I. INTRODUCTION
The rapid development of artificial intelligence technology and its widespread industrial applications have exponentially increased the demand for high-performance computing hardware. The industry is evolving from infrastructures based on Graphics Processing Units (“GPU”), predominantly used in the training phase of large language models, to inference-focused architectures that provide end-users with real-time, efficient services with low latency. This transformation, known in the literature as the “Inference Flip,” has brought new requirements regarding latency, energy efficiency, and parallel processing capabilities that differ from GPU architecture.
In this context, the Language Processing Unit (“LPU”) architecture developed by Groq has positioned itself as a potentially substitutable and disruptive alternative to NVIDIA’s GPU-based solutions, particularly in the inference market, thanks to its deterministic processing sequence and low-latency token generation capabilities. Independent performance tests have shown that the Groq architecture can provide significant advantages over competing accelerators in certain large language model applications. Groq’s LPUarchitecture, especially its capacity to process individual user requests in real-time, has turned it into a strategic threat capable of fundamentally challenging NVIDIA’s market dominance.
From an international competition law perspective, the transaction between NVIDIA and Groq was not structured as a traditional merger but was designed as a hybrid model involving a technology license and talent acquisition (acqui-hire). The parties stated that the transaction was not an acquisition but a resource procurement and licensing deal. Groq also confirmed on its website that it was a license agreement. The fact that the parties do not classify the transaction as an acquisition and its legal form is based on a license agreement does not eliminate the possibility of the de facto elimination of the competitive structure in the market.
In legal doctrine, such structures are discussed under concepts like “asset-light acquisition”, “reverse killer acquisition” or “licensing-based concentration,” where the focus is on whether a competing technology’s independent development capacity has been eliminated, rather than the legal form of the transaction. The increasing scrutiny of such structures by competition authorities worldwide has accelerated debates on whether competition law should treat these transactions as genuine mergers, as if they were sham transactions.
II. ARTIFICIAL INTELLIGENCE CHIP TECHNOLOGY AND THE LEGAL FRAMEWORK IN COMPETITION LAW
The development and marketing of AI chip technology have become an area of significant interest for international competition law. NVIDIA holds a market share of approximately 70-95% in GPU technology, which has enabled it to control the data center chip market. Its Compute Unified Device Architecture (“CUDA”) software ecosystem, with over four million developers, further strengthens NVIDIA’s market position. This dominant market position has led competition authorities to scrutinize the transaction in question. Indeed, when evaluating the results of independent tests, Groq LPUs were found to be approximately twice as fast as their closest competitors in models like Llama 3 (e.g., 877 tokens/second). This situation also paves the way for the emergence of the killer acquisition thesis.
Groq can be considered a real and serious competitor to NVIDIA. In this context, the EC Merger Regulation 139/2004 (“EUMR”) and Section 7 of the U.S. Clayton Act provide for strict scrutiny of acquisitions of competitors by dominant companies. At the international level, the Digital Markets Act (“DMA”) in the European Union (“EU”) mandates stricter scrutiny of transactions carried out by dominant platform companies (gatekeepers).
Although legal developments in this area are limited in Türkiye, the Turkish Competition Authority (“TCA”) has the legal authority to review similar transactions under the Competition Act. Article 7 of the Competition Act stipulates that mergers or acquisitions by dominant companies are subject to review by the Competition Board. Transactions that neutralize potential competitors like Groq should be examined under the effects-based doctrine. The TCA must consider the real competitive effects of the transaction, regardless of its legal form. This does not necessarily require an intent to deceive, as in the concept of sham transactions, but focuses on the ultimate effect.
III. SIMILARLY STRUCTURED TRANSACTIONS IN THE INTERNATIONAL ARENA
1. Analysis of the License + Acqui-hire Model: The Microsoft-Inflection Example
The Microsoft-Inflection transaction, announced on March 21, 2024, is a pioneering example of the license + key personnel transfer structure. In this deal, Microsoft hired most of Inflection’s employees, including its co-founders, and obtained a long-term, comprehensive commercial use license for the company’s intellectual property. The parties did not classify the transaction as a merger or acquisition, but it was widely assessed that Inflection’s independent competitive capacity was significantly weakened. This complex structure drew the attention of four major competition authorities worldwide:
•United Kingdom Competition and Markets Authority (“CMA”): Based on its powers under the Enterprise Act 2002, the CMA launched an investigation and, on September 4, 2024, cleared the transaction, concluding that it did not result in a substantial lessening of competition. The CMA stated that although Microsoft did not acquire legal control over Inflection, it had thoroughly examined the de facto influence and economic dependency relationship. The CMA assessed the de facto acquisition nature of the transaction but concluded its market impact was limited.
•German Federal Cartel Office (“FCO” - Bundeskartellamt): In its review under Section 37 of the German Act against Restraints of Competition, the FCO determined that the transaction theoretically constituted a concentration, meaning Microsoft had acquired de facto control over Inflection. However, on November 29, 2024, it issued a decision of no jurisdiction, on the grounds that Inflection lacked sufficient local activities in Germany.
•European Commission (“Commission”): Following a referral request from seven member states, the Commission assessed its jurisdiction under the EUMR but decided on September 18, 2024, that it lacked jurisdiction.
•U.S. Federal Trade Commission (“FTC”): An investigation launched in June 2024 is still ongoing. The FTC and the U.S. Department of Justice (“DOJ”) are reportedly debating whether to characterize this transaction as a genuine merger rather than a talent acquisition.
The NVIDIA-Groq transaction is currently being evaluated by the Commission, following an approach similar to the FCO’s in the Inflection case.
2. Abuse of Dominance and Vertical Merger: The NVIDIA/Run.ai Example
In April 2024, NVIDIA’s announcement of its acquisition of AI workload management software company Run.ai for approximately $700 million prompted action from global competition authorities due to NVIDIA’s 80-90% dominant position in the AI chip market:
The Commission completed its investigation under the EUMR in December 2024. It unconditionally cleared the deal, concluding that NVIDIA had neither the ability nor the incentive to make the Run.ai software incompatible with competing GPUs.
An investigation initiated by the DOJ under Section 2 of the Sherman Act is ongoing. The DOJ is examining whether this acquisition could lead to an abuse of NVIDIA’s dominant position and limit customer flexibility.
3. Indirect Control and Competition Infringement in Labor Markets: The Delivery Hero/Glovo Decision
On June 2, 2025, the European Commission fined these two giants of the food delivery sector a total of €329 million (€223.3 million for Delivery Hero, €105.7 million for Glovo) for forming an illegal cartel between 2018 and 2022. The Commission based its decision on Article 101 of the Treaty on the Functioning of the European Union (“TFEU”). This decision established three important legal precedents:
•“No-poach” agreements were recognized as a standalone competition infringement.
•Cartels targeting labor markets were brought under the scope of competition law.
•Delivery Hero’s 15% minority shareholding in Glovo was deemed to have facilitated information exchange and strategic coordination, making it part of the cartel and thus punishable.
This last point supports arguments that mechanisms in the NVIDIA-Groq transaction, such as NVIDIA gaining control over Groq’s assets and technology, could have a restrictive effect on competition similar to the coordination enabled by a minority stake.
IV. DECISION ANALYSIS: AN EXAMPLE OF THE EUROPEAN COMMISSION’S ANTITRUST PRACTICE
The Commission’s conditional approval of the Microsoft-Activision Blizzard transaction on May 15, 2023, under the EUMR, provides a comprehensive example of how the concepts of de facto control and concentration are applied in technology markets and how competitive concerns can be addressed. This decision serves as an important roadmap for analyzing the potential consequences of non-traditional technology and talent transfers like NVIDIA-Groq.
In its in-depth investigation, the Commission conducted a multi-faceted assessment of the potential competitive effects of Microsoft strengthening its dominant position in the game publishing market by acquiring Activision’s critically important game portfolio, including Call of Duty. The Commission’s main findings were:
•No Serious Risk in the Console Market: The Commission found that Microsoft had no incentive to withdraw Activision’s games from its rival console, Sony’s PlayStation. Sony’s market-leading position would make such a move commercially irrational for Microsoft. The Commission noted that even if Microsoft made such a decision, Sony could maintain its competitive position with its extensive game catalog and market power.
•The Main Competition Concern: The Cloud Gaming Market: The Commission identified the primary competition risk as being concentrated in the emerging cloud gaming services market. If Microsoft were to make Activision’s popular games exclusive to its own cloud gaming service, Game Pass Ultimate, and withhold them from rival platforms, this could significantly reduce competition in this innovative and growing market. This created a risk of market “lock-in” in favor of one player at an early stage.
To address these concerns, Microsoft offered comprehensive ten-year commitments to the Commission, which formed the basis for the approval:
•A Free License for Consumers: Consumers in the European Economic Area (“EEA”) would be granted a free license to stream all current and future Activision Blizzard PC and console games they own on any cloud game streaming service of their choice.
•A Free License for Service Providers: Cloud game streaming service providers would be granted a corresponding free license to allow EEA-based gamers to stream these games on their platforms.
These commitments prevented Activision’s games from being locked into a specific platform, ensuring the market remained open to competition. The Commission emphasized that this remedy constituted a significant improvement for the cloud gaming market compared to the status quo.
The key lessons from this decision in the context of the NVIDIA-Groq transaction are:
•De Facto Control and Input Foreclosure Risk: Just as Microsoft gained control over Activision’s game portfolio, NVIDIA is gaining de facto control over Groq’s LPU technology and the key personnel developing it through the license and talent transfer. The primary concern for competition authorities will be NVIDIA further strengthening its dominant position by foreclosing this superior inference technology to rivals or integrating it exclusively into its own CUDA ecosystem.
•Potential for a Commitment-Based Solution: The Microsoft-Activision decision shows that competition authorities may demand behavioral commitments to keep the market open to access, rather than blocking technology acquisitions outright. Following a similar approach, in reviewing the NVIDIA-Groq transaction, authorities could require NVIDIA to license Groq’s LPU technology to third-party chip manufacturers or AI companies on Fair, Reasonable, and Non-Discriminatory terms.
In conclusion, the Microsoft-Activision decision is a landmark case demonstrating that in technology acquisitions like NVIDIA-Groq, regulatory authorities may focus not only on the legal structure of the transaction but also on behavioral commitments that will shape the future market structure and eliminate the risk of “input foreclosure.”
V. CURRENT ISSUES AND DEBATES
1. Challenges Faced by Competition Authorities
A key issue in the NVIDIA-Groq transaction is that the license + acqui-hire structure does not fit squarely within the framework of existing antitrust laws. In the U.S., transactions falling below the Hart-Scott-Rodino (“HSR”) Act filing thresholds are technically exempt from federal antitrust review. However, the FTC and DOJ are increasingly inclined to review such deals using an effects-based approach. An email from NVIDIA CEO Jensen Huang to employees, which was leaked to the press, stating, “We are not buying Groq, we are just licensing,” creates an impression that runs counter to the authorities’ focus on ultimate effects.
2. Issues Regarding Business Models and Market Dynamics
A significant problem with the NVIDIA-Groq transaction from a market dynamics perspective is the diminishing number of alternatives in the AI chip market. Groq offered a genuine alternative with a different architecture from AMD, Intel, and Google’s TPUs (Tensor Processing Units). NVIDIA’s suppression of Groq’s competitive potential creates a dynamic that forces smaller startups to choose between merging with NVIDIA or exiting the market. While the increased valuations of other companies like D-Matrix, Cerebras, and SambaNova after the NVIDIA-Groq deal suggest it is not a “winner takes all” scenario, the dominant firm’s behavior appears questionable.
In various jurisdictions, the concept of killer acquisition is being taken increasingly seriously. Particularly in the pharmaceutical industry, the acquisition and “shelving” of rival drug development projects by dominant companies is considered an infringement of competition. A similar dynamic is observed in the NVIDIA-Groq transaction: NVIDIA is controlling the management and engineers who ensure the continued development of Groq’s LPU technology, thereby preventing its evolution as an independent, competing technology.
3. Legal Framework Uncertainty
Current competition law does not provide a complete framework for assessing whether a license + acqui-hire combination constitutes a competition infringement. Although it has been emphasized that competition authorities should evaluate license agreements with an effects-based approach, there is no sufficiently clear legal assessment of how to handle the combination of a license and a personnel transfer.
In Türkiye, while the definition of concentration in Article 5 of Communiqué No. 2010/4, issued under Article 7 of the Competition Act, is quite broad, whether the NVIDIA-Groq transaction would be evaluated under this definition will initially be at the discretion of the TCA. International competition law trends indicate that such transactions will be subject to increasingly strict scrutiny.
VI. THE SITUATION IN TÜRKİYE AND THE POTENTIAL APPROACH OF THE TURKISH COMPETITION AUTHORITY
1. Comparative Analysis: Türkiye vs. International Level
In Türkiye, a competition dispute involving an AI chip technology transaction with a structure identical to the NVIDIA-Groq deal has not yet occurred. However, the TCA possesses the legal authority to review such transactions under the existing competition law framework. Internationally, especially in the EU and the U.S., license + acqui-hire transactions are becoming subject to stricter scrutiny.
In Europe, under the DMA, dominant platform companies are required to inform the Commission of their intended concentrations. This is an information requirement, not a request for clearance, allowing the Commission to monitor market trends and intervene with traditional competition law tools if necessary. In the U.S., while transactions below the HSR filing threshold are technically exempt, they can still be reviewed by the FTC using an effects-based approach.
The TCA tends to follow the approaches of the EU and U.S. competition authorities. In particular, amendments made to the Competition Act in 2023 allow for a broader assessment of the conduct of dominant companies. The TCA’s past reviews of dominant companies in the technology sector indicate that potential interventions are possible. For instance, in its investigations in the telecommunications and energy sectors, the TCA identified similar market foreclosure and competitor neutralization behaviors and imposed sanctions, such as the penalties applied to Turkcell’s exclusivity agreements.
In Türkiye, a broad interpretation of Article 7 of the Competition Act, which sets limits on mergers and acquisitions, would allow for the review of transactions similar in nature to the NVIDIA-Groq deal. The common trend across jurisdictions is to focus on the real competitive effects of a transaction rather than its legal structure. This is also a manifestation of the general legal principle of discerning the true intent behind transactions.
VII. CONCLUSION
The NVIDIA-Groq transaction is a striking example of the complex transactional structures developed by 21st-century tech giants to push the boundaries of traditional competition scrutiny. While formally structured as a license agreement and personnel transfer, its substance results in NVIDIA neutralizing a rising competitor by absorbing its key personnel and core technology, thereby eliminating its competitive potential.
As examined throughout this text, such de facto acquisitions have drawn the attention of global competition authorities, particularly in the U.S. and the EU. The adoption of the effects-based principle against tech giants by U.S. authorities and the EU’s closer monitoring of designated gatekeepers under the DMA indicate the direction of enforcement trends. Precedent cases like Microsoft-Inflection are concrete proof that regulatory bodies are now focusing on the de facto market impact of a transaction rather than its legal label.
From a Turkish perspective, although the TCA has not yet ruled on a transaction of this nature, its tendency to reference EU competition law practices and the potential for a flexible interpretation of the concept of “control” under the Competition Act raise the possibility that a deal like NVIDIA-Groq could be deemed a concentration on the grounds that it confers the ability to exercise decisive influence over the strategic commercial decisions of an undertaking.
Ultimately, the NVIDIA-Groq case reveals the complexity of protecting competition in a strategic market like artificial intelligence. The legal fate of this and similar transactions will not only set a precedent for the companies involved but also for all future merger and acquisition practices in the digital economy, testing the adaptive capacity of competition law in the face of the technological revolution.
B. KEY TAKEAWAYS
(1)Transaction Structure and Evasion of Scrutiny: The NVIDIA-Groq transaction was structured as a license + acqui-hire model instead of a traditional merger or acquisition. While formally not an acquisition, this complex structure in substance leads to NVIDIA gaining de facto control over the key technology and human resources of a rising and serious competitor, Groq. This has sparked debates that the transaction is an asset-light acquisition designed to evade traditional merger control thresholds and procedures.
(2)The Central Role of the Substance Over Form Doctrine: The core legal principle at the heart of this case is the doctrine of substance over form, which focuses on the actual competitive effect of a transaction in the market rather than its legal label. Global competition authorities may treat such deals as if they were sham transactions, scrutinizing their consequences in reinforcing NVIDIA’s market dominance and eliminating a potential competitor.
(3)Strategic Timing and the Inference Flip: The transaction occurred at a critical juncture as the AI market shifts from the training phase to the inference phase. Groq’s LPU technology posed a strategic threat to NVIDIA’s GPU architecture precisely because it offered a tangible technological advantage tailored to the needs of this new and growing inference market.
(4)Killer Acquisition and Stifling of Innovation: NVIDIA’s control over this technology and the team that developed it creates the risk that access to the technology will be foreclosed to third parties and used exclusively to strengthen NVIDIA’s CUDA ecosystem. This is considered a killer acquisition that could lead to a reduction in market competition and innovation.
(5)Reinforcement of Existing Dominance: The competition law risk of the transaction is amplified by NVIDIA’s existing dominant position of 70-95% in the AI chip market and its CUDA software ecosystem, which creates a lock-in effect for developers. This transaction has the potential to further solidify NVIDIA’s existing market power.
(6)Precedent Decisions and Potential Remedies: Cases like Microsoft-Inflection and Microsoft-Activision Blizzard offer important precedents. The approach taken in the Microsoft-Activision decision, in particular, suggests that behavioral remedies—such as requiring NVIDIA to license Groq’s LPU technology to third parties on FRAND terms—could be a solution instead of outright prohibition.
(7)Potential Approach of the Turkish Competition Authority: Although no such transaction has been adjudicated in Türkiye, the TCA’s tendency to follow EU practices and the potential for a broad interpretation of the concepts of control and concentration under the Competition Act would allow for such a transaction to be subject to review. It is plausible that the Authority would initiate an investigation, considering the transaction’s potential to confer decisive influence over the strategic business decisions of an undertaking.
(8)Debate on the Inadequacy of the Legal Framework: This case highlights that existing notification thresholds, such as those under the U.S. HSR Act, may be insufficient to capture strategic transactions that can fundamentally alter market dynamics but fall below monetary thresholds. This has triggered a global debate on the adaptation of antitrust laws to modern technology markets.
(9)Restriction of Competition in the Skilled Labor Market: The talent acquisition dimension of the transaction shifts competitive concerns from the product market to the market for skilled labor. NVIDIA’s hiring of Groq’s key engineering team means it controls not just a technology, but also the human capital that could advance it further or develop a similar one at a rival firm.
(10)Precedent for the Future: The NVIDIA-Groq case tests the adaptive capacity of competition law in strategic and rapidly evolving markets like AI. The legal outcome of this transaction will set a precedent for the future M&A strategies of tech giants and will shape the global legal framework for protecting competition in the digital economy.
