Four things we learnt at GTC 2025
We’re back from a very successful GTC conference. The pace of technological change is remarkable! Here are four key trends I feel sum up what I learnt at GTC 2025 this year.
Traditional file storage is evolving
Memory is becoming the new storage. The industry is moving towards object-based data lakes, which enable AI to quickly process abstract queries. There are loads of really cool players out there that are well ahead of this evolutionary shift in storage - VAST Data being one of the leaders in the pack here.
Scheduling and Queue Management
We are seeing big shifts away from traditional infrastructure management solutions: Datacenters are looking for infrastructure, node management, and workload orchestration - which includes scheduling and load balancing - that allow them to serve up and assign custom services to customers. This is going to be critical when supplying infrastructure to companies and labs training AI Clusters* and serving inference, at speed, back to end users.
* An AI Cluster consists of a host server running RAVEL Orchestrate (the 'brain') that directs compute jobs to multiple target servers.
The rise of Neo Clouds
Neo Clouds - the new term for specialized datacenters - for example CoreWeave, Nebius, Lambda Labs, Together AI and Crusoe. These emerging NEO Clouds are aiming to significantly increase profits by transitioning from colocation services to workload-based solutions. They require straightforward applications that allow customers to utilize their bare metal infrastructure without the complexity and expense associated with traditional tools and service providers.
Most Important: POWER
This was the biggest theme of GTC. Talk of power savings and managing the power grids was top of mind. Liquid cooling is fast emerging as the way right now to save on power, and over usage of water to cool machines processing huge amounts of energy.
Want to discuss how we could help with your organization’s GenAI, or LLM training workload orchestration needs?
Useful links related to this article
- Liquid Cooling - scalable orchestration for liquid cooled infrastructure —> READ HERE
- Generative AI orchestration —> READ HERE