AI Dealmaking in an Era of Industrial Policy: What Parties Need to Know Across Jurisdictions

As AI capabilities become central to economic competitiveness, AI-related transactions are increasingly being examined through a broader policy lens. Alongside traditional competition considerations, authorities are placing growing emphasis on how these instruments may shape technological leadership, supply chain resilience, and national security. The US has adopted a more permissive posture aimed at accelerating domestic AI leadership, the EU is embedding technological sovereignty and resilience into its competition and industrial policy agenda, and China treats AI as a core strategic priority under its long-term industrial planning. These diverging approaches reflect a broader shift: control over AI technologies is now viewed as a matter of state capacity as much as market structure.

Against this backdrop, the assessment of AI-related transactions is becoming more complex and less predictable. Industrial policy considerations are increasingly shaping which deals attract scrutiny, the theories of harm advanced by authorities, and the intensity of intervention. At the same time, the expanding use of call-in powers and below-threshold review mechanisms in several jurisdictions means that regulatory exposure can no longer be inferred from filing thresholds alone.

This blog examines what these developments mean in practice for deal structure, transaction timelines, execution risk, and remedies in AI-related transactions.

Theories of Harm and the Influence of Industrial Policy on Merger Control 

Across jurisdictions, agencies are likely to examine a familiar set of competition theories of harm in AI-related transactions, but increasingly through the prism of industrial policy. Common concerns include input foreclosure, where an acquirer gains control over a critical input and can restrict or degrade rivals' access; conglomerate effects, where market power is leveraged across adjacent markets through tying or bundling; and the entrenchment of a dominant position, including through the acquisition of complementary assets that reinforce an incumbent's ecosystem rather than eliminating a direct competitor.

These theories of harm no longer operate in isolation from governments' broader strategic objectives. Increasingly, agencies are assessing whether a transaction affects access to technologies and assets that are regarded as critical to national AI capabilities or long-term economic resilience. This has brought greater attention to acquisitions involving key AI inputs and bottlenecks—including compute capacity, cloud infrastructure, semiconductors, high-quality datasets, technical expertise and downstream distribution channels—and whether a transaction could reduce innovation, foreclose future competition, reinforce technological dependencies or concentrate control over strategically important capabilities.

The significance attached to these concerns, however, differs markedly across jurisdictions because merger control is increasingly being used alongside industrial policy to advance broader economic and geopolitical objectives. In the EU, merger policy has become more closely aligned with the bloc's competitiveness and resilience agenda. The Draft Merger Guidelines expressly recognize scale, innovation, investment and resilience as potential procompetitive benefits, while also introducing theories of harm—such as dynamic foreclosure, loss of investment competition, and entrenchment of a dominant position—that are particularly relevant to AI ecosystems. More broadly, initiatives such as the European Technological Sovereignty Package and the proposed Cloud and AI Development Act reflect a growing concern with reducing dependence on non-EU providers of critical AI and cloud infrastructure.

The US is currently pursuing a different balance. The administration has consistently emphasized maintaining US leadership in AI and has cautioned against intervention that could impede innovation in rapidly evolving markets, while continuing to challenge transactions presenting clear competitive concerns. At the same time, it has signaled that foreign scrutiny of US technology companies may itself be viewed through a geopolitical lens, including through the possibility of trade measures where antitrust enforcement is perceived as disproportionately targeting US firms. China, meanwhile, treats AI as a strategic national priority under its 15^th Five-Year Plan. As a result, merger review of AI-related transactions may be intertwined with broader objectives relating to technological self-sufficiency, data governance, cybersecurity and control over strategically important technologies, particularly where transactions involve compute infrastructure, sensitive datasets or advanced technical know-how.

In effect, merger control in AI is no longer concerned solely with preserving competition in narrowly defined markets. It is also becoming a forum through which governments seek to shape the ownership, development and resilience of strategically important technologies.

Deal Structures and Jurisdictional Uncertainty

AI-related transactions increasingly take forms that fall short of outright acquisitions, including minority investments, acqui-hires, strategic partnerships and licensing arrangements. These structures may reduce or eliminate mandatory merger filing requirements, but they are also attracting increasing attention from competition authorities seeking to review strategically significant AI-related transactions regardless of form. As a result, merger control risk depends not only on the competitive significance of the transaction, but also on how it is structured and whether the assets involved are viewed as strategically important. This assessment may extend beyond companies whose primary business is AI. Businesses that market AI capabilities as a significant driver of value or future growth may attract closer scrutiny where those capabilities form part of the transaction rationale.

Minority investments illustrate this tension. While many jurisdictions require an acquisition of control before a filing obligation arises, others—including the US, UK, Germany, Austria, Australia, Brazil, Israel, Japan and Korea—can require notification of significant non-controlling shareholdings. Even where a filing is not required, governance rights, board representation or veto rights may increase scrutiny by suggesting the investor has acquired strategic influence over a key AI business.

Acqui-hires similarly occupy an increasingly important grey area. Hiring employees alone will generally not constitute a reportable concentration. However, where the transaction also transfers intellectual property, customer relationships or other assets that amount to a functioning business, authorities in the US, EU and China may look beyond the label and assess whether the arrangement effectively transfers the target's market position or nascent competitive potential. This is particularly relevant in AI, where the value of a business often lies in the combination of its technical talent, proprietary models and data rather than in substantial revenues or physical assets.

Licensing arrangements have likewise become a popular transaction structure, particularly where parties wish to collaborate without transferring ownership of AI technology. Non-exclusive licenses generally present lower merger control risk because they do not confer exclusive control over a critical input. Nevertheless, authorities are increasingly focused on the commercial reality of these arrangements, particularly where exclusive rights, long-term commitments or other contractual provisions may have effects similar to an acquisition.

This shift in deal structure, however, is being matched by a parallel expansion in enforcement reach. Some jurisdictions, such as Australia, have recently revamped their merger control regime, giving authorities broad jurisdiction to review transactions with a (remote) local nexus. And across jurisdictions, agencies are increasingly prepared to use an array of procedural tools to review AI-related transactions that would previously have escaped merger control. In the EU and China, in particular, agencies have demonstrated a readiness to investigate below-threshold transactions where they involve strategically significant assets and raise competition issues or nascent competitive threats. In the EU, although the Illumina/Grail judgment curtailed the European Commission's expansive use of Article 22 EUMR referrals, below-threshold scrutiny continues through Member States' call-in powers—including Italy's regime—and through transaction-value thresholds in Germany and Austria that capture acquisitions of high-value, low-revenue technology companies. In China, SAMR is increasingly relying on its call-in powers to investigate below-threshold transactions. Even in the US, the DOJ and FTC retain broad powers to challenge non-reportable transactions under Section 7 of the Clayton Act. 

For dealmakers, the message is twofold. Thoughtful transaction structuring remains an important tool for managing merger control risk, particularly where parties are considering minority investments, acqui-hires or licensing arrangements. But structure alone can no longer be relied upon to avoid scrutiny. Where AI-related transactions involve strategically important technologies, critical inputs or emerging competitive threats, parties should assume that authorities may look beyond legal form and consider whether call-in powers or other below-threshold review mechanisms could be engaged. That assessment should be built into transaction planning from the outset, including early identification of potential filing obligations and call-in risk, realistic transaction timetables, and the allocation of regulatory risk in the transaction documents. Parties should also exercise caution when characterizing transactions as AI-driven. While highlighting AI capabilities may support valuation or strategic rationale, it may also increase the likelihood of regulatory scrutiny where those capabilities become central to the investment thesis.

Navigating the Remedies Landscape

Where competition concerns are identified in technology and AI-adjacent transactions, remedies are increasingly being used to preserve access to critical inputs and constrain foreclosure risks without resorting to structural divestitures. In practice, as in China, agencies in the US and EU have shown a growing willingness to consider behavioral and access commitments—particularly where they safeguard interoperability, ensure rival access to key datasets, models or infrastructure, or address competitive concerns in a targeted and technically workable manner. The current US administration, in particular, has signaled a more pragmatic stance than its predecessor, moving away from an implicit preference for blocking transactions outright and showing greater openness to negotiated solutions in appropriate cases.

Recent decisions illustrate this increasingly flexible approach. In HPE/Juniper Networks, the DOJ accepted commitments requiring HPE to auction a perpetual, non-exclusive license to Juniper’s Mist AIOps source code, with optional transitional support and personnel transfers. This remedy resembles a structural one in effect, as it equips rivals with strong engineering teams to compete using the licensed asset, while reducing the need for complex regulatory monitoring. In Microsoft/Activision, the European Commission accepted a 10-year commitment requiring Microsoft to offer free licenses to EEA consumers and cloud game streaming providers for Activision PC and console games, directly addressing the identified foreclosure risk by preserving rival access to important content. 

The common thread among these cases is that remedies are becoming more design-sensitive and more outcome-oriented. Commitments are more likely to be accepted where they are tightly calibrated to the identified harm, technically workable, and capable of preserving competition without heavy regulatory monitoring. More broadly, access-based remedies increasingly sit at the intersection of competition enforcement and industrial policy, particularly in markets critical to digital and AI infrastructure.

Key Takeaways

To successfully navigate the shifting global enforcement landscape, dealmakers should adopt a proactive, sophisticated regulatory strategy:

• Assess the industrial policy dimension early. Parties should evaluate at an early stage whether the transaction implicates strategic AI inputs—such as compute, data, cloud infrastructure, semiconductors or technical talent—and whether it may be viewed as affecting technological sovereignty, resilience or long-term innovation capacity. These considerations, especially when promoted by the deal rationale, increasingly shape how theories of harm are framed—including with respect to FDI—and how aggressively authorities intervene.
• Develop a coherent global strategy. With differing priorities shaping enforcement in key jurisdictions, a coherent narrative that addresses jurisdiction-specific concerns while avoiding cross-jurisdictional contradictions is essential. Submissions in one jurisdiction may be used as evidence in investigations or litigation in other jurisdictions.
• Align transaction strategy and documentation with regulatory risk. Filing obligations, below-threshold investigations and call-in powers should be assessed at the outset so that transaction timetables, conditions precedent, efforts clauses, break fees and long-stop dates appropriately reflect regulatory risk. Deal documentation should explicitly allocate risk around regulatory intervention appropriately even where no filing is expected.
• Consider remedies proactively. Where substantial competition concerns are identified, parties should consider remedies early. Agencies have demonstrated some openness toward behavioral commitments—such as licensing source code, ensuring interoperability, or providing rivals access to key inputs—provided they directly and cleanly resolve the identified competitive harm.