The Myth of Nascency: Recalibrating Competition Law in the Age of AI Markets

I. Introduction

The rapid commercialisation of artificial intelligence has generated a paradox: technological novelty is combined with structural concentration. Companies often claim that the AI markets are in their infancy, which can be used to postpone the regulatory review and lead to weakening the competition law enforcement.

However, the empirical indicators suggest otherwise: the AI market in India is no longer in the business of boutique pilots, and there is a significant increase in market size and concentration, which discredits the defence of being new.

The market analysis of the Competition Commission of India (CCI) shows that the industry is experiencing rapid growth and consolidation of the major inputs of data, compute and foundation models, which point to regulators using nascency as a reason to intervene promptly instead of complacency.

Through this blog, the author will address the issue of how AI is affecting the market, highlight how another market deals with the same issue, and suggest some policies that could be adopted in India.

II. Why the “Nascent Market” Defence Misfires

Nascency has been understood to imply fluidity of markets, where the leader of the day can quite easily be overthrown tomorrow. This ideology normally causes the competition-regulating bodies to be cautious when regulating the emerging industries. Nevertheless, the assumption is not applicable in AI markets.

The forces that make AI a potent technology are powerful network effects, data economies of scale, and extremely high model training and computing fixed costs. Those characteristics enable the early movers to become advantageous very fast. Switching costs are immense due to the fact that firms use exclusive datasets, access to advanced computing infrastructure, and AI that is fully interwoven with business processes.

For instance, the UK Competition and Markets Authority has observed that foundation models require large computing resources and extensive data access, which provide a competitive edge for early adopters and raise barriers to entry. In practice, companies like OpenAI and Google benefit from access to hyperscale clouds, internal data sets, and frequent user interaction, which further extends feedback loops and strengthens early market advantages.Existing companies are in a position to align these benefits along the value chain as they have access to combine proprietary data, computing power, and distribution platforms that make them inaccessible to new firms. They can close the door on competitors by denying them critical information, driving up switch costs by locking themselves into an ecosystem, or monopolising particular industries, thus making it hard to compete with new entrants. As highlighted in the CCI’s AI study, the risk lies in treating AI as merely nascent, which may hide the early formation of durable market power and lead to regulatory blind spots.

III. Functional-Layer Market Definition

Courts and regulators should not perceive AI as a homogeneous and single-market. AI has many functional layers, which have competitive conditions that differ at different levels. The seven-layer conceptualisation of AI by the CCI- that spans raw data and compute infrastructure to deployment, user interaction, and governance - offers a feasible conceptualisation of competition analysis.

In the Indian competition law, market definition should indicate where the harm of competition exists. In this regard, market definition should be driven by the relevant functional layer (i.e., data, compute, or API), and, accordingly, measures should be taken with respect to the elements on which the exclusion or foreclosure allegations are made.

For example, an artificial intelligence (AI) company that operates a widespread online platform and gathers user data in huge volumes. In the case it withholds this data to train its own artificial intelligence models and deprives competitors of similar data, the damage will be at the data layer, rather than at the AI service market overall. Looking at the issue through a wide “AI services” market would wrongly suggest that competitors have many alternatives.

Hence, Market definition at the data layer ensures that the control of the firm is clear, and it can be properly assessed by regulators concerning dominance and foreclosure in the context of competition law.

IV. New Dominance Indicators in AI markets

Conventional measures of dominance in Section 4 of the Competition Act, 2002 (Act), like market share, barriers to entry and price control, are still applicable but do not apply adequately to the AI markets where dominance is often attained by manipulating inputs and technical structure instead of prices.

Therefore, there is a need for considering AI-specific markers of dominance, such as ownership of proprietary training data, access to advanced computing and specialised chips, API-level lock-ins and restricted portability, and platform integrations which increase switching costs due to technical dependence.

This strategy is reinforced by the CCI market analysis, which revealed data access and the high cost of the cloud as the major obstacles to Indian AI start-ups. Almost two-thirds of start-ups surveyed attribute the availability of data as a significant restraint of their activities, and the upstream cloud market concentration makes the situation even worse.

Similar concerns were echoed by international policy analyses, with the OECD noting that concentrations in cloud infrastructure, compute resources, and foundational models can lead to dependency on more than one resource and be barrier for small firms to enter. As a result, control over the AI inputs at the upstream stage of production, rather than downstream pricing, has been identified as the source of competitive advantage in AI markets.This fact shows that the implementation of Section 4 should go beyond price-focused tests and acknowledge the ability of dominance in AI markets through the maintenance of data, compute, and ecosystem dependencies.

V. How the European Union Balances Ex-Ante Regulation with Competition Concerns?

A different route was adopted by the European Union (“EU”); it is a mixture of ex-ante regulation and competition enforcement. Under the Artificial Intelligence Act, AI systems are grouped by the level of harm they might cause. Acts that are clearly unacceptable are banned, while high-risk systems are allowed but subject to strict rules.

The law also requires documentation (like model cards and risk assessments) and transparency for certain models, and it sets up governance both at the national levels of various EU nations and through an EU office for general-purpose AI. These up-front rules on transparency, design and deployment help reduce information gaps that make antitrust investigations difficult, and they work together with later competition investigations.

The EU also uses other tools alongside the AI Act. The Digital Markets Act targets unfair behaviour by big platform “gatekeepers” (usually indulging in self-preferencing or forced bundling). At the same time, public investment is being used to ease supply bottlenecks where money is being directed to build high-performance computing capacity, so Europe depends less on these non-EU Hyperscalers (American companies like Oracle, IBM, etc that provide massive scale cloud computing services but often raise data sovereignty concerns in Europe).

Together, these measures show that law alone isn’t always enough when the real bottlenecks are physical or financial. Industrial policy, public funding and strategic investment can help break structural supply constraints that otherwise reinforce incumbent dominance. This coordinated approach, combining regulation, competition enforcement, and industrial support, offers a model for other jurisdictions that want to stay competitive and shape technology strategically.

VI. Policy Prescriptions for India

India should not replicate the European Union approach to regulation in its wholesale form: a more nuanced, situation-specific policy will have superior results because the EU’s ex-ante regulatory framework is designed for mature markets characterised by entrenched digital gatekeepers, whereas Indian digital markets remain structurally evolving and constrained by infrastructural and institutional limitations. As recognised in the CCI report and NITI Aayog policy frameworks, a rigid regulatory transplant risks stifling innovation and entry, making a calibrated, context-specific approach more suitable.

The CCI must start with the adoption of a layered market definition under section 19(5)–(7) of the Act, whereby investigations into the alleged exclusion/foreclosure occur at a specific functional level, including foundation models, cloud-computing-based GPUs, or proprietary dataset markets instead of the overall AI market.

In conjunction with this, precise disclosure and auditor controls may be provided by firms as per section 19(4) of the Act, through confidential model cards, provenance reports and accredited technical audits with stringent non-disclosure safeguards. This would enable the CCI to evaluate harm in competition without compelling disclosure of trade secrets to the public.

In addition, there should be portability and interoperability norms through promotion of multi-cloud strategies, model weights and dataset metadata open interfaces, which will reduce switching costs and artificial lock-in. In the case of firms that hold the control of critical inputs, the narrowly-specific ex-ante obligations might be justified to avoid self-preferencing and exclusive arrangements formed, and such obligations might also be formulated taking into account India's market structure.

Last but not least is the establishment of cross-regulatory task forces under section 18 of the Act, and long-term capacity building: open coordination forums between competition, data protection, telecom, and sectoral regulators, as well as investment in technical labs, think-tanks and specialised training, would make the economic and technical inferences about algorithms based on evidence and not abstraction.

Simultaneously, the issues associated with the trade secrets, intellectual property, and innovation incentives cannot be disregarded. An effective regulatory design needs to create a balance between disclosure and security, which needs to be based on such mechanisms as metadata-level disclosures, third-party confidential auditing, and reversible remedies, such as access obligations, interoperability, and behavioural constraints.

Structural remedies must be extraordinary and must only be implemented when there is harm, and in circumstances where harm is proved, access, portability and non-discriminatory API conditions should be accorded priority. Firms, on their part, may decrease regulatory risk through documenting the data lineage systematically, making it portable where possible, and having their systems accredited with audits. The continued international coordination will occur via the platform like the OECD and bilateral regulatory discourses, whereby the best practices will be easily transferred without ignoring local market dynamics.

VII. Conclusion

To conclude, the “nascent market” defence in the context of AI argues that technological novelty should not be mistaken for competitive fluidity. The central claim advanced is that AI markets, particularly in India, already display structural concentration driven by control over data, compute, and foundational infrastructure. By examining why nascency misfires, adopting a functional-layered approach to market definition, and proposing AI-specific indicators of dominance, the analysis demonstrates that conventional competition law tools must be contextually recalibrated rather than cautiously deferred.

The comparative assessment of the European Union illustrates that competition enforcement can coexist with targeted ex-ante regulation and industrial policy to address early chokepoints. For India, the way forward lies not in regulatory replication but in adopting layered market analysis, calibrated disclosure, interoperability norms, and institutional capacity-building within the CCI.

Looking ahead, the real risk is regulatory complacency. If AI markets are continually treated as “emerging”, early advantages may harden into irreversible dominance, foreclosing competition before it meaningfully emerges. This raises a critical question: should competition law wait for harm to fully materialise, or intervene early to preserve contestability? The answer will shape not only AI governance, but the future relevance of competition law itself.