Intercompany financial transactions:
loans, guarantees and cash pooling in transfer pricing.

Debt or capital?

Before discussing the rate, Law 14.596/2023 (art. 27) requires delineating the operation: when a controlled transaction provides resources formalized as debt, one must determine whether it is in fact debt — or whether, in economic substance, it is a capital contribution. The contract label does not decide; substance decides.

The factors that pull recharacterization toward capital are well known in international practice (OECD TPG, ch. X): the absence of a defined maturity; the lack of the borrower's repayment capacity (cash flow cannot service the debt); the absence of guarantees or of any real consequence for default; subordination to all other creditors; and remuneration tied to results rather than fixed interest. The more of these signals, the more the "loan" looks like equity.

The sanction is in the sole paragraph of art. 27: once delineated as capital, the interest and other financial expenses are non-deductible for IRPJ and CSLL. It is not a price adjustment — it is the full loss of the deduction. That is why delineation is the first test, and the most expensive to get wrong.

The reason form does not control is economic: between related parties, the contract can call an instrument whatever is convenient, because there is no independent lender on the other side negotiating its own protection. An unrelated bank would never advance resources with no maturity, no repayment plan and no recourse on default — so when an intragroup instrument carries those features, the label "loan" is describing a relationship that the market would have priced as equity. Delineation simply asks what an independent party would have agreed to, and treats the operation accordingly. None of the factors is decisive on its own; they are read together, and the analysis weighs how the funds actually behave over time — whether interest is really paid on the agreed dates, whether the principal is ever expected to return, and whether the lender acts like a creditor or like a shareholder waiting for the business to perform.

The rate: the risk is the borrower's

Once delineated as debt, the operation follows arm's length (art. 28): the deductible interest is the market rate. And "market", here, is the rate the borrower would obtain on its own — not the rate the parent raises. Using the parent's funding cost (as a rule cheaper, because the parent has a better rating) is the most common error and the first target for disallowance.

Building the rate follows four steps (OECD TPG, ch. X): (1) the borrower's standalone credit rating, from its own fundamentals — leverage, interest coverage, cash generation; (2) the adjustment for implicit support: belonging to the group improves the rating by a few notches (notching), because the market assumes the group would rescue the subsidiary — this effect exists even without a formal guarantee and lowers the rate; (3) the selection of market comparables (bonds and loans of borrowers with an equivalent rating, in the same term and currency); (4) building the spread by term, currency, guarantees and covenants.

Illustrative example: a subsidiary with a standalone rating equivalent to "BB", raised to "BB+" by the group's implicit support, should borrow at the cost consistent with "BB+" — and not at the parent's "AAA" cost. The gap between the two rates is exactly the interest expense the tax authority tends to disallow.

The logic of implicit support deserves to be made explicit, because it cuts both ways. It is a passive benefit: it flows from the simple fact of belonging to the group, not from anything the parent actively does, and for that reason it is never charged for. But it is also real, and it has to be priced in. A lender looking at the subsidiary in isolation would see weaker fundamentals; a lender who knows the subsidiary sits inside a strong group prices in the expectation that the group would not let a strategically important member default. That expectation lowers the rate — which is why the standalone rating is the floor, not the answer, and why ignoring implicit support and lending at the parent's cost is wrong in the opposite direction from ignoring it and lending at a punitive standalone cost. The arm's length rate lives in between: the rate of a borrower with the subsidiary's own fundamentals, improved by the few notches that group membership credibly buys, and no further.

The no-substance lender

And when the related lender is just a "box" passing resources through? The law addresses this in art. 29, separating financial capacity from risk control:

• Item I — if the lender has no financial capacity and does not control the risks of the financing, its remuneration is the risk-free interest rate. It is a mere conduit; it assumes no risk, so it earns no risk spread.
• Item II — if it has the capacity to provide the capital but neither assumes nor controls the risk, it receives the risk-adjusted rate for that financing, without the residual return. The surplus is reallocated to whoever actually controls the risk.
• Item III — if it performs only an intermediation function (the resources in fact come from another party), it is remunerated according to its functions, assets and risks (FAR), under art. 2 — and not by the difference between funding and passing through.

The law itself calibrates the floors: the risk-free rate is the return of a lowest-risk investment (as a rule, government bonds in the lender's functional currency with the lowest rates); the risk-adjusted rate is that risk-free rate plus a premium that reflects the risk actually assumed (art. 29, sole paragraph).

The reasoning behind the intermediation case (item III) is worth drawing out, because it explains why a "lender" can end up earning the least. When an entity only books the loan in and out — the money in fact originates elsewhere, and the decisions about whether to lend, on what terms and against what risk are taken by others — it is performing a service, not bearing the financing. Its reward must match what it actually contributes: the functions it performs, the assets it uses and the risks it controls (FAR), under art. 2. That return is a service margin on the activity of arranging and administering the flow, not the interest spread between the cost of funds and the rate charged downstream. Capturing that spread would pay the conduit for risk it never took and capital it never truly committed, and the surplus is reallocated to whoever does carry the risk.

It is the application, to the financial world, of the same principle as intangibles: the return follows the function and the risk control, not the mere title to the money. Structures with a group finance company in a low-tax jurisdiction, with no people and no substance, are exactly the target of this provision.

Two filters: price × amount

Every intercompany financial transaction passes through two independent controls that do not replace each other — and may apply to the same loan:

Filter 1 — price (transfer pricing, Law 14.596/2023): is the operation debt or capital? And, if debt, is the rate arm's length? It corrects the interest rate.

Filter 2 — amount (thin capitalization, Law 12.249/2010): it limits how much debt with foreign related parties generates deductible interest. Indebtedness is deductible up to twice the net equity of the related party (art. 24) — a limit that drops to 30% of net equity when the lender is in a tax haven or privileged tax regime (art. 25), assessed month by month. Excess debt has its interest disallowed, even if the rate is perfect. A point that confuses the market deserves attention: Law 14.596 did not repeal these rules — on the contrary, it gave them new wording (art. 42) so that they coexist, cumulatively, with the new transfer pricing.

In practice the two add up: a loan may have its rate adjusted by TP and part of its amount disallowed by thin capitalization. Modeling the operation means passing both filters at the same time.

Cash pooling: who earns what

The centralization of the group's cash balances has its own rule in art. 32. There are two forms: physical cash pooling (account balances actually migrate to a master account) and notional (balances are merely offset for accounting, without transfer). In both, three OECD principles (ch. X) guide remuneration:

• Short-term. The pool is a short-term liquidity instrument. Balances that stay credit or debit for long periods stop being pooling and become a long-term loan — with the arm's length rate that requires.
• The leader does not capture the spread. A cash pool leader that merely coordinates, without assuming risks, is remunerated as a service provider (a coordination margin) — it does not keep the difference between funding and lending.
• The synergy belongs to the participants. The gain of obtaining better rates through aggregated volume (synergy benefit) is shared among the members that generate it, in proportion to their contributions — not concentrated in the leader.

Guarantees and implicit support

When a related party provides a guarantee — a legally binding commitment to honor the debt upon default — there is a transaction to price (arts. 30 and 31). But the technical point is what is not charged.

The borrower already obtains better conditions merely by belonging to the group — the implicit support. This benefit is passive, arises from association, and is not a service: it is not charged (OECD ch. X). The formal (explicit) guarantee only adds value above that level. So the guarantee fee remunerates only the incremental benefit of the explicit guarantee — the difference between the cost of the debt with and without the formal guarantee, net of what implicit support already delivered.

And there is a statutory cap: the guarantee remuneration, when delineated as a service, cannot exceed 50% of the benefit that surpasses the implicit support (art. 31, sole paragraph), unless reliable evidence of a more appropriate approach exists. Furthermore, the additional amount of resources the borrower was only able to obtain because of the guarantee is, as a rule, delineated as a capital contribution — and no fee is due on it (art. 30, sole paragraph).

The usual methods to measure this increment (OECD ch. X): the yield approach (the interest-differential approach — how much the guarantee lowers the rate), the cost approach (the guarantor's expected cost of default) and pricing by comparables. Charging a guarantee fee on the effect that would already exist without a formal guarantee is the error the tax authority targets.

Intragroup insurance

Art. 33 reaches insurance transactions between related parties — typically captive insurers (captives), created by the group to insure its own risks. The test is the same: the captive must have substance (capacity to assume the risk, real diversification, regulatory capital) for the premiums paid to it to be deductible at arm's length value. A captive without substance, which merely accumulates premiums in a low-tax jurisdiction without bearing real risk, is recharacterized — the premiums stop being a deductible expense to the extent they do not correspond to a genuine transfer of risk.

A worked example: the parent loan

Take an illustrative case, as if it were a client of the firm. ImportCo BR takes a US$ 50 million loan from its foreign parent, at 9% per year. ImportCo's standalone rating, already raised by the group's implicit support, supports a market cost of around 6% for a borrower of that profile, term and currency.

Filter 1 (transfer pricing). The tax authority adjusts the rate to arm's length: the 3 points of excess (9% − 6%) on US$ 50 million become non-deductible interest — about US$ 1.5 million per year. The price of the debt has been corrected.

Filter 2 (thin capitalization). In parallel, ImportCo's net equity is R$ 80 million. The loan, once converted, is equivalent to R$ 275 million — well above the 2× net equity limit (R$ 160 million). The interest on the excess indebtedness is disallowed, even though the rate had already been corrected.

Both adjustments fall on the same loan — with the technical care of not double-counting the same band of expense (the order in which the tests are applied matters). On the disallowed portion: IRPJ (25%) + CSLL (9%), a 75% ex-officio penalty and Selic interest. The lesson is direct: an intragroup loan must be designed to pass both filters at the same time — rate and amount. (Illustrative figures.)

What should ImportCo have done differently? Two things, and both before signing. On the rate, it should have priced the loan from the start at the cost a borrower of its own profile, term and currency would actually pay — the roughly 6% the comparables support — instead of accepting the 9% set by reference to the parent, and it should have kept the rating study and the comparables on file to show how that 6% was reached. On the amount, it should have sized the debt against its own net equity before drawing it down: a R$ 275 million loan against R$ 80 million of equity was over the limit on day one, so part of the interest was lost no matter how clean the rate. Pricing the rate correctly does not save the excess amount, and keeping the amount within the limit does not save an off-market rate — the two have to be planned together, at the moment the operation is structured, not reconstructed once the assessment arrives.

What CARF shows

Administrative case law on intragroup interest is, for now, concentrated on thin capitalization — the arm's length rate of Law 14.596 is too new to have consolidated case law. But the thin cap decisions are a good risk map, since thin capitalization remains in force and cumulative.

Two recurring points. First, CARF confirms that interest paid to a related party in a privileged tax regime is subject to the stricter limits of art. 25 of Law 12.249 (30% of net equity), and not to the 2× of art. 24 — as in Acórdão 1301-006.874. Second, the excess indebtedness is assessed month by month, not at the end of the fiscal year, which changes the calculation of what is deductible — as in Acórdão 1301-007.551. There is also a signal that the tribunal prefers to apply the specific anti-avoidance rules (thin capitalization, transfer pricing) rather than disqualifying the operation as a generic "sham".

The lesson carries over to the new regime: with the arm's length rate added to thin capitalization, defending an intragroup loan requires documenting the debt/capital delineation, the borrower's rating and the fit within both limits. The tax authority has specific tools for each flank — and the taxpayer needs evidence for each one. (The cited decisions concern thin capitalization, the regime in force in parallel; each case requires verification against the full text.)

What the tax authority expects

Every controlled financial transaction enters the transfer pricing documentation (arts. 34 and 35, with penalties for omission or inaccuracy). For the financial block, the defensible file gathers:

• the debt/capital delineation analysis, with the factors that sustain the debt treatment;
• the borrower's standalone rating, the implicit-support adjustment and the notching memo;
• the rate comparables (bonds, loans) with the selection rationale and the adjustments;
• the contracts — term, currency, covenants, guarantees — consistent with the delineation;
• for pooling, the arrangement and the synergy-benefit allocation criterion;
• for guarantees, the guarantee fee calculation and the isolation of implicit support.

The thread running through all of these items is consistency: the contract, the delineation memo, the rating study and the rate comparables have to tell the same story. A loan documented as debt but written with no maturity, or priced off a rating the supporting study does not justify, hands the tax authority the contradiction it needs. The TaxUp team assembles this file as a single, coherent Local File on the financial block — built when the operation is structured, not improvised after a notice — so that each figure in the return can be traced back to the analysis that produced it. That is what turns a deduction from something asserted into something defended.

Without this backing, the adjustment falls on the IRPJ and CSLL base — and, at the limit, recharacterization to capital eliminates the entire deduction. For an overview of the regime and the other methods, see Transfer Pricing and the methods.