If AI is a bubble that bursts, Kong's CEO says all the infrastructure spending will have been worth it (www.businessinsider.com)

Kong CEO Augusto “Aghi” Marietti warns that the current AI spending frenzy could be a bubble that “might blow up,” but argues the hyperscaling push is ultimately necessary. He likens today’s capex rush to 19th‑century railroad buildouts: some projects may be deployed ahead of demand, and there could be a down moment, but the infrastructure will be used later. Wall Street worries about sustainability — an analysis estimated Amazon, Microsoft, Meta and Google alone could spend roughly $320 billion on AI‑related capex — and leaders including OpenAI’s Sam Altman have acknowledged bubble risks. The core technical constraint Marietti highlights is energy: power availability is likely to be the primary bottleneck for the next phase of GPU‑heavy training and inference. Firms are already designing large data centers and self‑contained power solutions to feed fleets of GPUs, and OpenAI’s Greg Brockman has suggested eventual demand could approach “one GPU per person” scenarios. For the AI/ML community this signals two concrete implications: expect continued investment in compute, networking and power engineering (and corresponding supply‑chain and ops challenges), and plan for long‑term reuse of oversized infrastructure even if short‑term returns waver.