DA AI AT AN DC EI N FTR AE SCTOR OU LCITNUGR E
What AI means for infrastructure – and why the channel needs to get involved
Increasing use of AI is putting pressures on servers, cooling systems and the like that they weren’t designed for, which means infrastructure is an important conversation – and resellers can add value here, as Martin Ryder, channel sales director Northern Europe at Vertiv, explains.
Artificial Intelligence (AI) is no longer emerging technology. It’s active in everything from financial modelling to image recognition and customer demand is growing. But while most of the attention has focused on software, the impact on physical infrastructure is becoming impossible to ignore. The servers, cooling systems and power setups that support traditional enterprise IT weren’t designed with AI in mind. The shift towards high-density, high-heat, power-hungry compute environments is already underway – and it’s not limited to hyperscalers. For channel partners, including resellers, systems integrators and MSPs, the implications are technical and commercial. As customers start to ask how AI fits into their existing infrastructure, they’ll want practical guidance. The channel is in a strong position to provide the infrastructure needed – but need to move quickly if it is to keep up with demand and take advantage of the opportunities this presents. AI is changing infrastructure assumptions One fundamental change is that AI workloads behave differently from most enterprise applications. They run faster, draw more power, and fluctuate far more aggressively in comparison to conventional workloads that typically
operate at a steady baseline. AI training and inference jobs can shift from low to high utilisation in seconds. This creates difficulties for existing power and cooling systems, especially in environments where capacity planning was based on older compute patterns. Racks designed for 10 kW–15kW loads are now being pushed towards 50kW or more. The infrastructure required to support this shift looks very different to traditional systems. Liquid cooling is starting to be implemented to work along side traditional airflow models. Power distribution units and uninterruptible power supplies need to handle surges and unpredictable spikes. Meanwhile, software-defined power management is emerging as a way to smooth loads and improve resilience. For most customers, adapting to these conditions will require more than a hardware upgrade. It may mean revisiting the underlying architecture. Customers feel pressure AI is now an essential part of infrastructure conversations. Some organisations are exploring early-stage deployment of AI tools, while others are further along the road and starting to hit real limitations – whether in their power systems, thermal management or physical capacity. In both cases, they’re looking for advice.
Martin Ryder channel sales director, Northern Europe
vertiv.com
“
Some organisations are exploring early- stage deployment of AI tools, while others are further along the road and starting to hit real limitations...
”
54
Powered by FlippingBook