Tracking Trillions & Jevons’ Paradox

The broader investment community has its ears to the ground as we finish up the summer, tracking ever-climbing yields on Treasury obligations against the backdrop of midterm elections coming fast. Political cycles tend to generate noise, but we are keeping our focus on what we believe is the defining investment story of 2026 – the extraordinary scale of capital being deployed by America’s largest companies in pursuit of AI-driven profits. 

A recent report from Goldman Sachs did a good job of outlining the full scale of the money that is being spent – and yes, the number is multiple trillions. If your curiosity is piqued, you can find the full article here. 

One insight that caught our attention was a key first hurdle:  the depreciation schedule to which American corporations are subject under financial accounting standards (GAAP), and the estimated economic useful life of AI hardware.

Put simply, capital expenditures are written off on a set schedule. The Goldman Sachs report highlights that because AI silicon evolves rapidly, chips may become economically obsolete in three years rather than five or six. If companies shorten useful life assumptions from five years to three years, annual depreciation expense spikes by hundreds of billions of dollars, directly suppressing reported earnings.

The consequences for not meeting this hurdle could be dire. It is an ironic twist that the laws intended to help corporations might undermine the duration they need. 

We view the depreciation schedule as a proper check and balance — a structural discipline that will separate the companies building durable AI businesses from those chasing a cycle.