Artificial intelligence is undoubtedly the most important technological revolution of our time, with effects on our lives and the functioning of our economies that we have probably only just begun to glimpse. In the coming years, many economic sectors will be transformed by the adoption of AI, benefiting not only from greater efficiency but also from a radical redesign of production processes.
The problem is that, during the transition period, like the one we are experiencing now, an unprecedented investment cycle to build data centres and modernise power transmission networks means that this expansion of infrastructure runs the risk of intensifying inflationary pressures.
The demand for capital to finance AI-related infrastructure is of a magnitude much greater than that required during the expansion of the Internet, when measured in relation to the size of the economy as a whole. Therefore, although AI may have a deflationary effect in the long term, in the short term it acts in the opposite direction, driving up the prices of chips, energy, and raw materials like copper.
Another possibility points in the opposite, deflationary direction, and relates to the return on invested capital (ROI). If the trillions of dollars currently being allocated to AI infrastructure spending end up generating only a modest economic return, a large part of that investment will have to be written off, negatively affecting the balance sheets of those who made it.
The ultimate size of the market for AI service subscriptions remains difficult to quantify. Above all, there is a risk that competition from open-source models, coming from China and elsewhere, could significantly reduce the margins of industry operators, potentially undermining the sustainability of their business models.
We have seen this before, for example, during the boom and bust cycle of telecommunications and the Internet in the 1990s. Investment in broadband, fibre optics, and the underlying Internet infrastructure ultimately generated significant productivity gains, but not in an orderly fashion: first it contributed to the creation of a bubble in the stock markets and a sharp rise in bond yields —the yield on 10-year US bonds approached 7% during that period— and, subsequently, to the bursting of the bubble, with recessionary and deflationary consequences for the global economy.
In conclusion, there seem to be few doubts about the final destination of the transition to an AI-transformed economy: stronger growth and greater productivity, with the potential to mitigate the inflationary pressures arising from the macroeconomic challenge posed by rising public and private debt and increasingly accommodative and less sustainable fiscal policies.
However, the transition to this long-term favourable scenario could develop unevenly: on the one hand, it could intensify existing price pressures due to strong demand for raw materials and energy; on the other, it could lead to a recessionary scenario if the investment made soon proves excessive and is financed with too high a level of debt.