Research

This paper examines how compatibility across standalone technology products anchors consumers to brands. Using a novel experiment, I identify the causal effect of compatibility, showing that willingness to pay for smartphones increases by 9% of the retail price when compatible with laptops. Utilizing these results, I construct a smartphone demand model incorporating compatibility with laptops to evaluate the welfare effects of (i) mandating cross-brand compatibility (“open ecosystems”) and (ii) cross-market mergers. I find that in 2018–2019, closed ecosystems benefit Samsung by locking non-Apple laptop owners into lower-quality Samsung smartphones, while open ecosystems increase Apple’s dominance. However, in 2020–2023, closed ecosystems benefit Apple, as Samsung’s superior smartphones induce Apple laptop owners to switch to Samsung under open ecosystems. Across periods, open ecosystems increase consumer surplus. A counterfactual cross-market merger between Samsung and HP reduces smartphone market concentration but raises Samsung prices, disadvantaging consumers who value compatibility less.

Artificial intelligence is emerging as a general-purpose technology with the potential to reshape production, innovation, and competition across the economy. Whether this potential yields widespread prosperity depends on the competitive environment for AI products. This article identifies four economic forces that will shape competition in AI: complementarities between products, endogenous sunk costs, diminishing returns to scale in data, and the burden of consumer protection regulation. Complementarities create both opportunities for modular systems and risks of tying or foreclosure. Endogenous sunk costs, particularly escalating investments in compute, models, and talent, may sustain concentration even as markets grow. While data exhibits diminishing returns, feedback loops can reinforce incumbency advantages. Consumer protection rules address privacy, bias, and safety but may impose compliance costs that disadvantage entrants. Together, these forces suggest that AI’s trajectory will reflect underlying economics, as well as legal enforcement choices with respect to tying and modularity.