WHAT TO EXPECT

The compute track is designed for the world’s leading AI architects and hardware pioneers.

Explore the ultimate engine room of the AI Infra Summit, focused on making AI training and inference faster, more efficient, and more affordable for enterprises.

For architects, engineers, and technical leaders building and deploying AI hardware and systems. 


 

Compute Speakers

Who Attends?

How Will You Benefit?

COMPARE, VALIDATE & BENCHMARK

Compare, validate, and benchmark emerging hardware architectures. Access real-world performance data and deployment war stories across compute, networking, and storage layers, enabling better decisions on architecture trade-offs.

TAKE A FRONT ROW SEAT IN THE NEWSROOM

Take a front row seat in the AI hardware newsroom. Since 2018 we’ve been AI infrastructure’s ultimate stage for launching companies and products, from Habana Labs in our inaugural year, to NVIDIA’s Rubin CPX in 2025. Stay ahead of the curve with concentrated Q3 industry intelligence you won’t get anywhere else.

SUPERCHARGE YOUR ENGINEERING EFFORTS

Supercharge your engineering efforts. Learn how the world’s best engineering teams are pushing the limits of AI systems design, tuning data movement across memory hierarchies, optimizing interconnect topologies, and balancing power and thermal budgets for blistering training and inference performance.

FAQs

The AI Infra Summit’s compute track is focused on the primary area of concern in the AI infrastructure market – dominating the inference era.

It covers efficient and cost-effective inference, how using AI in chip design can develop specialised chips with a shorter time to market, how scheduling facilitates the “always-on” requirements of inference, and the critical need for monitoring in keeping enterprise AI usage accountable. 

 

The AI Infra Summit is a vendor-neutral conference for all start-ups and industry incumbents to present at. We showcase the entire AI infrastructure ecosystem, from bell weather brands to emergent start-ups, providing them with opportunities for both thought leadership and commercial announcements. 

The Compute track, as a whole, addresses how to make large-scale inference systems efficient, reliable, and economically viable.

The track will address how to efficiently utilise compute by minimising idle memory bandwidth, and how to match workloads to hardware capabilities. The track will also address how to efficiently allocate limited, expensive compute across competing inference workloads. 

  • The hyperscale and enterprise audience will leave the compute track having heard the cutting-edge developments in compute technology and having met the vendors selling the solutions necessary for scaling their AI infrastructure. 

  • Infrastructure start-up C-Suite can be confident that they have met critical hyperscale and enterprise customers and gained the press exposure pushing competitive advantage.  

  • For senior architects at established vendors, they will leave with deep insights into market development, their position against competitors, and the challenges enterprise customers need addressing.