False Comfort of Nascent Market: Re-assessing the Position of Strength in the Age of AI Market

[Shivam and Priyanshi are students at Dharmashastra National Law University.]

As the recent Competition Commission of India’s (CCI) Market Study on Artificial Intelligence (AI) and Competition (CCI’s AI Study) reveals that the AI sector has “rapidly transitioned from experimental use to mainstream adoption”, with India’s AI market growing from USD 3.20 billion in 2020 to USD 6.05 billion in 2024 and projected to reach USD 31.94 billion by 2031. Despite this staggering growth and concentration, dominant AI players invoke the excuse of the AI market being a “nascent market” to deflect stricter antitrust scrutiny. The argument goes: AI is new, unpredictable and evolving so rapidly that market power today may not exist tomorrow. However, competition regulators must not be relaxed by this “false comfort”. 

The exponential growth trajectory of the AI markets directly contradicts claims of “nascency” and instead signals a maturing, highly concentrated market dominated by a handful of upstream players that demand early antitrust intervention rather than regulatory complacency. As of 2025, just a handful of firms: OpenAI (with Microsoft), Google (DeepMind/Gemini), Anthropic & a few others command the vast majority of generative AI ecosystem, controlling not only the most powerful foundation models but also the underlying compute infrastructure & critical training datasets.

In the MCX Stock Exchange Limited and Others v. National Stock Exchange of India Limited (NSE), CCI rejected the “nascent market” defense as taken by NSE in the currency derivatives segment and held that "nascence" refers to the immediate existence after birth, while "infancy" is the time after. Markets are considered nascent during the first few months, followed by infancy for another year, then maturity when the market is fully developed. The AI market began evolving in the late 2010s, peaked in the early 2020s (notably 2023), and therefore is no longer in its nascent phase.

This blog argues that, in the age of AI, market definition must adapt, separating hype from structural reality and ensuring that claims of “nascency” do not become convenient cover for an anti-competitive market. 

Defining Relevant Market in the Era of AI Services with Unclear Boundaries

The emergence of nascent competition is a double‑edged sword: on the one hand, it can spur innovation and improve quality; on the other, if its risks are not addressed, it can suppress innovation, reduce consumer choice, and entrench market concentration. It is well established in competition law that market definition should be considered in the context of the theory of harm in the case at hand  and assessed in light of the overall economic context. Historically, the competition authorities have defined relevant markets narrowly, for example, in M/S Ess Cee Securities Private Limited v. DLF Universal Limited, CCI defined the relevant market narrowly as the provision of services relating to development and sale of residential apartments in Delhi. By contrast, AI markets are both exceptionally broad, spanning diverse products and technologies, and at the same time highly concentrated.  

Foundation models may not be fully interchangeable, or interchangeable at all, depending on their training data. They may be general purpose or domain specific. Emerging technologies and innovation at an unprecedented pace in the realm of AI are leading to unclear boundaries, especially where two different AI models are linked together. For example, large language models (LLMs) differ from cloud-based AI systems.

The CCI’s AI Study’s 7-layer AI stack is analytically decisive for market-definition questions and argues against treating AI as a single product market.  The CCI distinguishes the data and compute deployment frameworks from foundation models and the downstream layers of fine-tuning, user interaction, and governance. In the upstream layers, control over high‑quality data, compute resources, and specialized AI chips is concentrated among a few global providers Amazon Web Services (AWS), Microsoft Azure, Google Cloud, and NVIDIA, creating structural bottlenecks and limiting access for smaller Indian firms.  At the development layer, the number of foundation‑model developers remains small, with global incumbents such as OpenAI, Google, Meta, and Anthropic dominating model creation and licensing.  In the downstream layers, firms that fine‑tune or deploy these models often become locked into providers’ ecosystems through proprietary application programming interfaces (APIs), closed‑source architectures, and cloud dependencies.  This vertical layering of dependence leaves little substitutability across layers, creating intra‑stack dependencies that can lead to input foreclosure, ecosystem lock‑in, and self‑preferencing. 

Accordingly, the relevant market should be defined narrowly at the functional layer where exclusionary control actually operates, for example, foundation‑model provision or cloud GPU compute for large‑scale model training, and evidence of control should be adduced at that specific layer, rather than relying on a diffuse AI market definition that blurs foreclosure and leveraging across adjacent inputs.  If the relevant market is not defined narrowly, firms may escape liability for anti‑competitive behaviour. 

New Factors of Dominance Emerging with Tech Innovations in AI Sector to be Taken into Account

It is well settled that holding dominant position under Section 4 of the Competition Act 2002 (Competition Act) is per se not prohibited by law, but abusing the same is punishable under the law. In digital platform competition, the roles of gatekeepers, network effects, and tipping—combined with the tendency of digital markets to extend into adjacent markets—can blur the distinction between legitimate competition on the merits and abuse of dominance, particularly when “nascency” is invoked as a defence. However, the CCI’s AI Study provides concrete evidence that two most decisive inputs in the contemporary AI markets are: access to data and access to compute. Exclusive control over training data and cloud capacity creates durable market power.  The empirical findings are notable: 68% of surveyed startups identified data availability as a key barrier; 61% cited the cost of cloud services and talent shortages as critical constraints.  In India, the cloud market is led by AWS (~32.6%), Microsoft Azure (~20.8%), and Google Cloud Platform (~11.5%), with the remainder dispersed among smaller providers. 

The CCI’s AI Study further warns that “the control of a few large firms across the AI stack may create barriers to entry for smaller players, as established entities own vast, high‑quality datasets that may not be accessible to new entrants, and high training costs tend to favour well‑funded incumbents". As the CCI cautions, this asymmetry has produced ecosystem lock‑in: control over data, infrastructure, and development layers by major technology companies limits transparency, reduces consumer choice, and creates dependence on a few hyperscalers such as AWS, Microsoft Azure, and Google Cloud.

Data serves as the foundation for AI and past behavioural patterns preserved in data are necessary to train and increase the efficiency and accuracy of algorithms' predictions. This source data, combined with third-party data, is critical for improving existing AI, performance, particularly in generating accurate behavioural insights. At present, OpenAI’s GPT models—trained on vast datasets using Microsoft Azure’s cloud—are significantly ahead in capabilities, an advantage that startups cannot easily match. This technological lead attracts substantial investment, reinforcing OpenAI’s market position and valuation (reported at around USD 80 billion). Scale and advanced resources give such enterprises a persistent advantage and enable the creation of barriers for rivals.

The CCI has observed that, in a data‑driven ecosystem, unreasonable data collection and sharing may grant undue advantages to dominant players, as recognized in In Re Privacy Policy for WhatsApp Users. Similarly, in In In re: Delhi Vyapar Mahasangh and Flipkart Internet Private Limited, the CCI recognized how data collection and enterprise scale can reinforce market power. Massive volumes of data, combined with machine learning and AI capabilities, strengthen the position of data‑driven technology enterprises. When users transition to new platforms, a lack of data portability and interoperability creates significant barriers to entry and high switching costs. New entrants therefore often need to introduce genuinely disruptive technologies to overcome these frictions and attract users from established markets. Given these dynamics, competitors may struggle to match the cost structures and capabilities of products released by large technology firms.

The CCI’s AI Study finds that deep integration of foundation models into enterprise workflows, for example, embedding model outputs into productivity suites or integrating via APIs, creates lock‑in effects and raises non‑monetary switching costs. According to the CCI’s AI Study’s startup survey, 37% of respondents identified AI‑facilitated collusion as a concern, 32% cited price discrimination, and 22% pointed to predatory pricing.  These risks are amplified when dominant firms enter exclusive data‑access agreements to train large language models or tie separate products, making potential abuses harder to detect and deter.  As an illustration from Indian competition enforcement, the NSE was previously found dominant in part due to extensive vertical integration—showing how control across layers (from data to high‑technology services) can entrench market power. Accordingly, “nascency” should not be the sole factor in assessing dominance, particularly in AI markets characterized by rapid technological change and vertical integration across critical inputs.

Superior technology is considered an indicator of dominance in the market by the CCI in Shri Shamsher Kataria v. Honda Siel Cars India Limited. Recently, Google’s DeepMind’s solution to the protein-folding problem through AlphaFold was a calculated decision that gave it a formidable edge in bioinformatics and pharmaceutical AI. The CCI’s AI Study’s primary research reveals that approximately 67% of India AI startups operate at the application layer, building end-user solutions on the top of existing models while only about ~3% reported any activity in developing foundation models. Further 20% function in the data layer and around 10% in the compute and infrastructure segment. This shows that upstream layers: data, compute and foundation model development remain almost entirely dominated by global incumbents such as OpenAI, Google, Meta and Microsoft. 

This further reflects how AlphaFold do not simply create superior research capacity but also, they control foundational inputs that downstream innovators cannot replicate or substitute. In short, the CCI’s AI Study’s evidence shows that the AI ecosystem in India has already evolved beyond nascency, exhibiting entrenched upstream dependence, information opacity and reduced substitutability, indicating digital monopolies. 

Other factors include economic power of the enterprise, including commercial advantages over competitors, and market structure in size of the market. For example, OpenAI signs deals with news publishers, Reddit, Stack Overflow to use unique content for training because of its economic capacity to enter into such agreements. Competitors without these deals can’t match training data quality. Recently, Amazon dominates voice AI and cloud AI by sheer size. AWS holds 32% of the global cloud market, giving it unmatched scale to train massive AI models.

In the Belaire Owners' Association v. DLF Limited,  the CCI considered “dependence on customers” as relevant to dominance. Given the limited interchangeability of LLM services, users risk lock‑in to particular ecosystems, with providers continually innovating to reinforce user dependence.

For example, OpenAI’s integration into Microsoft 365 CoPilot is an example of ecosystem lock-in. Due to internal reliance on particular APIs and data flows, switching can be expensive for businesses once embedded in this ecosystem. In the same way, companies are completely locked in to Microsoft's environment because OpenAI's models are only restricted to Microsoft cloud only. As the incumbents (OpenAI) rely on foundational models, infrastructure, and platforms provided by a few large players (Microsoft CoPilot), the lack of visibility into how these systems operate, including opaque APIs, black-box algorithms, and unclear pricing structures, creates uncertainty and dependency. This lack of transparency may hinder innovation and may complicate compliance. Simultaneously, reduced choice in infrastructure and model access limits the ability of smaller players to compete effectively, forcing them into ecosystem lock-in and reducing overall market dynamism.

No other cloud provider can currently host OpenAI’s best models, conferring a unique AI cloud advantage on Microsoft. While this may make business sense, it raises concerns about decreased substitutability and potential innovation stagnation at the ecosystem level. Over‑reliance on a single stack may direct future developments toward a few gatekeepers and create systemic fragility, where a single failure (e.g., a Microsoft outage or API change) can disrupt entire AI workflows. The CCI has also noted that data can further strengthen and entrench network effects, limiting inter‑platform competition.

Although mergers and partnerships can enhance technological development, they may raise concerns about reduced substitutability and innovation if they entail exclusivity over essential inputs. As the CCI’s AI Study highlighted, exclusive supply arrangements, distribution restrictions, or IP exclusivity can deny market access and heighten dependencies. Given the high concentration in compute and cloud infrastructure, such arrangements may reinforce incumbents’ dominance and deter entry by increasing switching cost. Future developments could then tilt toward incumbents, while a single link failure (such as a Microsoft outage or API change) could disrupt AI workflows across industries.

Conclusion

The “false comfort” in AI is the belief that the market remains open because entry barriers are high and ecosystems are closed. Price discrimination and predatory strategies are not merely theoretical in this market; rather, the legal implications are twofold. First, regulators must remain attentive to differential pricing that targets vulnerable cohorts and may prohibit new entry. Second, economic evidence should be adduced to show how predatory pricing is anti-competitive, which the CCI’s AI Study assists by providing typical algorithmic behaviours. Such practices significantly distort market dynamics by limiting competition, innovation, and consumer choice, creating conditions that will cause long-term AAEC in the market. This directly impacts cost-sensitive, smaller developers by depriving them of cost-effective options and increasing their dependency, which restricts innovation and discourages new entrants from developing superior alternatives.

This results in fewer advancements in AI technologies and reduces the availability of diverse, high-quality, and affordable solutions for consumers. A reduction in choice will lead to consumer harm, contrary to the legislative intent of the Competition Act, which seeks to safeguard consumer interests by fostering healthy competition in the market.

There is a rising need for multidisciplinary review involving data, telecom, and other sectoral regulators alongside competition authorities, given the multidimensional nature of these markets. The CCI’s AI Study calls for the self-audit of AI systems for competition compliance. The annexed self-audit checklist provides an indicative template for organizations to assess algorithmic transparency, data access, and competitive neutrality. Adopting this checklist as part of internal compliance protocols can operationalize responsible AI governance and enable ex ante detection of competition risks within algorithmic design and deployment. Further, it advocates transparency measures to reduce information asymmetries, institutionalize regulatory coordination, and enhance international cooperation through platforms such as the OECD, ICN, and UNCTAD. As concrete next steps, the CCI should organize focused advocacy workshops on “AI and Competition Compliance,” establish a dedicated think tank to draw expert insights on evolving AI market structures, and strengthen its technical capabilities to evaluate algorithmic conduct and market outcomes. Many mature competition regimes have adopted ex ante regulations to address data-related concerns and the gatekeeper status of some online platforms. The CCI will also strengthen its technical capabilities to evaluate algorithmic conduct and market outcomes. Fast-changing markets require timely decisions to avoid further abuse of dominance in such markets.