India, Aug. 26 -- Artificial Intelligence (AI) is beginning to change more than the compute layer of the data centre. As enterprises move from experimentation to production, storage architectures, networking fabrics, power densities, cooling systems, and even application development choices are being reconsidered in light of the requirements of AI workloads.

Swastik Chakraborty, Vice President, Technology, Netweb Technologies, believes this requires enterprises to stop viewing Graphics Processing Units (GPUs) as the measure of AI readiness and instead optimise the complete system around data movement, utilisation and business outcomes.

In this interaction with Voice&Data Editor, he explains why training and inference need different infr...