Sameer Shende (University of Oregon)