Components
Technologies that change how systems behave and may change the game itself
This section looks at key components and architectural approaches that are shaping how modern HPC/AI and enterprise systems are built and used.
The focus is not on standard components, but on technologies that materially change performance, efficiency, capabilities or how resources are deployed.
We have a few categories to start with but will be adding more over time.
Accelerators
Compute accelerators, primarily GPUs, are at the center of current HPC/AI system design. As AI moves deeper into enterprises, we will increasingly see accelerated systems landing in those data centers as well.
But there are different approaches to how acceleration is delivered and used.
- Nvidia – dominant, full-stack approach
- AMD – competitive alternative, growing ecosystem
- Cerebras – wafer-scale, different architecture, completely different approach to AI processing
- NextSilicon – software-defined acceleration approach
Composable & Disaggregated Infrastructure
As accelerator costs and power consumption rise, improving utilization becomes critical. Composable and disaggregated infrastructure approaches aim to make expensive resources like GPUs more efficiently shared and allocated.
Players:
- GigaIO – GPU composability / pooling
- (optionally)
- Liqid
- HPE (GreenLake / composable angle)
- others later
The key question isn’t which component is “best.” It’s how these technologies are used together to improve performance, efficiency, and overall system behavior.
As systems become more complex and expensive, the ability to deploy and utilize these components effectively becomes the real differentiator.
This section will continue to evolve as new technologies and approaches emerge—and as we see what actually works in real-world deployments.
Vendors included on this site are selected based on technical relevance and real-world deployments.
If you have something worth showing, there are opportunities to participate through Deep Dive profiles.