AI Everywhere 2025

In the data center, the shift to inference is in full swing. Infrastructure for inference is being built out at scale around the world in what Jensen Huang calls “AI factories,” optimized for the best possible tokens per dollar performance on a small but growing number of open-source LLMs. Sovereign AI efforts are also gaining momentum. Several chip companies in this market have become AI cloud service providers, shifting their business models to providing APIs for fast tokens, where they have an advantage versus Nvidia. Is now the time for data center inference-only chips to shine? What’s the limitation on tokens per second, and what are the avenues for chip and system design that can break these bottlenecks? And will cutting-edge techniques like reasoning and agentic AI drive demand for more tokens? ​

At the edge, tokens per watt is becoming a dominant metric as all kinds of appliances get the ability to communicate in natural language. However, the rollout of AI-powered vision and audio analytics technologies, as well as time series sensor data analytics applications, is also becoming widespread in industrial, automotive, healthcare, and consumer electronics. Are edge chips keeping pace with the demands of new open-source models, and how do techniques like reasoning and agentic AI apply at the edge? How do we support edge AI applications from R&D to deployment and throughout the product’s life cycle? What challenges remain for chip and system design in the spectrum of edge use cases, and how can we discourage fragmentation and encourage efficiency, usability, and portability?​

These topics and more will be explored during this two-day virtual conference on December 10-11.

More information about the conference agenda will follow soon.

What to expect?

With a conference space and resource center, AI Everywhere functions similarly to a live exhibition and conference. The technical conference will include keynotes, panel discussions and technical presentations on a variety of subjects, including significant technical trends, design ideas and new application areas. Along with the conference, there will be a resource center with microsites from relevant businesses that will include useful links and AI-focused technical materials available for download.

Who should attend?

This event will cater for AI system designers across the spectrum of AI use cases, embedded engineers and software developers, data scientists and anyone with an interest in AI hardware, software and system design. The aim is to present the latest in hardware and software for this extremely fast-moving sector and discuss the challenges that remain and how the industry should respond to them.

Contact