Overview

The development of frontier AI systems is dominated by a handful of labs associated with Amazon, Microsoft, and Google, the dominant cloud computing providers. Competition regulators are concerned that these technology giants are using anticompetitive behaviours to dominate the cloud computing market, allowing them to capture downstream markets such as AI.

The stakes are high: economically and geographically concentrated AI development and deployment does not bode well for how inclusive, accessible, and democratically accountable these systems can be in the future. Yet evidence of actual anticompetitive behaviours remains elusive. If not anticompetitive behaviours, then what could explain this concentration in the market for computation?

This project posits economic models to address this question, arguing that the we are currently seeing a reverse personal computing revolution in which data and computation are moving from millions of end-user devices and on-premises servers back inside large data centres, this time known as “cloud computing”.

During the 1980s, the so-called personal computing revolution saw large mainframes give way to personal computers and local servers. This was driven in part by the increasing affordability of semiconductor chips capable of handling an amount of data so great that the costs of connecting to central mainframes were no longer feasible. It became more economical for each organization and even each worker to perform computations locally, on their own computers.

Then from late 1990s onward fibreoptic cables and later 5G networks slashed the cost of connectivity again, while chip shortages, energy prices, and climate awareness put upward pressure on the cost of computation. The economics once again favoured centralization, now manifesting as cloud.

This “reverse personal computing revolution” is now reversing some of the political and economic consequences of the original personal computing revolution, so understanding this process as cloud computing plays a crucial role in frontier AI development is of crucial importance to guiding policymaking in this area.