Hannah Pittock argues that current analysis of reverse acquihires misses the core conceptual debate over antitrust’s antiquated treatment of hiring as benign vertical agreements between the laborers (the supplier) and employers (the buyer), in which labor is treated as one input among many.
There has been a wave of artificial intelligence deals between market leaders and smaller startups: Microsoft’s arrangement with Inflection, Google’s deals with Character.AI and Windsurf, Meta’s Scale AI transaction, Nvidia’s deal with Groq. These transactions—one firm hiring the target firm’s staff, particularly its founders, software engineers, and researchers, while leaving its products in a husk subsidiary or independent firm that it can license out without restriction—have split antitrust commentators into two camps.
On one side of the debate, enforcers and lawmakers like Senators Elizabeth Warren, Ron Wyden, and Richard Blumenthal have called these deals “de facto” mergers. The Department of Justice’s top antitrust enforcer recently described these transactions, often known as reverse acquihires (a traditional acquihire absorbs a whole firm to acquire its talent; a reverse acquihire works in the other direction, hiring the talent and leaving the firm’s products in a husk subsidiary or independent entity) as a “red flag” when they appear designed to sidestep ordinary merger review. On the other side, AI industry voices argue the deals simply reflect competitive hiring practices in a market that increasingly values people over products or intellectual property. If the firm losing talent wanted to keep it, they would pay more or offer more attractive benefits.
The enforcement camp is correct to identify that reverse acquihires are often tantamount to a merger, but both sides miss the deeper issues reverse acquihires present: antitrust’s conventional treatment of employment as a presumptively benign vertical arrangement between a buyer and supplier (firms buy labor, among other inputs) can cause it to miss transactions in which a firm hiring another firm’s team amounts to absorbing its competitor. The enforcement camp assumes that the underlying employment arrangements are vertical and that the issue is merely that the parties dressed a merger—a horizontal agreement between competitors—in different clothes. Meanwhile, the industry camp places the transaction in the “hiring” category—as a purely vertical arrangement—before asking whether that category fits.
Neither side asks whether antitrust’s conventional vocabulary (vertical for hiring, horizontal for competitor agreements) is doing substantive analytical work. What both camps fail to recognize is how labor can constitute the full boundary of a firm, rendering labor not as one input among many or an entity separate from a firm’s identity, but the firm itself. Antiquated categorization of labor risks dating antitrust analysis in markets where firms are dependent almost entirely on the intellectual capabilities of its staff and the acquisition of key staff is the acquisition of the firm itself. For the pro-enforcement camp, reliance on dated categorization of hiring may also preclude it from articulating the full harm of the reverse acquihire in policy and court.
Antitrust’s conventional treatment of hiring
Ordinarily, antitrust law treats hiring as a vertical transaction resulting from competition for an input, and that presumption exists for good reasons. Firms should generally be free to bid for workers, and workers should generally be free to take better offers. The antitrust agencies’ labor guidance therefore targets the opposite problem: horizontal agreements among employers that suppress competition for labor, such as wage-fixing and no-poach agreements. Those arrangements are unlawful precisely because they are horizontal agreements among rival firms, not bilateral contracts between a firm and a worker. The conventional map is therefore clear: unilateral recruiting is vertical and presumptively benign, and coordination among competing employers over labor terms is horizontal and potentially per se unlawful. The “this is just hiring” defense draws its considerable force from that background assumption. Once a deal is placed in the category of hiring, antitrust’s default posture is deference, but that deference rests on a structural premise that may not hold.
In a traditional industry, the distinction between a firm and its employees maps onto a real economic difference between a firm’s multiple inputs. In most industries, competitive capacity is spread across capital, intellectual property, distribution, brand, and labor. In other words, no single input is the enterprise. When a rival hires away an engineer, it is acquiring one input among many, and the target firm ordinarily continues to compete without the engineer. There is, in traditional industry, a real vertical separation between the enterprise that competes and the people who work for it.
In frontier AI, that separation can collapse. The binding competitive constraint in foundation-model development is a small number of researchers with the expertise to design architectures, manage training runs, and solve alignment problems. Compute and data, while expensive, are increasingly available as commodity inputs purchasable from cloud providers. A frontier AI startup’s competitive capacity is therefore concentrated in its team in a way that has no parallel in capital-intensive industries, where physical plant, supply chains, and distribution networks all constitute independent sources of competitive advantage that survive the departure of any individual or group of employees.
When the team is the competitive capacity and the remaining inputs are insufficient to allow the firm to continue to compete, the employees are not one input among many but are the competitor itself. However, conventional antitrust analysis still treats them as an input. It imposes a vertical distance that does not reflect organizational reality, making the acquirer appear to be purchasing labor-market inputs when it is in fact absorbing a whole product-market rival. This structural collapse is precisely what the current debate fails to register.
The horizontal/vertical distinction has always been a proxy: horizontal deals are presumed primarily competition-reducing, vertical deals primarily efficiency-enhancing, and the law calibrates its skepticism accordingly. That proxy made reasonable sense when the categories tracked real economic differences, when acquiring a supplier looked meaningfully different from absorbing a rival. But in markets where a rival’s entire competitive capacity is embodied in its team, the proxy breaks down: a transaction that is vertical in form can be horizontal in substance, and the law’s built-in leniency toward vertical arrangements does the rest. Antitrust’s systematic preference for vertical integration is already contested on its own terms; frontier AI is the setting where the costs of that preference become hardest to ignore.
The enforcement camp’s response, to subject these deals to merger review, is necessary but insufficient. Merger review itself relies on the vertical/horizontal distinction to assess competitive effects. If the framework treats the acquired team as a vertical input purchase, the deal will receive the lenience that framing implies even inside a formal merger investigation. The problem is not only procedural but analytical: antitrust needs to update how it understands the relationship between labor and the firm before merger review can do the work enforcers expect of it.
Kevin Kwok’s influential 2025 analysis illustrates how this classificatory move works in practice. He argues that reverse acquihires (he calls them a “hire-and-license-out” or “HALO,” but the structure is the same) are not, “acquisitions contorted into bizarre form. They are in fact hiring.” Yet that framing performs the classificatory move that matters most for antitrust: it places the transaction in the “hiring” category before asking whether that category encompasses all parts of a reverse acquihire.
Kwok’s framing has a legal counterpart in an argument that reverse acquihires do not meaningfully evade antitrust review and are thus not a consequential identification of market behavior for antitrust analysis. Rather, “talent hoarding” theories identify inefficiencies rather than competitive harm. The ability for acquired labor to jump to other firms render the acquiring firm’s ability to foreclose competition less plausible than when firms acquire physical assets. This account is useful because it exposes the precise premise at stake: the relevant conduct is one firm acting quickly to secure scarce labor. If that description is right, the conclusion mostly follows that antitrust should hesitate before condemning a company for hiring talent faster than rivals.
However, the description is not self-proving. In frontier AI, the core research team may itself constitute the competing enterprise, so paying to bring that team inside the incumbent can function less like a vertical input purchase than like a horizontal payment for a rival’s noncompetition, and the talent at issue may be, for all intents and purposes, the competitor.
Antitrust’s problematic understanding of hiring
Sanjukta Paul’s paper, Antitrust as Allocator of Coordination Rights, helps explain why “just hiring” is such a powerful defense, and why that defense may rest on a mistaken antitrust intuition. Paul argues that antitrust does not merely preserve “competition” in some abstract sense. More fundamentally, it allocates coordination rights. It decides which forms of economic coordination will be permitted, which will be discouraged, and which will be condemned. In her account, the business firm enjoys a “firm exemption” where coordination within the organizational boundaries of the firm (such as subsidiaries) are immunized or naturalized, while coordination beyond those boundaries, especially horizontal coordination between competing firms, is treated with suspicion. Paul further shows that modern antitrust has increasingly favored vertical control and hierarchical organization within a firm over more dispersed forms of cooperation. That is why the language of “just hiring” carries so much intuitive force. Once conduct is moved inside the firm, it becomes ordinary management, ordinary integration, and ordinary command. A firm’s boundary does not merely locate rights; it naturalizes a hierarchy and renders alternatives anomalous.
The classificatory problem runs even deeper than merger evasion. Consider a reverse acquihire that is subjected to merger review. Even then, the transaction may present as a vertical acquisition because what the acquirer obtains, formally, is a set of inputs: talent and possibly intellectual property. The conventional vertical frame treats input acquisition with relative lenience. The conventional account of hiring feels mismatched with our intuition that there’s competitive harm. To see the potential harm, one has to be able to see that hiring the team can be horizontally significant: the “input” being purchased may itself be the rival’s competitive capacity. Without that step, merger review simply reproduces the same classificatory error in a more formal setting.
Recent AI deals suggest an interesting extension of Paul’s insight. In this setting, the organizational boundary of the firm is not only where coordination becomes privileged, it is also the mechanism that reclassifies the conduct in the first place. On the startup side, the acquired firm often remains formally alive after the transaction. Character.AI continued to exist after its Google agreement; Windsurf remained a company after its founders and part of its R&D team moved to Google; Inflection remained an entity after its core personnel moved to Microsoft. On the acquirer side, however, the rival’s key researchers and their expertise are absorbed into the incumbent, and their future coordination becomes intrafirm by definition. The same legal boundary that preserves the appearance of a continuing rival also converts what may be, in substance, the elimination of rivalry into an ordinary employment relation. In this way, the firm’s boundary simultaneously performs classificatory and obscuring work.
This is easier to see once one stops imagining these arrangements as spontaneous defections by individual employees. They are often structured, multi-part bargains involving the startup itself. Microsoft’s Inflection transaction involved, in addition to employment contracts, a non-exclusive intellectual property license, a non-litigation agreement under which Inflection and its equity holders waived claims against Microsoft for soliciting and hiring the team, a line-of-credit amendment, and transition services.
Although the United Kingdom’s Competition and Markets Authority ultimately cleared the Microsoft/Inflection transaction, the analytical path it took actually sharpens the claim here. The CMA did not treat Microsoft’s conduct as ordinary recruiting. Instead, it concluded that the transaction created a relevant merger situation because the transfer of Inflection’s core team, their know-how, and the associated IP license gave Microsoft more than it could have obtained through open-market hiring, particularly because that development team was “at the core” of the business, and that Microsoft acquired economic continuity of at least part of Inflection’s business when the vast majority of the team moved. The deal was cleared only after that classificatory issue was resolved, on the narrower and fact-specific ground that Inflection was not, on the evidence, a material competitive constraint. That is, the CMA found that Inflection was too early-stage and too limited in scale to be exerting meaningful competitive pressure on Microsoft in the relevant market for foundation models. The clearance turned on the specific competitive significance of the target, not on the character of the transaction. That is fully consistent with the claim here. Antitrust should not let the label “hiring” answer the antecedent question of what has actually been transferred.
Within that frame, the speed-and-matching version of the “just hiring” defense, the claim that the acquirer simply moved faster than rivals in a competitive labor-market auction and offered superior compensation, is incomplete. It assumes the deal concerns labor in an ordinary input market, when the more fundamental question is whether the team itself was an independent center of competitive capacity. If a dominant firm pays billions in licensing fees, hires a startup’s founders and research leaders, and leaves behind a diminished shell, the fact that it moved faster in a labor-market auction does not answer the antitrust concern; it may simply describe the mechanism of consolidation. Nor does later labor mobility solve the problem. The competitive loss occurs when the startup ceases to function as an independent center of decision-making, and the possibility that some employees might later depart does not recreate the firm, its capital, its roadmap, or the experiments it would otherwise have pursued. In that setting, “hiring” is not a neutral description but a way of obscuring the internalization of a rival.
The agencies’ treatment of wage-fixing and no-poach agreements also points toward the right reorientation. Antitrust already recognizes that labor-related arrangements can be horizontal restraints when they suppress inter-firm rivalry. The missing step in the AI debate is to see that some arrangements styled as “hiring” may also be horizontal in competitive substance, not because the bilateral employment contract ceases to be bilateral, but because the overall transaction reallocates rivalry among actual or potential competitors. When a dominant firm pays for technology, secures waivers, and absorbs the rival’s core team, the law should hesitate before assuming that the vertical form of the employment contracts exhausts the horizontal substance of the arrangement. The pro-enforcement side has recognized this possible harm to competition, but it has not acknowledged the foundational work that the “hiring-as-vertical” assumption does in making the harm possible and hard to reach. Paul’s framework is especially useful here because it focuses attention on the allocation of coordination rights and on how incumbents can transform rivalry into hierarchy by moving key actors across the firm boundary.
The current conversation about “de facto mergers” is directionally right but analytically incomplete. Closing the procedural gap by subjecting these deals to merger review does not resolve the deeper problem if the analytical framework applied within that review continues to default to vertical-input logic whenever it encounters an employment contract. The contribution of a coordination-rights lens is to show that the firm boundary is not a neutral fact but a classificatory mechanism: it converts what may be the elimination of an independent rival into ordinary internal management, and it does so automatically once the employment relation is formalized. That insight extends beyond frontier AI. In any knowledge-intensive market (e.g. biotech, quantitative finance, specialized consulting) where a small team’s collective expertise constitutes the enterprise, the same classificatory slippage is available. Antitrust’s task is not simply to catch mergers that disguise themselves as hiring but to recognize that its foundational distinction between vertical input purchases and horizontal rival elimination can be collapsed by the formal structure of employment itself.
Author Disclosure: The author reports no conflicts of interest. You can read our disclosure policy here.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
Subscribe here for ProMarket’s weekly newsletter, Special Interest, to stay up to date on ProMarket’s coverage of the political economy and other content from the Stigler Center.
