AI & Enterprise
In the AI inference era, corporate infrastructure shifts its centre of gravity
More than half of corporate AI investment is now going into running models rather than building them, as infrastructure shifts from training to inference, Kim Dong-hyuk writes. Inference runs at scale all day and can reshape data centre baselines and costs. As inference moves into production, latency, data location and governance drive more workloads toward on-premises or hybrid designs. Bottlenecks often emerge in data access, not computing, raising risks when data is copied or moved.
