AI capex cycle risks show circular investment strain

A critical debate is emerging on Wall Street about whether the AI capex boom is truly productive or merely circular. Nvidia's investments in CoreWeave, Iren, and Corning have sparked warnings that the chip giant is essentially funding its own demand via infrastructure bets. CoreWeave CEO publicly stated that without Nvidia's support, the company would not exist, highlighting the dependency chain. Critics argue that this circular model inflates capex spending without creating durable competitive advantage; once the initial buildout phase ends, return on invested capital deteriorates. Goldman Sachs and other strategists have flagged this risk explicitly, noting that AI infrastructure is becoming a closed loop rather than a genuine innovation multiplier. The concern is that we're in the hype phase of capex deployment, and once saturation arrives, capex collapses and multiples compress. Proponents counter that this is a necessary phase of technological adoption and that infrastructure investments will unlock genuine productivity gains in enterprise software and data centres. However, the fact that insiders are now openly discussing the circular nature of the spending is a yellow flag. If energy supply constraints or margin compression force capex discipline, the AI narrative could quickly turn from 'supercycle' to 'overcapacity trap'. For now, the momentumThe empirical fact that winners keep winning over the medium term. is still positive, but scepticism is rising among quality-focused investors who worry about capital destruction.