AI Infra Summit 2026

AI Infra Summit 2026 takes place from 15 to 17 September 2026 at the Santa Clara Convention Center, in the middle of Silicon Valley. Organised by Kisaco Research, it is the descendant of the AI Hardware Summit and has grown into what the organisers bill as a full-stack AI infrastructure conference — by their published figures, 8,000 attendees, 400 speakers, 250 partners and eight stages. The tagline for this edition is “Where Inference Meets Influence”, and the framing is blunt: the age of inference has arrived, and the job now is to connect infrastructure investment to enterprise return.

The summit’s positioning is what distinguishes it from the crowded field of AI events. It sits deliberately at the infrastructure layer — the intersection of hardware and models that determines what the software layer can actually do — and concentrates on serving AI at scale, with the deployment and operational problems that come with it. The organisers put it as bluntly as anyone in the events business: there is no can-kicking or vague investigation of AI in some vertical industry here. You will not find a healthcare-AI track or a retail-AI track. What you will find is a room full of people whose stated mission is making AI faster, cheaper, more efficient and more sustainable, on the grounds that AI’s cost problem, energy consumption problem and scalability problem are all solvable at the infrastructure layer, and that the quality of that infrastructure ultimately determines whether the investment case holds.

Content is organised around five themes, described by the organisers as the bottom layers of the AI stack: AI Data Center (power, cooling, siting and build-out), Compute (silicon, accelerators, systems), Data Movement (networking, interconnect, storage — the layer that increasingly gates training and inference performance), Data & Models (the software and model layer that runs on top), and Physical AI (robotics and embodied systems). Each has its own dedicated stage, alongside a main stage, an expo theatre with its own agenda, and a hackathon.

The confirmed main-stage line-up is unusually senior for an infrastructure event. Jeff Dean, Chief Scientist and technical lead for Gemini at Google, and Lip-Bu Tan, Chief Executive of Intel, headline. They are joined by Ian Buck, VP of Hyperscale & HPC Computing at NVIDIA; Peter DeSantis, SVP for Foundational AI Models, Custom Silicon and Quantum Computing at Amazon; Santosh Janardhan, Head of Infrastructure at Meta; Kripa Krishnan, VP of AI Infrastructure at Meta; Fidelma Russo, CTO and EVP for Hybrid Cloud at HPE; Tony Pialis, EVP and GM for Data Center at Qualcomm; Monique Picou, Chief Product Supply Officer at Google; Ori Goshen, co-founder and co-CEO of AI21 Labs; Alexis Black Björlin, Chief Strategy Officer at General Catalyst; Ram Nagappan, VP for AI Infrastructure at Oracle; Chris Dolan, Chief Data Center Officer at Crusoe; and Shez Partovi, Chief Innovation Officer at Philips. The enterprise-buyer side is represented by figures including Prashant Mehrotra, EVP and Chief AI Officer at US Bank, Arun Nandi, Chief Data and AI Officer at Carrier, and Keith Sarbaugh, EVP and Chief Digital and Technology Officer at Zoetis — a deliberate signal that the agenda is about deployment economics as much as silicon roadmaps.

The attendee mix is correspondingly broad. The organisers list enterprises including Airbus, CVS Health, eBay, Eli Lilly, JP Morgan, McDonald’s, Netflix, PayPal, Uber and Walmart, and model builders including AI21 Labs, Anthropic, Cohere, GitHub and Google Gemini, alongside the chip, cloud and data-centre vendor community that has always been the summit’s core.

Registration is tiered. An expo ticket is listed at $247 and a full-access conference ticket at $1,497; a free Enterprise End-User Pass is offered for qualifying attendees on the buyer side — worth checking your eligibility before paying, as the criteria are set by the organiser. Full agenda, speaker and floorplan detail is published on the official site.

Go if you build, buy or operate the layer beneath the models: silicon and systems engineers, data-centre and network architects, infrastructure leads at model labs and hyperscalers, and the enterprise teams trying to make inference costs work at scale. It is one of the few events where you can hear from a chip chief executive, a hyperscaler infrastructure head and the person actually paying the inference bill on the same day.