On March 10, 2023, Silicon Valley Bank — the sixteenth largest bank in the United States — collapsed in 48 hours. The proximate cause was a bank run, but the underlying cause was something far more technical: SVB had classified $91 billion of bonds as Held to Maturity. As interest rates rose through 2022 and 2023, those bonds lost market value — their unrealised losses eventually reached $17 billion. But because the bonds were in HTM, none of those losses appeared in SVB's financial statements. The bank reported healthy profits even as its economic net worth evaporated. When depositors finally noticed the gap between reported book value and actual market value, they withdrew $42 billion in a single day. Indian banking regulators watched this unfold with particular attention, because Indian banks held their own massive HTM portfolios — dominated by government securities acquired during the pandemic liquidity surge — and faced the same interest rate risk. The RBI's response came in stages: the 2023 Investment Portfolio Directions (since withdrawn) overhauled the classification framework, and the November 2025 entity-specific Investment Portfolio Directions (RBI_MD_13148) completed the restructuring across every regulated entity type.
See also: What SLR and CRR Actually Do | What Basel III Actually Changed | How India Built Its Bond Market
The three categories and why they exist
Every Indian bank must classify its entire investment portfolio into three categories. The November 2025 Directions (RBI_MD_13148) state the rule plainly:
"A bank shall classify its entire investment portfolio (except investments in its own subsidiaries, joint ventures, and associates) under three categories, viz., Held to Maturity (HTM), Available for Sale (AFS), and Fair Value through Profit and Loss (FVTPL)."
Each category carries different accounting treatment, different profit-and-loss impact, and different regulatory capital consequences. A bank holding Rs 10,000 crore of identical government securities can report vastly different profits depending on how it distributes those securities across the three buckets. This is not a theoretical distinction — it is the mechanism through which interest rate movements translate (or do not translate) into reported bank earnings.
Held to Maturity (HTM): Securities acquired with the intention and objective of holding to maturity. These are carried at cost and not marked to market after initial recognition. Any discount or premium is amortised over the remaining life of the instrument. If interest rates rise and the market value of a 10-year government bond drops from Rs 100 to Rs 85, an HTM-classified bond still appears at Rs 100 (less amortised premium, if any) on the bank's balance sheet. The unrealised loss is invisible.
Available for Sale (AFS): Securities acquired with an objective that is achieved by both collecting contractual cash flows and selling. These must be fair valued at least quarterly. But the valuation changes do not hit the profit and loss account — they go to a separate reserve called AFS-Reserve. The bank's reported profit is shielded, but the unrealised gain or loss is visible in the balance sheet through the reserve. Upon sale, the accumulated gain or loss transfers from the reserve to the P&L.
Fair Value through Profit and Loss (FVTPL): This is the most transparent category. Securities here must be fair valued, and the net gain or loss goes directly to the Profit and Loss Account. Within FVTPL, securities classified as Held for Trading (HFT) must be valued daily; others at least quarterly. Every price movement is immediately reflected in reported earnings.
Why three categories? Because banks hold securities for fundamentally different purposes. A bank that buys a 10-year government bond to meet its SLR requirement and intends to hold it until maturity should not be forced to report quarterly mark-to-market fluctuations — the fluctuations are economically irrelevant if the bank will hold to maturity and receive the full face value. But a bank that buys the same bond for trading should report every price movement immediately, because the bank's profit depends on selling at the right time.
The SPPI criterion: what can go into HTM?
The 2025 directions introduce a critical test for HTM eligibility: the SPPI criterion — Solely Payments of Principal and Interest. A security can be classified as HTM only if two conditions are met: the bank intends to hold it to maturity, and the security's contractual terms give rise to cash flows that are solely payments of principal and interest on specified dates.
This sounds simple but has profound consequences. Government bonds typically meet SPPI because they pay fixed or floating coupons at defined intervals and return the principal at maturity. But instruments with convertible features, contractual loss absorbency features (like Additional Tier 1 capital instruments), or equity-linked coupons do not meet SPPI and cannot be classified as HTM. Equity shares, by definition, cannot be HTM.
Why does this matter? Because before the SPPI framework, banks could — and did — park a wide range of instruments in HTM to avoid mark-to-market volatility. The SPPI test forces a principled distinction: only instruments that behave like traditional debt (predictable cash flows of principal and interest) get the shelter of HTM accounting.
The RBI's Discussion Paper on Investment Portfolio norms (RBI releases Discussion Paper on Prudential Norms), released in January 2022, explained the rationale for this shift:
"The extant regulatory instructions on classification and valuation of investment portfolio by commercial banks are largely based on a framework introduced in October 2000 drawing upon the then prevailing global standards and best practices."
Two decades had passed. Global standards had moved to IFRS 9. India needed to catch up.
Why HTM is a regulatory gift — and a potential trap
HTM classification is the most sought-after accounting treatment in Indian banking. It shields the bank from reporting unrealised losses when interest rates rise and eliminates the earnings volatility that comes from quarterly mark-to-market of large bond portfolios. For a bank holding Rs 50,000 crore of government securities in HTM, a 100-basis-point rise in yields might create Rs 3,000-4,000 crore of unrealised losses — losses that never appear in the income statement.
During COVID-19, the RBI relaxed the HTM limit significantly. Banks that had been operating under a 19.5 percent ceiling (as a percentage of NDTL — net demand and time liabilities) were permitted to hold up to 23 percent in HTM. The March 2020 circular on spreading of MTM losses RBI/2019-20/175 allowed banks to spread mark-to-market losses on their AFS and HFT portfolios over multiple quarters — a direct response to the bond market turbulence caused by pandemic uncertainty.
Why was the HTM cap raised? Because the government was borrowing massive amounts to fund pandemic relief, and banks were the primary buyers. Without the HTM relaxation, banks buying large quantities of government bonds would have faced immediate mark-to-market losses as yields moved — creating a perverse situation where the act of funding the government's pandemic response would reduce reported bank profits.
But the trap is real. Every rupee of government bonds that sits in HTM at cost rather than market value represents an unrealised loss (or gain) that is invisible to depositors, shareholders, and analysts. The Silicon Valley Bank failure demonstrated the extreme case: when the gap between book value and market value becomes large enough, and depositors become aware of it, the resulting loss of confidence can destroy the bank faster than any loan default.
The April 2018 IFR circular RBI/2017-18/147 and the 2013 HTM clarification (Investments under HTM Category - Clarification) show the RBI repeatedly intervening to manage the tension between HTM's accounting shelter and its risk-hiding potential. The Investment Fluctuation Reserve — which requires banks to maintain at least 2 percent of their AFS and FVTPL portfolio — is one buffer. But the fundamental tension remains.
How AFS changed under the 2025 framework
The most significant change in the 2025 directions is the treatment of AFS valuation gains and losses. Under the new framework, AFS securities must be fair valued at least quarterly, and the net appreciation or depreciation goes directly to the AFS-Reserve — a component of the balance sheet that does not pass through the P&L account.
"The net appreciation or depreciation (adjusted for the effect of applicable taxes, if any) shall be directly credited or debited to a reserve named AFS-Reserve without routing through the Profit & Loss Account."
This is aligned with IFRS 9, where AFS-equivalent instruments (classified under FVOCI — Fair Value through Other Comprehensive Income) behave the same way. The significance for Indian banks is that AFS-Reserve is now reckoned as part of Common Equity Tier 1 capital, but with a restriction: unrealised gains in AFS-Reserve cannot be used for dividend distribution or coupon payments on Additional Tier 1 instruments.
Why does this matter? Because under the old framework, AFS valuation changes were aggregated and netted, and the net depreciation (if any) was charged to the P&L. This created an asymmetry: losses hit the P&L, but gains stayed in reserves. The new framework eliminates this asymmetry by routing both gains and losses through AFS-Reserve, providing a more complete picture of the bank's economic position without the earnings volatility of FVTPL.
For equity instruments, the treatment is even more unusual. A bank can make an irrevocable election at the time of purchase to classify an equity instrument under AFS. If it does so, any gain or loss on sale of that equity instrument is never transferred to the P&L — it moves from AFS-Reserve directly to Capital Reserve. This means a bank can hold equity investments without ever reporting trading gains or losses in its income statement.
The Expected Credit Loss framework: the next transformation
The current provisioning framework for Indian banks is based on an "incurred loss" model — provisions are recognised only after a loss event has occurred. A loan is provisioned only after it is classified as NPA. A bond is provisioned only after it shows signs of impairment. The problem with this model is that it recognises losses too late. By the time a loan is classified as NPA, the economic loss has already happened — often months or years earlier.
The RBI's push toward an Expected Credit Loss framework represents the next fundamental shift. The Working Group on ECL-based provisioning (Reserve Bank of India constitutes an external Work) was constituted in October 2023, following a discussion paper released in January 2023:
"The ECL approach to provisioning is a paradigm shift from the current incurred loss-based provisioning regime. The Discussion Paper envisages a forward looking approach."
Under ECL, a bank must estimate expected losses on day one of acquiring a loan or a bond, based on forward-looking probability models. If conditions deteriorate, the expected loss estimate increases — and provisions increase immediately, even before the borrower actually defaults. This forward-looking approach aligns with IFRS 9 and is designed to prevent the situation where banks accumulate unreported losses until a crisis forces them to recognise everything at once.
For the investment portfolio specifically, ECL would mean that a bank holding corporate bonds would need to estimate and provision for the probability of default over the bond's expected life — not just when the bond actually defaults. The interaction between ECL provisioning and the HTM/AFS/FVTPL classification framework is complex: HTM bonds subject to ECL would carry both their amortised cost and an ECL provision overlay, while AFS bonds would carry fair value plus ECL provisions on credit-impaired instruments.
The SVB lesson and what India learned
Silicon Valley Bank's failure was not caused by bad loans. It was caused by the interaction between interest rate risk, accounting classification, and depositor confidence. SVB held $91 billion of bonds in HTM — mostly long-duration agency mortgage-backed securities bought when rates were near zero. When rates rose, the market value of those bonds dropped by $17 billion. But because they were in HTM, the losses were unrealised and unreported. SVB's regulatory capital ratios looked fine. Its reported earnings were acceptable. The gap between appearance and reality was hidden by the HTM classification.
Indian banks faced a structurally similar situation — one that the RBI had already begun to address by replacing the old classification framework that had been amended piecemeal since 2000. During 2020-2021, banks purchased large volumes of government securities at historically low yields, parking much of the acquisition in HTM. When the RBI began tightening in May 2022, yields rose and the market value of those bonds fell. The difference was that Indian regulators had already begun the process of reforming the classification framework.
The September 2023 Investment Portfolio Directions implemented the SPPI criterion, the AFS-Reserve mechanism, and the requirement to maintain an Investment Fluctuation Reserve. The November 2025 entity-specific directions extended the framework across every entity type — commercial banks (RBI_MD_13148), NBFCs (RBI_MD_12950), All India Financial Institutions (RBI_MD_12975), and co-operative banks at all levels.
The regulatory chain
The investment classification regulatory chain runs through a clear sequence: the October 2000 framework established the original three-category system, the December 2015 HTM relaxation RBI/2015-16/261 adjusted limits in the context of SLR reductions, the April 2018 IFR and MTM spreading circular RBI/2017-18/147 introduced the Investment Fluctuation Reserve, the June 2019 HTM sale norms RBI/2018-19/204 addressed premature disposals from HTM, the January 2022 Discussion Paper (RBI releases Discussion Paper on Prudential Norms) proposed the overhaul, the September 2023 directions (Reserve Bank of India (Classification, Valuation a) implemented it, and the November 2025 Master Directions (RBI_MD_13148) consolidated the regime on an entity-specific basis.
At every link, the driving force was the same question: how transparent should the gap between book value and market value be? The old framework tilted toward concealment — HTM hid losses, MTM spreading delayed their recognition, and the absence of a CET1-linked AFS reserve meant that valuation changes affected capital only indirectly. The new framework tilts toward transparency — SPPI limits what can enter HTM, AFS-Reserve is recognised in CET1, FVTPL captures everything else immediately, and the IFR provides an explicit capital buffer against market volatility.
The investment classification framework may sound technical, but its practical consequences are enormous. It determines how much profit your bank reports, how much capital it claims to hold, and how vulnerable it is to a sudden loss of confidence. Silicon Valley Bank proved that accounting classification is not merely an accounting question — it is a solvency question. India's 2025 framework is built on that lesson.