The Novelty of Innovation: Competition, Disruption, and Antitrust Policy

GAICIL Insights

Antitrust usually counts how much innovation a market produces.¹ This paper asks how daring it is. The Novelty of Innovation: Competition, Disruption, and Antitrust Policy, by Steven Callander and Niko Matouschek, models an innovator deciding both whether to innovate and how boldly. The more novel an innovation, the more different it is from the past and the more uncertain its outcome. Standard models often miss this dimension entirely.

The results overturn several intuitions. A variant of the Arrow replacement effect holds: new entrants pursue more innovative technologies than incumbents. Yet the entrant is less likely to disrupt an incumbent than the incumbent is to disrupt itself, and less likely to fail in the market. Boldness and disruption are distinct, and the link between them depends on market structure. The giant may be its own most dangerous rival.

Acquisitions deliver the plot twist. When the incumbent can acquire the entrant after innovation, the Arrow effect reverses. The prospect of acquisition makes innovation more profitable, and simultaneously suppresses its novelty, as the entrant seeks to maximize her value to the incumbent. Entrepreneurs innovate in ways that complement the incumbent's position, reducing boldness. The authors conclude that the reversal suggests a positive role for a strict antitrust policy that spurs entrepreneurial firms to innovate boldly. A startup building to be bought builds to fit.

The questions are the ones dominating current debates. Is a market more innovative when controlled by an incumbent with market power? Or is innovation better served by space for entrepreneurial upstarts and competition? They peak when policymakers decide whether to allow an incumbent to take over an upstart. The authors cite Facebook and Instagram, Amazon and Zappos, and AT&T and T-Mobile.

Now the AI angle. Artificial intelligence fits the model uncomfortably well. The AI industry features large incumbents and many startups, with frequent acquisitions and partnerships. The prospect of acquisition can shape the direction of startup innovation, encouraging work fitting incumbents' needs over radically novel approaches. If the most transformative AI advances require bold departures, the structure of exit opportunities matters for the field's trajectory. The exit market writes part of the research agenda.

Incumbents get a fairer hearing too. An incumbent may be more likely to disrupt itself than an entrant is to disrupt it. Large AI firms investing heavily in new approaches may drive significant change. The finding cautions against assuming incumbents always resist innovation, or that only startups can disrupt.

Policy needs nuance. Strict antitrust policy toward acquisitions can encourage bolder innovation by entrants. Acquisitions also make innovation more profitable, which can encourage entry. The balance depends on the value society places on novelty relative to the volume of innovation. Reviewers of AI acquisitions should weigh both: whether a deal strengthens incentives to innovate and whether it tilts innovation toward the incremental. Merger review gains a new question: what kind of ideas does this deal reward?

The authors note what the model sets aside, including market competition dynamics emphasized by disruption researchers. Combining the novelty dimension with richer competitive dynamics would deepen understanding of how market structure shapes innovation over time.

For AI governance, novelty is a missing lens. Competition policy often focuses on prices and market shares. Understanding how market structure and acquisition prospects affect the boldness of innovation lets policymakers see more deeply the incentives that antitrust creates.

Seen whole, the paper's message is plain: the quality and novelty of innovation, beyond its quantity, depend on market structure. Entrants innovate boldly, incumbents can disrupt themselves and acquisition prospects can suppress novelty. In a field as dynamic and consequential as AI, these insights offer a rigorous foundation for sustaining the bold experimentation that drives transformative progress.

Something for founders and enforcers alike to sit with: a thriving acquisition market can make an AI ecosystem look more innovative while making it less adventurous. Counting deals and patents will miss that; asking how far from the incumbent's roadmap the startups dare to wander will not.