Crossing into AI: When Incumbents Build, Partner, Acquire, Absorb, or Wait A History-Friendly Agent-Based Model of the Generative-AI Market Transition
This paper employs a history-friendly agent-based model to demonstrate that incumbent firms' strategic choices in the generative AI market—ranging from building and acquiring to the novel "absorption" route—are primarily determined by the contractibility of tacit capabilities and the intensity of regulatory scrutiny, with no single strategy dominating across all conditions.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the business world, companies often face a critical moment when a new technology threatens to upend their industry. They must decide how to get the skills they need to survive. The traditional playbook offers three paths: build the technology themselves from scratch, borrow it by partnering with someone who already has it, or buy a company that possesses it. This decision is usually treated as a one-time choice made by a single company looking at a static market. However, the rise of generative artificial intelligence between 2022 and 2026 revealed that these choices are not made in a vacuum. When many companies rush for the same scarce talent and technology at once, their decisions change the landscape for everyone else. If one company locks up a key partner, rivals cannot reach them. If one buys a startup, that startup is gone from the pool. Furthermore, as the market becomes dominated by a few big players, regulators often step in to make buying companies more difficult and expensive. This creates a complex, shifting environment where the best move for one company depends entirely on what the others are doing.
To understand how this chaotic race played out, researchers built a computer simulation that acts like a digital ecosystem. Instead of trying to predict the future or fit data to a specific company, they created a virtual world populated by thousands of competing firms. These digital companies follow simple, realistic rules: they sense a threat, look at their options, and choose a path based on what seems most logical at the moment. The researchers started with a basic model of how companies usually switch from old technologies to new ones, and then added four specific rules that mirrored the unique conditions of the artificial intelligence boom. These rules included the rising cost of buying companies as the market got crowded, a new type of deal that had emerged in the real world, a sudden drop in the cost of building technology, and the risk that a new technology could eat away at a company's existing revenue. By running this simulation thousands of times, the researchers could watch how the population of companies evolved and see which strategies actually worked.
The simulation revealed that the winning strategy depends entirely on how easy it is to write a contract for the technology. If a company can simply rent the technology through a standard agreement, it usually partners with an outside provider. But when the technology is too complex or "tacit"—meaning it lives in the minds of a specific team and cannot be easily written down or rented—the company must own it. Here, the researchers discovered a new pattern that had appeared in the real world but was not part of the original theory. Instead of buying a startup outright, which triggers strict government reviews, many companies began to "absorb" them. In this arrangement, a company licenses the startup's technology and hires its entire team, leaving the original company as an empty shell. This allowed them to capture the necessary talent and know-how without triggering the same regulatory hurdles as a full acquisition. The simulation showed that this "absorb" route was not just a legal trick; it was often the most economical way to get the job done, even before regulators made buying expensive.
As the simulation progressed, it reproduced the exact sequence of events seen in the real artificial intelligence race. It began with a wave of partnerships, as companies tried to rent the technology while it was still available. As the market concentrated and regulators began to scrutinize big mergers, the simulation showed a sharp shift toward the "absorb" strategy. The total amount of capability that companies managed to bring inside their own walls remained steady, but the method of getting it changed. Instead of buying whole companies, they bought pieces of them through licensing and hiring. The researchers also tested what would happen if the cost of building technology dropped suddenly, as it did when some models were released to the public. The simulation showed that this price drop only revived the "build" strategy if it happened early, before the industry had settled on a single standard design. If the drop came too late, after a dominant design had already taken hold, it had very little effect.
Finally, the study looked at the fate of companies that chose to wait and see. The results were stark and depended on what the company actually sold. If a company's product could be directly replaced by the new artificial intelligence, waiting was a disaster. The simulation showed that these companies lost nearly all the value of their core business, even though they did not go bankrupt. Their revenue was simply eaten away by the new technology. However, for companies whose products were not easily replaced, waiting was a safe strategy that cost them very little. The research suggests that the danger of waiting is not a universal rule but a specific trap for those whose businesses are directly exposed to the new technology. The study concludes that there is no single best way to cross into a new market. Instead, the path a company takes is shaped by a combination of how easy the technology is to contract, how much regulatory pressure exists, the timing of cost changes, and how vulnerable the company's current business is to being replaced.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.