AI's NPV > $40 Trillion

The question of whether today’s artificial-intelligence buildout resembles a bubble has been haunting investors for months. The sheer scale of capex commitments creates an uncomfortable atmosphere of déjà vu – one that typically ends in tears. Yet, a simple line of reasoning may help settle the nerves before the winter break.

By 2030, some $7 trillion in data centers will have been deployed globally to meet surging demand for computing power. By then, companies and institutions will have climbed a steep learning curve and achieved tangible gains in productivity. If one assumes that AI ultimately improves labor efficiency by a modest 10% (either by reducing labor input for the same output or by increasing output for the same effort), then, on a global GDP of $130 trillion and with labor accounting for more than half of it, the annual benefit exceeds $6 trillion. This stream of productivity gains is, in effect, perpetual.

Further assuming an annual maintenance cash bill of around $1 trillion (roughly equivalent to depreciation on the installed base) and a conservative 10% real cost of capital, a net present value for society exceeding $40 trillion can be calculated.* The economic value unlocked for users is so large that suppliers, including hyperscalers, will earn attractive returns on capital while leaving an enormous surplus for users. Q.E.D.

This is not to say that sentiment related to this capex will be steady. Markets will oscillate, supply chains will stretch, and political narratives will shift. But based on common sense, investments in data centers appear sound, and the underlying cycle is fundamentally robust. In fact, it may well extend well beyond 2030: while incremental productivity gains inevitably face diminishing returns, rapid advances in models, agents, and automation, including robotics, should continue to shift the boundaries of economic activity.

At this point, Frédéric Dard, a wildly prolific French noir novelist with a remarkably sharp sense of human nature, would have likely interrupted the argumentation with characteristic style to warn that relying solely on logical projections carries its own risks: ‘Common sense is what allows those who are not intelligent to be heard.’ 

Mr. Dard was the author of the crime comedy ‘San Antonio’ (1949-2001), an irreverent French series about a detective who reveled in mocking institutions, politics, and social norms. Commissaire San Antonio, his mother Félicie, and his burlesque companions Pinaud and Bérurier were loyal friends during the long hours of my military service, sitting in my M113 as its driver. Loaded with slang, human insights, and innocent vulgarity, San Antonio is probably impenetrable to anyone who was not born and bred in a French-speaking and Francophile environment. But for those who were, it is a small treasure chest of anthropology and psychology.

Ahead of the festive season, Mr. Dard could remind anyone who may dread year-end table discussions that ‘two intelligent people with opposing ideas will find much more to talk about than two idiots who belong to the same party.’

And, to carry into 2026, a favorite piece of advice by San Antonio: ‘Live your present and leave your past for the future.’