Conference Agenda

We are in the midst of one of the most significant infrastructure investment cycles in the aggressive buildout of generative AI (“GenAI”) capacity. Many commentators are worried that it is a bubble that will pop and wreak havoc on the economy. We take a longer view, drawing from the history of past infrastructure booms and busts, to identify potential pathways ahead for GenAI after the investment cycle crests. Past cycles suggest what to look for. Societies captured the gains from large infrastructure buildouts when two conditions held: the assets outlasted their builders, and regulatory constraints prevented concentrated owners from taking the downstream surplus. Whether those conditions hold for GenAI determines which of two paths we take. One path leads to a concentrated market, where the structure favors vertically integrated players who can lock in their advantages and foreclose new entrants. These concentrated firms can ride the treadmill of short-lived assets because they price by usage, see which applications succeed through API access, and compete directly in the downstream markets their customers occupy. The other path leads to a flourishing open market. The durable assets here are the open-weight models that persist across hardware generations and can be shared at low marginal cost. If open models remain at or near frontier capability and computing capacity becomes more of a commodity, the societal benefits of GenAI can diffuse more broadly. Crucially, the path ahead depends on policy choices made during the current formation window, a finite period before the structure hardens. This paper advocates for structural safeguards, including enhanced disclosure requirements, interoperability mandates, and protection of the open-weight ecosystem, to preserve the conditions for broad-based benefit from GenAI infrastructure.