Transfer Pricing For AI, Cloud, And Data-Driven Service Centres: Benchmarking The Next-Generation Global Capability Centre

Introduction

The Global Capability Centres (GCCs) in India have advanced significantly since the time when they were in the form of voice-based call centres and back office operations, during the early 2000s. Many of these centres have been set up as centres of excellence in the areas of artificial intelligence, cloud engineering, and data science, enabling the processes of model training, database management, and platform development for the organization as a whole. This represents a departure from the practice of implementing transfer pricing framework in connection with ordinary activities involving low-level risk ICT, enabled service providers, who would be compensated at low margins.

In this blog, we will therefore consider the issues that are related to benchmarking the processes of support services and R&D in connection with the DEMPE framework, what problems arise with respect to intangible ownership and markup, as well as what documentation issues are likely to result in disputes.

Legal Provisions

1. The Transfer Pricing Guidelines of the OECD

The first chapter among the OECD Transfer Pricing Guidelines requires an accurate relevancy between the transactions being dealt with through functional analysis, asset analysis, and risk analysis. The sixth chapter of these guidelines specifies the concept of DEMPE, which is important in determining what the actual nature of the transactions is. The eighth chapter comprises guidelines regarding COST contribution arrangements, thus showing some provision for such transactions.

2. The Indian Domestic System

The system prescribed in India for determining transfer prices has been provided in the Income Tax Act, 1961 through sections 92 to 92F, whereas recently this system of transfer pricing has been reclassified in accordance with the provisions of the Income Tax Act, 2025 in sections 161 to 173. It is necessary to perform the calculations for income received from “international transactions” occurring between “related enterprises” according to the arm’s-length principle with the applied regular procedures (CUP, resale price, cost plus method, profit split, TNMM, or any other one) supported by the master file, local file, and country-by-country report corresponding to the recommendations of BEPS Action 13. The introduction of a new multi-year method of evaluating transactions will enable doing the arm’s-length determination in one year of operation, and then carrying it forward to similar transactions occurring in subsequent years.

2. Safe Harbour and APA Developments

The CBDT keeps enlarging the scope of Safe Harbour Rules stated in section 92CB (adapted by the 2025 Act) for IT/ITES companies, and the Union Budget 2026 proposals suggest applying more homogenous margin and multi-year certainty concept, which is important for AI and cloud companies because the nature of their work and type of services changes much more frequently than the classification of safe harbours expects, making it necessary for companies to check their current margins using information from the respective CBDT notification before claiming they qualify for the respective safe harbour.

Legal Analysis

1. Cost Centre to Value Creator

India has always recognised the fact that services commonly referred to as "information technology-enabled services" cannot always be treated as interchangeable in terms of comparability. In “Rampgreen Solutions Pvt. Ltd. v. CIT”, the Delhi High Court stated that a knowledge process outsourcing (KPO) service provider cannot be compared with a business process outsourcing (BPO) service provider performing standard tasks, even if both services are included under the umbrella of ITES. Comparative analysis should have a basis in functional similarity rather than in the fact that both services qualify under the same name. The focus on artificial intelligence amplifies this issue even more, such as a service centre performing model training or data curation and working with valuable content is performing jobs far removed from standard BPO or KPO services and cannot be compared with the trials of relatively traditional ITES service providers, otherwise the true nature of its work may be seriously undervalued.

2. DEMPE and Ownership of AI Intangibles

Depending on the OECD's functional definition, datasets for training, model parameters, and specialized systems can constitute intangibles despite their formal registration. The major issue for DEMPE is whether the Indian entity merely acts according to the given instructions and what justifies the application of the regular return. If personnel in India make decisions related to architecture design or data use without supervision or in respect of which they are responsible for potential losses, the entity is entitled to an intangible profit share, not merely in accordance with the method of providing “support services.”

3. Mark-up and Method Choice

The cost-plus TNMM approach is purportedly the most widely-applied approach for IT/ITES and knowledge-based captives since truly comparable companies in the field of AI development are rare. Thus, the taxpayers tend to resort to IT service or KPO companies that can be only somewhat similar to the entity representing AI development functions; as a result, the levels of mark-ups are either insignificant or insufficiently justifiable from the market-environment perspective. For situations in which a centre is involved only in simple tasks with limited risks and operates in a well-structured manner, it is not a problem to argue in favour of the cost-plus method; at the same time, if a centre assists in the creation of collections of knowledge in a more significant manner, the profit split or contribution analysis becomes more applicable as more detailed considerations of investment have to be made.

4. Use of the Same Types of Platforms

AI and cloud-based centres are benefiting from a single system of platforms, using computer power and models which are accessible to various countries. Shared AI infrastructures bring about a comprehensive issue of allocation. Cost Contribution Arrangements (CCAs) need to properly allocate the costs and expected benefits among all participants involved. In case the allocation keys are not consistent with the actual contributions made by the DEMPE entities, the tax authorities are likely to object to this arrangement.

Case Laws

“Rampgreen Solutions Pvt. Ltd. v. CIT,” (2015) 377 ITR 533 (Del): Delhi high court held that while analyzing comparable entities, a distinction must be made between KPO and BPO service providers and KPO entities mentioning that the very fact that the certain entity belongs to ITES industry does not lead to the conclusion that two entities have functional similarities.

“McKinsey Knowledge Centre India Pvt. Ltd. v. PCIT” (Del HC): Delhi high court in this case upheld Rampgreen ruling indicating that entities comparing on the basis of similar industry labels must be disregarded based on the fact that what matters is the scale of operations.

“Sony Ericsson Mobile Communications India Pvt. Ltd. v. CIT” (Del HC) and the Revenue’s challenge against marketing intangibles in GSK transfer pricing case are indicative of an overarching principle: the Indian tax authorities have been recharacterizing onshore activities as rendering intangibles to offshore associated entities based on the level of economic activity rather than the form of contract.

Practical Implications

AI service centres of multinationals may be transformed from regular support to innovation centres, while transfer pricing documentation cannot accommodate those changes. The inconsistency of recognition across the jurisdictions can be used by tax offices to challenge pricing. Minutes of a board meeting, engineering documents, product documentation should therefore be treated as transfer pricing documents. In addition, the absence of established benchmarks per the intangible assets in AI leads to the necessity of using the judgement-based methods, meaning the impossibility of using cost-plus-style methods.

A practical checklist for transfer pricing in AI-driven and data-driven service centres:

• Perform a review of actual functions against the contractual definition each year.
• Identify intangible assets linked to AI and document DEMPE for every one of them.
• Re-assess the transfer pricing method when the DEMPE contributions seem to be meaningful.
• Use practically comparable benchmarks to justify any deviations from IT/ITES comparables.
• Document the arrangements of cost contributions with regard to similar AI infrastructure.
• Make sure that internal documents are in tune with the documentation of transfer pricing before audits.
• Consider using APAs or Safe Harbours only if the functions are not changing over time.

Conclusion

The growth of AI, cloud computing, and data-driven service centres has rendered obsolete the traditional transfer pricing playbook for captive back office functions. The DEMPE framework shifts the focus from legal ownership and contractual terms to the substance of who creates, improves, maintains, manages, and uses the intangibles that AI-based functions create. The best course of action is not to wait for a dispute but to proactively review the classification, method, and documentation as you grow your centre, all the while gathering the requisite governance records and applying the same substance-over-form test, which Indian courts have been applying to ITES and KPO disputes for a long time.

Footnotes

^{1  OECD, OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations chs. I, VI & VIII (2022).}

 ^{2 OECD/G20 Base Erosion & Profit Shifting Project, Aligning Transfer Pricing Outcomes with Value Creation: Actions 8–10 – 2015 Final Reports (OECD Publishing 2015).}

 ^{3 OECD/G20 Base Erosion & Profit Shifting Project, Transfer Pricing Documentation and Country-by-Country Reporting: Action 13 – 2015 Final Report (OECD Publishing 2015).}

^{4 OECD, Pillar One—Amount B: Simplified and Streamlined Approach for Baseline Marketing and Distribution Activities (2024).}

^{5 Income-tax Act, No. 43 of 1961, §§ 92–92F (India).}

^{6 Income-tax Act, 2025, §§ 161–173 (India).}

^{7 Income-tax Rules, 1962, rr. 10A–10E, 10D & 10DA (India).}

^{8 Safe Harbour Rules, Income-tax Act, 1961, § 92CB (India), as amended.}

 ^{9 Rampgreen Sols. Pvt. Ltd. v. Comm'r of Income Tax, (2015) 377 I.T.R. 533 (Del.).}

 ^{10 M/s McKinsey Knowledge Ctr. India Pvt. Ltd. v. Pr. Comm'r of Income Tax, ITA Nos. 461 & 526 of 2017 (Del. H.C. Aug. 9, 2018).}

 ^{11 Sony Ericsson Mobile Commc'ns India Pvt. Ltd. v. Comm'r of Income Tax-III, (2015) 374 I.T.R. 118 (Del.).}

^{12 Finance Bill, 2026 (India); Central Bd. of Direct Taxes, Safe Harbour Notifications (India).}

https://chatgpt.com/s/m_6a7afcb40f0c81919c236c7270e392ea

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.