The gap is the business

An insurance company is, mechanically, a bank that nobody calls a bank. It takes money in today against a promise to pay money out on a date it does not know, in an amount it does not know, to a person who may never claim at all. The premium arrives first. The claim arrives later, or never. The pool of money sitting in that gap is called float, and for the industry as a whole it is enormous, permanent and, unusually for borrowed money, often free.

Warren Buffett did not invent float and has never claimed to. What Berkshire Hathaway did was state plainly what the gap is for. The 2025 annual report defines it in one line: insurance float is “the capital we hold to pay future losses and, in the meantime, invest for Berkshire’s benefit”. At the end of 2025 that stood at $176 billion, up from $171 billion a year earlier and $88 billion in 2015. The 10-K adds the part that makes it remarkable rather than merely large: across the three years to December 2025, Berkshire’s insurance operations made an underwriting profit every year, so “the average cost of float was negative in each year”.

That is the whole trick, and it is worth being precise about why it is a trick and not magic. Float is not profit. Every rupee of it is someone else’s money, owed against a claim that has not yet been made. An insurer that mistakes float for capital and invests it as though it will never be called is not clever; it is the reason solvency regulation exists. The skill is not in having float. It is in having float that costs nothing — which requires underwriting discipline — and then having somewhere useful to put it.

Three balance sheets, one chain

The money does not sit in one place. It moves down a chain, and each link holds a different slice of the same risk.

The bank leg is the one people forget. In India banks are among the largest distributors of insurance, which means a single group can originate the policy, hold the float and lend against the same customer.

The reinsurer is where the chain stops being intuitive. A primary insurer that writes a motor book in Tamil Nadu is exposed to Tamil Nadu. A reinsurer that accepts slices of a hundred such books across forty countries is exposed to something else entirely: the correlation between them. Most of the time that diversification is real and the reinsurer is the safest link. In the rare event where losses arrive everywhere at once — a pandemic, a synchronised market fall — the diversification stops working precisely when it is needed, and the reinsurer is the most exposed link instead.

India’s version of this layer is unusually concentrated. Every general insurer must cede a fixed share of each policy to the state-owned national reinsurer, GIC Re. IRDAI has set that obligatory cession at 4 per cent for 2026-27, down from 5 per cent a few years earlier, and has retained GIC Re as its sole recipient. GIC Re holds roughly two-thirds of the Indian reinsurance market. So a compulsory, non-negotiable sliver of nearly every Indian insurance policy accumulates on one balance sheet — which is a deliberate policy choice about keeping reinsurance capital onshore, and also a concentration.

2020: the year the float ran backwards

The popular account of March 2020 is that insurers had to sell shares to pay claims. It is a tidy story and it is mostly wrong, in a way worth correcting because the real mechanism is more troubling.

Work by Federico Apicella, Raffaele Gallo and Giovanni Guazzarotti at the Bank of Italy examined Italian insurers’ securities portfolios from 2017 to June 2020 and found the deciding variable was not claims experience. It was the solvency ratio. Before the pandemic, insurers with high and low solvency behaved alike, both increasing exposure by roughly 13 per cent when an asset’s price fell 1 per cent — textbook stabilising behaviour, buying what others were dumping. After the outbreak the two groups split. Better-capitalised insurers went on buying, raising exposure about 8 per cent. Less-capitalised ones went the other way, cutting exposure about 4 per cent. The same shock, the same asset, opposite behaviour, sorted by how much capital headroom the firm had.

The reason is that under a risk-based capital regime such as Solvency II, a falling asset price does two things at once. It reduces your assets, and it raises the capital you are required to hold against what remains. An insurer close to its solvency floor can be forced to sell the falling asset simply to stop the capital requirement climbing — which pushes the price down further, which tightens the requirement for the next insurer. The study found the same firms cutting exposure to BBB-rated corporate bonds, the rung just above junk, where a single downgrade would have triggered a much larger capital charge.

This is procyclicality, and it is a design flaw rather than a failure of nerve. The rule that makes each insurer individually safer makes the system collectively less stable, because it turns every large holder into a forced seller at the same moment. The claims were real too — event cancellation, business interruption litigation, mortality — but the selling was driven by the capital rulebook, not the claims department.

What LIC did instead

India ran the opposite experiment, for reasons that were partly structural and partly not.

As markets bottomed in March 2020, the value of the Life Insurance Corporation’s equity holdings fell to around $55 billion, its lowest in six years. LIC did not sell. It bought — adding to positions through the March quarter and, by its own account, putting roughly ₹55,000 crore into equities from April onwards. By the first half of 2020-21 the portfolio was worth about $77 billion, a 40 per cent recovery driven by both price appreciation and fresh buying.

Two things made that possible, and only one of them is admirable. LIC’s liabilities are long-dated life and annuity contracts, not short-tail general insurance, so a bad quarter in equities does not translate into an immediate need for cash. And LIC is majority state-owned, which means it can behave counter-cyclically in a way a listed insurer answering to quarterly solvency disclosure cannot. The stabilising is real. It is also, in part, a function of who owns it — and the same feature that lets LIC buy a falling market is the one that has periodically made it the buyer of last resort for government divestments.

The Indian float, in the actual numbers

Here is what India’s general insurers wrote in 2024-25, from the segment-wise return compiled by the General Insurance Council.

Gross direct premium underwritten within India, segment-wise, provisional and unaudited, for the period up to March 2025. One caveat travels with the total: IRDAI changed the reporting format from 1 October 2024 to exclude premium from long-term policies. Adding the ₹6,989 crore of excluded long-term premium back gives ₹3,14,648 crore for 2024-25 against ₹2,89,673 crore, or 8.6 per cent growth rather than the 6.2 per cent the raw comparison shows. The headline understates the year because the ruler changed mid-measurement.

That footnote is not a technicality. A reported 6.2 per cent that is really 8.6 per cent is the difference between a sector decelerating and a sector growing at trend, and anyone comparing 2024-25 against earlier years without adjusting for the format change will draw the wrong conclusion.

Why the third-party half is the float half

Split motor in two and the float story separates cleanly from the insurance story.

Own-damage cover — ₹40,355 crore — is short-tail. Your car is dented, a surveyor inspects it, a garage is paid, the file closes. Premium in, claim out, often inside the same financial year. Float from own-damage exists but it is thin and it turns over fast.

Third-party cover — ₹58,711 crore, or 59 per cent of the motor book — is the opposite. It is compulsory under the Motor Vehicles Act, it is priced by the regulator rather than the market, and its claims are settled through Motor Accident Claims Tribunals. Those cases commonly run one to three years and can run considerably longer. The insurer collects the premium on day one and may not know the size of the liability for several years, and in the meantime it holds the money.

This is also where the analogy to Berkshire is strongest and least noticed. Berkshire’s float did not come primarily from car insurance premiums arriving; it came from long-tail liability and reinsurance business where claims develop over years or decades. The Indian motor third-party book has exactly that shape, at scale, and it is growing.

What the car boom actually adds

India sold a record 43 lakh passenger vehicles in 2024-25, and 46.43 lakh in 2025-26 — a 7.9 per cent rise, with every vehicle segment posting its best year in seven. The float consequence of that is not the premium from the sale. It is the premium from every year after.

A car sold in 2026 joins a compulsory-insurance pool and stays in it for as long as it is on the road, which in India means well over a decade. The flow that matters is not new sales, it is the parc — the total insured fleet — and new sales only ever add to it. Motor premium grew 8 per cent in a year when passenger vehicle sales grew 2 per cent, which is the arithmetic of a stock growing faster than the flow into it, helped along by regulated third-party rate revisions and rising sums insured.

One honest caveat belongs here, because it cuts against the tidiness of the story. India has a persistent uninsured-vehicle problem, concentrated in two-wheelers and older commercial vehicles, and estimates of the uninsured share vary widely enough that this piece will not quote one. The insured parc is smaller than the registered parc by a margin nobody measures well. That does not change the direction — it means the float is being built from a base that is less complete than the registration data implies.

The rulebook is the whole difference

So India has a growing, long-tail, compulsory float. Why does no Indian insurer look anything like Berkshire Hathaway?

Because float is only half the machine. The other half is what you are permitted to do with it, and there the two systems are not comparable at all.

Indian figures are the investment norms under the IRDAI Investment Regulations; Berkshire figures are cash and investments held in insurance businesses at 31 December 2025 per the 2025 annual report, which also shows $212.65 billion in cash, cash equivalents and US Treasury bills.

Read that last pair again. Berkshire’s insurance arm holds roughly seventeen times more equity than bonds. A regulated Indian insurer is required to run closer to the inverse. These are not different investment philosophies applied to the same freedom; they are different rulebooks producing different institutions.

The floors are not the same for both halves of the industry, and the difference matters for this article. The Insurance Act requires a life insurer to hold not less than 50 per cent in approved securities and a general insurer a minimum of 30 per cent — built from a 25 per cent and 20 per cent central-government-securities base respectively. Motor is general insurance. The float this piece is about therefore sits under the looser of the two regimes: a general insurer has roughly 70 per cent of its assets outside the approved-securities floor, against 50 per cent for a life insurer. That is still nothing like Berkshire’s latitude, but it is meaningfully more room than the life-side numbers imply.

The Indian constraint is not obviously wrong. A mandatory government-securities floor makes insurers structurally safe, gives the sovereign a captive domestic buyer for its debt, and would have prevented precisely the forced-selling spiral the Bank of Italy documented in Italy — you cannot be squeezed out of an equity position you were never allowed to build. The cost is that the return on the float accrues largely to the bond market, which is to say to the government, rather than compounding in the insurer. India has chosen stability and cheap sovereign funding over compounding. That is a defensible trade. It is worth being clear that it is a trade.

The scale is not small either. LIC alone reported assets under management of ₹54.52 lakh crore at 31 March 2025, rising to ₹57.23 lakh crore by 30 September. That is a pool larger than Berkshire’s entire float, overwhelmingly parked in government paper by regulation.

What actually counts as an approved security

“Approved securities” and “approved investments” sound like synonyms and are not. The difference is the single most useful thing an investor can know about how insurance money moves, because it decides which rupee can ever reach the equity market.

IRDAI publishes an exhaustive Category of Investments schedule, and every rupee an insurer holds is reported against one of these codes. The buckets run A to E.

Condensed from the Category of Investments annexure to the IRDAI master circular on actuarial, finance and investment functions. The schedule is longer than this — category D alone runs to more than forty codes and category C to nearly fifty — and each carries its own exposure and rating conditions. This table names the classes, not the limits.

Three consequences follow, and each is checkable in an insurer’s published returns.

Ratings are the gate, not judgement. To sit in category D an instrument generally has to clear a minimum credit rating; drop below it and the holding is reclassified into category E, which is capped. There are explicit reclassification codes for exactly this — a downgrade mechanically moves the asset, whatever the insurer thinks of the credit. This is the Indian analogue of the BBB effect the Bank of Italy found in 2020, and it is why a wave of downgrades is an asset-allocation event for insurers rather than merely a mark-to-market one.

Equity is bounded twice. First by what is left after the approved-securities floor, then by the exposure norms inside category D and by the approved-equity universe being confined to larger listed companies. An insurer cannot express a view on a small-cap the way a mutual fund can.

Promoter-group holdings are tracked separately at every level. Note how often “Promoter Group” appears as its own code — equity, debentures, mutual funds, ETFs, each has a distinct one. Overall promoter-group exposure is capped at 5 per cent of investment assets. The schedule is built so that self-dealing has nowhere to hide in the reporting.

The reinsurance layer, and who is arriving

The obligatory cession to GIC Re concentrates a slice of Indian risk on one balance sheet. The countervailing development is happening 900 kilometres away, in Gujarat.

An earlier piece on India’s two tracks of health cover found the supply side moving faster than the demand side was being reported: through 2026 the International Financial Services Centres Authority approved a wave of foreign reinsurance branches at GIFT City — International General Insurance, Singapore Re, Santam, Abu Dhabi National Insurance, Eurasia, Korean Re, Kuwait Re, Peak Re, Saudi Re, Allianz, Generali, Starr International, Qatar Re, Doha Re and Lloyd’s among them — taking the count of active insurance offices from roughly eight to around twenty-four in a single financial year. Several entrants named health reinsurance, surety, parametric cover, marine and cyber as the lines they intended to build.

That matters for the float argument in a specific way. Reinsurance capacity is where the long-tail, capital-hungry risk goes. If it sits offshore, the premium leaves the country and so does the float; if it sits in GIFT City, the risk is still ceded but the money stays within an Indian jurisdiction. Going from eight to twenty-four active offices in a year is the state trying to onshore the float, not merely the risk — the same instinct as the obligatory cession, pursued by the opposite method. One compels a share to a state reinsurer; the other competes for the rest by building a venue.

Whether it works is a question for the settlement data, not the licence count, and licences are what we currently have. An approved branch is an intention. The figure worth watching is ceded premium retained in India, which will take a few more annual returns to read.

The monopoly ends: private reinsurers arrive

The larger change is not the branches. It is that for the first time since nationalisation, GIC Re has domestic company.

In March 2025, in what was Debasish Panda’s final board meeting as IRDAI chairperson, the regulator granted the first reinsurance licence ever issued to a private Indian player: Valueattics Reinsurance, backed by Kamesh Goyal — founder and chairman of Go Digit — and Prem Watsa’s Fairfax group, starting with reported paid-up capital of about ₹210 crore. A year later, at its 134th board meeting on 10 March 2026, IRDAI cleared Allianz Jio Reinsurance, an equal joint venture between Allianz and Jio Financial Services, marking Allianz’s return to India after the unwinding of its long partnership with Bajaj Finserv. The same meeting licensed Kiwi General Insurance, a primary insurer backed by WestBridge Capital and former Tata AIG chief executive Neelesh Garg, writing motor, health and property.

Domestic entities only. Foreign reinsurers additionally participate through GIFT City branches and as cross-border reinsurers, of which roughly 280 provide capacity into India.

Note who is behind the first two, because it is not incidental to this article’s subject. Fairfax Financial has spent four decades running an explicitly Buffett-derived model: underwrite for a profit, hold the float, invest it with concentration and patience. Prem Watsa has been called Canada’s Buffett often enough that he has publicly tired of it. The first private reinsurance licence in India has gone to precisely the school of investor that treats float as the product rather than the by-product.

The Allianz–Jio venture points somewhere different and arguably more consequential. Jio Financial Services is the financial arm of a group that already owns the distribution — telecom subscribers, retail footprint, payments. Pair that with Allianz’s underwriting and reinsurance capability and you have the bank-insurer-reinsurer chain described earlier collapsed into a single corporate group, at national scale. That is efficient. It also concentrates origination, float and risk-transfer in one place, which is the structure prudential regulators generally watch most closely.

What none of this changes yet is the asset side. A private reinsurer in India invests under the same IRDAI investment norms as everyone else. Fairfax’s model needs latitude on the asset side to work, and that latitude is what Indian regulation is specifically designed not to grant. The interesting question over the next few annual returns is not whether Valueattics writes business. It is whether a float compounder can generate a Fairfax-like return from a portfolio that must be half government paper — and if it cannot, whether it argues for the rules to change.

What to watch

• Ceded premium retained in India. The licence count at GIFT City is an intention; retained premium is the outcome. Two or three more annual returns will show whether the onshoring worked.
• Motor third-party rate revisions. Third-party pricing is regulated, so the largest float-generating line in the country is repriced by administrative decision rather than by the market. Each revision moves the float.
• The long-term premium reporting change. Any 2024-25 comparison that ignores the 1 October 2024 format change understates growth by nearly two and a half percentage points.
• Whether the investment norms move. The arrival of investors who exist to compound float creates, for the first time, a domestic constituency arguing for a higher equity ceiling.
• Solvency behaviour in the next drawdown. The Bank of Italy result predicts that thinly-capitalised insurers sell into falling markets. India’s government-securities floor should blunt that. The next real market stress will test whether it does.

The honest summary

Float is not a strategy. It is a structural feature of any business that is paid before it performs, and it exists identically in Omaha, Milan and Mumbai. What differs is entirely downstream: what the rulebook lets the holder do with it, and what the rulebook forces the holder to do when asset prices fall.

In 2020 the European rulebook forced holders to sell into weakness, and did so most severely to the firms least able to withstand it. India’s rulebook made that particular accident close to impossible, at the price of routing most of the return on the country’s largest capital pool into government paper. India is now building more float than it ever has, from a record vehicle fleet and a compulsory long-tail liability line, at the same moment it is letting in the kind of investor who exists to compound exactly that. Those two facts sit slightly awkwardly together, and the next few years are where they get reconciled.

Sources and caveats