For more than two decades the Indian transfer pricing regime laboured under a paradox. The arm's length principle, imported into the Income-tax Act, 1961 by the Finance Act, 2001 and housed in Sections 92 to 92F,2 was elegant in statement and intractable in application. It asked a deceptively simple question, namely what would independent parties, dealing at arm's length, have charged; and it answered that question through a machinery of comparables, filters, adjustments and tolerance bands so contentious that transfer pricing became, for a season, the single largest source of tax litigation in the country. The captive technology sector bore the brunt of it. Software development houses and business process units, though performing routine and low-risk functions, found themselves defending margins year upon year before Transfer Pricing Officers, the Dispute Resolution Panel and the Tribunals.
The State's response evolved along two tracks. The first was the safe harbour, introduced under Section 92CB of the 1961 Act and the rules framed thereunder, which offered eligible taxpayers a bright-line margin that the Revenue undertook to accept without inquiry. The second was the Advance Pricing Agreement under Sections 92CC and 92CD, a negotiated compact fixing the pricing methodology in advance for a term of years. Both were instruments of certainty; both, in their original design, were hedged with thresholds and conditions that kept the larger taxpayer outside their embrace.
It is against that history that the Income-tax Act, 2025 must be read.1 The new statute, which received legislative sanction in 2025 and takes effect on 1 April 2026, was conceived as a work of consolidation and simplification rather than of doctrinal upheaval. Yet in the field of transfer pricing, and read alongside the Union Budget 2026-27 and the Income-tax Rules, 2026, its effect is more ambitious than simplification alone. The Finance Minister's Budget announcement of 1 February 2026, that the safe harbour threshold for Information Technology Services would be raised from ₹300 crore to ₹2,000 crore and that a single margin of 15.5 percent would apply across the sector,3 signalled a decisive shift of policy: from adjudicating the arm's length price transaction by transaction, towards administering it by formula and at scale. That shift, and its consequences for the Advance Pricing Agreement, are the subject of what follows.
A word on sources is owed at the threshold. The statutory provisions are settled, and the operative safe harbour rules are settled also. The Income-tax Rules, 2026 were notified on 20 March 2026, after public consultation, and take effect with the Act on 1 April 2026, supplemented by the frequently asked questions and guidance notes issued by the Board on 23 March 2026. The margins and thresholds discussed below are therefore not proposals but enacted law. Certum est quod certum reddi potest, that is certain which can be rendered certain; and here the draftsman's ink is now dry.
The first thing the practitioner will notice is that the furniture has been moved. The transfer pricing provisions, formerly Sections 92 to 92F of the 1961 Act, now occupy Sections 161 to 173 of the Income-tax Act, 2025.1 The renumbering is not mere housekeeping. Section 161 restates the charging rule, that income from an international transaction or a specified domestic transaction shall be computed having regard to the arm's length price. Section 162 consolidates into a single provision the definition of associated enterprise; Section 163 restates the international transaction, retaining and in places widening its residuary reach; Section 164 addresses the specified domestic transaction; and Section 165 gathers the methods by which the arm's length price is to be determined. Sections 171 and 172 carry forward the documentation obligation and the accountant's report. The substance of the old law survives; its map has been redrawn for legibility.
For present purposes three provisions are cardinal. Section 167 confers upon the Board the power to make safe harbour rules and, importantly, defines a safe harbour as the circumstances in which the income-tax authorities shall accept the transfer price, or the income attributable under Section 9(2), declared by the assessee. The imperative is deliberate. Under the 1961 dispensation the mandatory character of acceptance lived principally in the rules; the 2025 Act elevates it to the statute itself, so that once the prescribed circumstances obtain, the authority is denuded of discretion to recompute. When the safe harbour is validly claimed, cadit quaestio, the question of the arm's length price simply falls away.
Section 168 re-enacts the Advance Pricing Agreement. The Board, with the approval of the Central Government, may agree in advance the arm's length price, or the manner of its determination, in relation to an international transaction, and equally the income reasonably attributable to the Indian operations of a non-resident. The agreement is valid for a period not exceeding five consecutive tax years, binds both the taxpayer and the Revenue in respect of the covered transaction, and ceases to bind if there is a change in the law or in the facts having a bearing upon it.1 Section 169 supplies the machinery of implementation: where income is modified in consequence of such an agreement, the person, and now the associated enterprise, may furnish a modified return limited to the agreement within three months of the end of the month in which it is entered. The binding force of the compact is an application of pacta sunt servanda to the fiscal sphere, tempered by the rebus sic stantibus qualification that a material change in circumstances dissolves the bond.
One further thread of the reset deserves mention, because it explains the animating philosophy. The Finance Act, 2025 had already introduced, for the 1961 Act, the option to apply the arm's length price determined for a base year to similar transactions in the two succeeding years, a species of multi-year or block determination designed to spare the taxpayer and the administration the annual ritual of re-benchmarking.4 The 2025 Act inherits that logic and amplifies it. When one places the block determination beside the five-year safe harbour and the five-year Advance Pricing Agreement, a coherent design emerges: the movement is uniformly away from the annual contest and towards multi-year certainty. The reset is, at bottom, a reset of the unit of account, from the assessment year to the block of years.
The safe harbour, in the OECD's own formulation, is a provision applying to a defined category of taxpayers or transactions that relieves eligible taxpayers of certain obligations otherwise imposed by a jurisdiction's transfer pricing rules, substituting simplified compliance for the ordinary arm's length inquiry.7 India first embraced the device in 2013 and refined it in 2017, prescribing sector-specific margins across roughly a dozen enumerated categories, among them software development services, information technology enabled services, knowledge process outsourcing, contract research and development relating to software and to generic pharmaceuticals, intra-group loans, corporate guarantees, and the manufacture of auto components.6 The margins were neither uniform nor modest. Software development and information technology enabled services attracted operating margins in the region of 17 to 18 percent depending upon scale; knowledge process outsourcing was tiered as high as 24 percent by reference to the ratio of employee cost; and contract research and development commanded a margin of 24 percent. The very multiplicity of categories bred boundary disputes, for the line between software development, knowledge process outsourcing and contract research is porous, and the taxpayer whose activity straddled two heads faced uncertainty as to which margin, and which threshold, governed it.
The Budget 2026-27 cuts that Gordian knot. Observing that software development, information technology enabled services, knowledge process outsourcing and contract research relating to software are, in the Finance Minister's words, quite inter-connected, the Government has clubbed the four into a single category of Information Technology Services and prescribed a common safe harbour margin of 15.5 percent, computed on operating cost.3 The Income-tax Rules, 2026, notified on 20 March 2026, give effect to the consolidation, locating the safe harbour framework in Rules 86 to 102 under the rule-making power of Section 167, with the margin for the consolidated category prescribed by Rule 89.5 Two features of the design are striking. First, the unified 15.5 percent margin sits below the entire range of the superseded margins, so that the taxpayer trades a contested and sometimes higher margin for a lower but assured one. Second, the whole determination is to be made, in the Finance Minister's phrase, by an automated rule-driven process, without any need for a tax officer to examine and accept the application, and the election once made may be continued for five years at the taxpayer's choice.3 Approval is thus removed from the discretionary sphere altogether; expressio unius est exclusio alterius, the enumeration of the eligible class excludes the officer's residual judgment upon it.
The Rules travel beyond the technology sector. They introduce, consonant with India's ambition to host global data and electronics infrastructure, a safe harbour for the provision of data centre services to an associated enterprise at a margin of 15 percent on cost, and a safe harbour, cast as an attribution of income to the non-resident, for the bonded warehousing of electronic components at 2 percent of invoice value.5 What unites these new heads with the technology consolidation is a single instinct: to convert whole classes of low-risk, high-volume, routine activity into administered rather than adjudicated pricing.
It would be complacent to greet the unified margin with unqualified applause, and the reflective practitioner will not. The OECD, having in 1995 regarded safe harbours with frank suspicion, revised Section E of Chapter IV of its Guidelines in 2013 to acknowledge that properly designed safe harbours can relieve compliance burdens and confer certainty, particularly for smaller taxpayers and less complex transactions, and that bilateral or multilateral safe harbours are to be preferred precisely because a unilateral one binds only the enacting State.8 The same guidance is candid about the costs. A prescribed margin, being an approximation, will overstate the arm's length result for some taxpayers and understate it for others; it may occasion double taxation where the counterpart jurisdiction declines to accept the figure; and it raises questions of equity and uniformity, since taxpayers in materially similar circumstances may be priced differently according to whether they elect in or out.7 A margin of 15.5 percent that fits the median captive comfortably may fit the genuinely high-value research unit poorly, and the latter, if it elects for the safe harbour, purchases peace at the cost of over-taxation, while if it declines, it forgoes the certainty the regime was designed to confer. The unified margin, in short, is an instrument of administrative convenience whose justice varies inversely with the heterogeneity of those it sweeps in.
If the unified margin is the reset's instrument, the threshold is its ambition. For the better part of a decade the eligibility ceiling for the service safe harbours stood at ₹200 crore of international transaction value, a figure that confined the regime to the smaller taxpayer and left the substantial captive to fend for itself in audit or to seek an Advance Pricing Agreement. By Notification No. 21/2025 the Central Board of Direct Taxes raised that ceiling to ₹300 crore for assessment years 2025-26 and 2026-27, and in the same stroke brought lithium-ion batteries for electric and hybrid vehicles within the definition of core auto components.6 That was an incremental loosening. The Budget 2026-27 is not incremental. It raises the ceiling for the unified Information Technology Services category from ₹300 crore to ₹2,000 crore, a nearly sevenfold enlargement at a single stroke.3
The policy logic is transparent and, on its own terms, defensible. India is host to a vast population of captive delivery centres and global capability centres whose transactions with their overseas parents run well into the hundreds and thousands of crores. To leave such entities outside the safe harbour was to consign the most significant slice of the sector, by value, to the very disputes the regime existed to prevent. By setting the bar at ₹2,000 crore the Government draws the great majority of the technology captive into the administered fold, and it does so at a moment when it wishes to present India as a predictable jurisdiction for the global capability centre. The threshold is, in effect, an industrial policy expressed in the grammar of transfer pricing.
Yet the very magnitude of the enlargement sharpens the theoretical objection. The arm's length principle is, in its essence, individuating: it asks what these parties, in this transaction, would have agreed. A threshold performs the opposite operation. It declares that above a certain scale, and within a certain class, the individual inquiry shall be displaced by a common figure. There is nothing objectionable in that for the small and routine taxpayer, where the cost of individuation exceeds its yield. But as the threshold climbs to ₹2,000 crore it begins to embrace taxpayers whose transactions are large enough, and whose functional profiles are various enough, that the loss of individuation is no longer trivial. The United Nations Practical Manual on Transfer Pricing for Developing Countries records the same tension, acknowledging the administrative attractions of simplification measures while cautioning that they must be calibrated so as not to depart materially from arm's length outcomes for the taxpayers they cover.9 The larger the class and the higher the threshold, the greater the strain upon that calibration.
There is, further, a quiet asymmetry in the bargain. The safe harbour is elective, and a rational taxpayer will elect into it only where the prescribed margin is at or below what an arm's length study would yield; the taxpayer whose true margin is lower will decline. The consequence, over the aggregate, is a mild adverse selection in the Revenue's favour: those who accept 15.5 percent are disproportionately those for whom 15.5 percent is a good bargain for the exchequer, while those for whom it is a poor one negotiate an Advance Pricing Agreement or take their chances in audit. The threshold does not cure that asymmetry; by enlarging the eligible population it enlarges the field over which it operates. This is not a criticism of the reset so much as an observation about its economics, and it leads naturally to the question with which this article is chiefly concerned: what, after all this, is left for the Advance Pricing Agreement to do.
The Advance Pricing Agreement has been the quiet success of Indian transfer pricing administration. Introduced in 2012, it has grown from a novelty into a mature programme of genuine international standing. The Central Board of Direct Taxes concluded 95 agreements in the financial year 2022-23, 125 in 2023-24, and a then-record 174 in 2024-25, the last of which included 65 bilateral agreements and the country's first multilateral agreement, carrying the cumulative tally to 815.10 The momentum did not abate. In the financial year 2025-26 the Board signed a further record of 219 agreements, of which 84 were bilateral, concluded with thirteen treaty partners and including India's first bilateral agreements with France, Ireland, Indonesia and Sweden; with that year the programme crossed the symbolic threshold of one thousand, standing at 1,034 agreements comprising 750 unilateral and 284 bilateral instruments.11 These are not the figures of a mechanism in retreat. They are the figures of a mechanism whose demand has, until now, exceeded the certainty available by any other route.
Financial year
Total APAs
Bilateral APAs
Cumulative
2022-23
95
—
2023-24
125
2024-25
174
65
815
2025-26
219
84
1,034
Table 1. Advance Pricing Agreements concluded by the Central Board of Direct Taxes. Source: APA Programme Annual Report 2024-25 and the Ministry of Finance press release of 31 March 2026.
Into that demand the reset now intrudes. The safe harbour, enlarged to a ₹2,000 crore ceiling and simplified to a single 15.5 percent margin approved automatically, offers a great swathe of the technology sector precisely what it once sought the Advance Pricing Agreement to obtain, namely multi-year certainty without the labour of negotiation. For the mid-sized captive whose margin comfortably clears 15.5 percent and whose transactions fall within the ceiling, the calculus is now simple. The safe harbour is costless, near-instantaneous and mandatory in its acceptance; the Advance Pricing Agreement, by contrast, entails an application fee, a functional and economic analysis, and a negotiation measured in years. For such a taxpayer the safe harbour is the obvious course, and the reset will, to that extent, divert traffic away from the unilateral Advance Pricing Agreement. This is not an unintended consequence; it is the design. The State would rather administer the routine captive by rule and reserve its scarce negotiating capacity for the cases that truly require it.
But the calculus does not resolve so tidily for every taxpayer, and it is in the residue that the new subtlety lies. Three considerations complicate the choice. The first is that the safe harbour is a creature of domestic law and binds only the Indian Revenue. It does nothing to constrain the tax authority of the counterpart jurisdiction, and it carries no relief from double taxation. The taxpayer who adopts a 15.5 percent margin in India has no assurance that the United States, the United Kingdom or Germany will regard that margin as arm's length in the hands of the paying associated enterprise; where they do not, the enterprise is exposed to economic double taxation with no treaty mechanism to relieve it. The Rules make the point unforgiving, for they bar the taxpayer who has accepted the safe harbour from invoking the Mutual Agreement Procedure in respect of the covered transaction, and they withhold the safe harbour altogether where the associated enterprise sits in a notified jurisdictional area under Section 176 or in a no-tax or low-tax jurisdiction.5 The bilateral Advance Pricing Agreement suffers no such infirmity: negotiated between competent authorities under the treaty, it binds both fiscs and extinguishes the double taxation at its root. For the exporter whose principal counterpart sits in a jurisdiction with an active transfer pricing administration, the bilateral Advance Pricing Agreement therefore retains a value the safe harbour cannot replicate, however generous the threshold.
The second consideration is one of coverage and bespoke fit. The safe harbour speaks only to enumerated categories within prescribed thresholds and at a prescribed margin; it is, by nature, a standard-form instrument. The Advance Pricing Agreement is bespoke. It can accommodate the transaction that eludes the enumerated categories, the margin that departs from the prescribed figure for good functional reason, the complex or bundled arrangement, and the intangible whose valuation no formula can capture. It offers, moreover, a rollback of the agreed methodology to as many as four preceding years, reaching backwards into open assessments in a way the prospective safe harbour cannot.1 For the taxpayer with a difficult back-history, or a functional profile that the unified margin fits poorly, the Advance Pricing Agreement remains the instrument of choice, and the reset does nothing to diminish it.
The third consideration is the Government's own answer to the first two. Alert to the risk that an enlarged safe harbour might hollow out the very Advance Pricing Agreement it wishes to keep vital for the harder cases, the Budget 2026-27 fortifies the Agreement in three ways. It promises a fast-tracked unilateral Advance Pricing Agreement for Information Technology Services, to be concluded as an endeavour within two years, extendable by a further six months at the taxpayer's request; and it extends to the associated enterprises of the applicant the facility of the modified return, so that the consequences of the Agreement may be given effect in their hands as well.3 The first of these directly attacks the Advance Pricing Agreement's historic weakness, its slowness, and narrows the temporal advantage the safe harbour would otherwise enjoy. The second, read with Section 169, tidies the downstream mechanics of implementation.1 The Government's strategy, in sum, is not to make the taxpayer choose between a fast safe harbour and a slow Agreement, but to offer a fast safe harbour for the routine case and a faster Agreement for the case that needs one.
The upshot is a decision matrix of some elegance. The mid-sized captive within the ceiling, transacting with counterparts in jurisdictions unlikely to contest 15.5 percent, will take the unified safe harbour and be done. The large captive above the ceiling, or the taxpayer whose activity resists the enumerated categories, will still require an Agreement, now with the comfort of a fast-tracked unilateral route. And the exporter whose counterpart jurisdiction is likely to disagree, and for whom double taxation is the live risk, will reach for the bilateral Advance Pricing Agreement, whose value the safe harbour cannot touch. The old question, safe harbour or Advance Pricing Agreement, is thus not abolished but refined: it becomes a question of which safe harbour, which Agreement, and above all whether the transaction's exposure is domestic or cross-border in its true incidence. That the jurisprudence long ago recognised the arm's length remuneration of a captive as capable of extinguishing further attribution, as the Supreme Court held in the context of a captive back-office in Director of Income-tax v. Morgan Stanley and Co. Inc.,12 only underscores the point: where the routine captive is fairly rewarded, certainty is the paramount good, and the reset supplies it in bulk.
A reset of this magnitude cannot be without friction, and candour requires that the friction be named. Four questions will occupy practitioners in the seasons ahead.
The first concerns the adequacy of 15.5 percent. A single margin, however carefully chosen, cannot be arm's length for all who adopt it. For the high-end research or engineering captive whose independent comparables would yield a materially higher return, the unified margin risks understating Indian profit and inviting, in due course, the attention of the counterpart jurisdiction under Pillar Two and the global minimum tax, where an artificially low Indian margin may simply shift the top-up elsewhere. For the thin-margin support unit, conversely, 15.5 percent may over-tax. The Government has chosen breadth over precision; the wisdom of that choice will be tested taxpayer by taxpayer.
The second concerns double taxation and the forfeiture of the Mutual Agreement Procedure. The bar on invoking that Procedure once the safe harbour is accepted5 is intelligible as a matter of administrative logic, for a State cannot negotiate with its treaty partner over a figure it has agreed to accept without inquiry. But it places upon the taxpayer the entire burden of foreseeing the counterpart jurisdiction's reaction before electing, and it converts what looks like a costless election into a considered waiver of treaty protection. The unwary taxpayer who elects for the safe harbour without regard to its cross-border incidence may find the relief it purchased in India purchased at the price of relief it has surrendered abroad.
The third concerns the automated, rule-driven approval. That the safe harbour for Information Technology Services shall be granted without an officer's examination3 is a considerable advance in ease of compliance and a genuine curb upon discretion. It also raises, in embryo, questions that the Rules and the Board's guidance will have to resolve in operation: how eligibility is verified where it is asserted rather than examined, what recourse lies where an application is wrongly rejected by the system, and how the automated grant interacts with the ordinary powers of assessment and reassessment. Lex non cogit ad impossibilia, the law does not compel the impossible; but automation must not become a black box that compels the taxpayer to accept an outcome it cannot interrogate.
The fourth is a question of coherence. The safe harbour is a special regime; the general arm's length machinery of Sections 161 to 166 endures beside it. Where the two meet, the maxim generalia specialibus non derogant supplies the ordering principle, the special safe harbour governing the transaction it covers and the general provisions the residue. But the boundaries will require tending, particularly as the block determination, the safe harbour and the Advance Pricing Agreement, three overlapping instruments of multi-year certainty, come to be operated in tandem. The practitioner's task, for some years, will be to know which instrument answers which case, and to counsel the client accordingly.
The Income-tax Act, 2025, read with the Budget 2026-27 and the Income-tax Rules, 2026, does not so much amend the Indian transfer pricing regime as re-orient it. The arm's length principle remains the lodestar, but the manner of its administration has turned decisively towards the formulaic and the multi-year. The unified safe harbour for Information Technology Services, the sevenfold enlargement of the eligibility threshold to ₹2,000 crore, and the fortification of the Advance Pricing Agreement are best understood not as three separate measures but as a single strategy: to administer the routine captive by rule, to reserve negotiation for the cases that need it, and to draw as much of the sector as possible into the domain of advance certainty.
For the profession, the lesson is that certainty is now abundant but no longer undifferentiated. The safe harbour, the unilateral Advance Pricing Agreement and the bilateral Advance Pricing Agreement are distinct goods answering distinct risks, and the reset has made the choice among them sharper rather than simpler. The taxpayer's true question is no longer whether certainty is available, for the State now presses it upon him, but which certainty his particular exposure requires, and at what price in individuation, in treaty protection, and in profit correctly attributed. To answer that question well is the counsel's contribution, and no formula, however elegant, will discharge it for him. Fiat justitia, let justice be done, remains the object; the reset has merely changed the instruments through which it is pursued.