Float is other people's money that sits in your account before you have to pass it on — and the returns you earn on it in the meantime. The concept is simple. The consequences are extraordinary. Warren Buffett built the greatest compounding machine in financial history on float, and most people who study Berkshire Hathaway still underestimate how central it was.
The insurance industry invented the dynamic. When you pay your car insurance premium in January, the insurer doesn't need that money until a claim arrives — maybe in March, maybe never. In the gap between collecting premiums and paying claims, the insurer holds the cash. That cash is the float. It doesn't belong to the insurer. It will eventually be paid out. But while it sits there, the insurer can invest it — in bonds, stocks, real estate, entire companies. The investment returns belong to the insurer, not the policyholder.
Buffett recognised this in the 1960s and spent the next six decades engineering the largest float pool in corporate history. Berkshire's insurance subsidiaries — GEICO, General Re, Berkshire Hathaway Reinsurance — collected premiums from millions of policyholders. By 2023, Berkshire's float exceeded $167 billion. That is $167 billion of other people's money that Buffett invested at his historical return on equity — north of 20% for decades. The float cost Berkshire nothing in most years because its insurance operations broke even or turned an underwriting profit. In years when underwriting was profitable, the float had a negative cost: Berkshire was being paid to hold other people's money. The economic equivalent is a bank that charges negative interest on deposits — the depositor pays the bank to keep their cash, and the bank invests it. No rational person would accept that deal, yet insurance policyholders accept it every day because they're buying risk transfer, not making an investment decision.
The insight extends far beyond insurance. Amazon holds seller funds for approximately two weeks before disbursing — during which time Amazon invests or deploys that cash. With over two million active sellers generating hundreds of billions in gross merchandise value, the float is enormous. PayPal held $36 billion in customer balances as of 2023 — funds sitting in PayPal accounts between receipt and withdrawal. Credit card companies operate on float by design: Visa and Mastercard process transactions where merchants wait 1–3 business days for settlement while the card network holds the funds. Travel agencies, subscription platforms, prepaid card issuers, loyalty programmes — float exists anywhere money changes hands with a time delay between collection and disbursement.
The power of float is not the money itself. It is what happens when float meets compounding. A one-time $167 billion investment is valuable. A $167 billion pool that renews automatically — because new premiums flow in as old claims flow out — compounds year after year without requiring the capital to be raised, borrowed, or earned through operations. Float is self-replenishing capital. That is what makes it structurally different from debt, equity, or retained earnings, and that is why Buffett called it "better than free money."
Section 2
How to See It
Float reveals itself wherever money is collected before the obligation to disburse it arrives. The diagnostic is the timing gap: who holds the cash between the moment it is received and the moment it must be paid out? If the answer is the business — and the business can invest or deploy the cash during the gap — you are looking at float.
The signal is not just the existence of the gap but its duration and predictability. A 48-hour settlement delay generates trivial float. A multi-year gap between premium collection and claim payment generates float that can fund entire investment portfolios. The businesses that generate the most valuable float are those where the gap is long, the volume is high, and the replenishment is automatic.
Insurance
You're seeing Float when an insurer reports its investment portfolio as a multiple of its equity base. Berkshire Hathaway's equity in 2023 was approximately $561 billion, but its total invested assets — funded largely by float — were far larger. The float-to-equity ratio reveals how much of the insurer's investment capacity comes from policyholder money rather than shareholder capital. Progressive, Markel, and Fairfax Financial all use float as an investment engine, but none approaches Berkshire's scale or discipline. The diagnostic: if an insurance company's investment portfolio dwarfs its net worth, float is the mechanism.
Payments & Fintech
You're seeing Float when a payment platform reports billions in "funds held for customers" on its balance sheet. PayPal reported $36 billion in customer funds in 2023. Stripe, Square, and Adyen all hold merchant funds during the settlement window. Apple holds gift card balances and Apple Pay Cash deposits. The amounts are enormous, the holding periods are short (hours to days), but the aggregate float across millions of transactions generates meaningful investment income at scale. When Stripe launched Stripe Treasury, it formalised what payment companies had always done informally: treat the settlement gap as a source of capital.
Marketplaces
You're seeing Float when a marketplace delays seller payouts while collecting buyer payments immediately. Amazon collects payment at checkout and disburses to sellers on a 14-day rolling cycle. Airbnb collects guest payments at booking and releases host funds 24 hours after check-in — for bookings made months in advance, Airbnb holds the float for weeks or months. Uber collects rider payments in real time and pays drivers weekly. Each delay creates a float pool proportional to the marketplace's transaction volume. Airbnb's cash and investments balance of $11.4 billion in 2023 included substantial funds held on behalf of hosts — float that Airbnb invested in short-term securities.
Investing
You're seeing Float when an investor values a company's float at zero because it appears as a liability on the balance sheet. Insurance float shows up as "unpaid losses and loss adjustment expenses" — a liability. But if the float is obtained at zero or negative cost, the liability functions as permanent, free capital. Buffett's central insight was that GAAP accounting misrepresents float by treating it as a burden rather than an asset. The investor who adjusts for the true cost of float — recognising that zero-cost float is economically equivalent to free equity — will arrive at a materially different valuation than the one who reads the balance sheet at face value.
Section 3
How to Use It
Decision filter
"Before designing any payment flow, transaction structure, or customer billing cycle, ask: can the timing between collection and disbursement be extended in a way that is fair to all parties — and can the resulting float be invested productively? Even a few days of float, at sufficient scale, generates meaningful capital."
The discipline is designing payment architectures that generate float without exploiting customers or counterparties. The ethical boundary matters: float that exists because of a legitimate processing delay is fair. Float that exists because a company deliberately delays payments it could make instantly is rent extraction. The best float businesses generate the gap naturally through the structure of the transaction, not through artificial friction.
As a founder
Design your payment flows with float in mind from day one. Most founders optimise payment architecture for speed — instant payouts, instant settlement, instant everything. Speed is a feature for the user. But every instant payout is float you're surrendering. The alternative is not to slow payments down unfairly — it is to design default payout cycles that balance user needs with float generation. Shopify pays merchants on a two-business-day cycle by default. Amazon uses a 14-day cycle. Neither is exploitative; both generate float that funds operations and investments. The calculation is straightforward: if your marketplace processes $1 billion annually and you hold funds for an average of seven days, your average float pool is roughly $19 million. At a 5% risk-free rate, that is $950,000 in annual investment income — money generated by payment architecture, not by selling more products.
As an investor
Float is the most frequently undervalued asset on insurance company balance sheets and the most commonly overlooked asset in marketplace and payments businesses. When analysing an insurance company, calculate the cost of float: (underwriting loss) ÷ (average float). If the cost is below the risk-free rate, the float is cheaper than Treasury bills — the company is borrowing at better-than-government rates. If the underwriting operation is profitable, the cost of float is negative — the company is being paid to borrow. Berkshire's float has had a negative cost in the majority of years since 1967. When analysing a marketplace or payments business, estimate the float pool by multiplying daily transaction volume by the average settlement delay. The float pool, invested at the risk-free rate, is a revenue line that most analysts ignore because it appears in interest income rather than operating revenue. It deserves separate valuation.
As a decision-maker
Recognise float in your own business even if you don't think of yourself as a financial company. Any business that collects payment before delivering the product or service generates float. SaaS companies that bill annually upfront hold twelve months of revenue before fully delivering the service. Gift card issuers hold cash until redemption — and a meaningful percentage of gift cards are never redeemed (breakage), turning the float into permanent capital. Subscription platforms, prepaid models, deposit-heavy businesses — all generate float that can be invested or deployed rather than sitting idle in an operating account. The decision is whether to treat that float as working capital (defensive) or as investable capital (offensive). The answer depends on your confidence in the stability and predictability of the float pool.
Common misapplication: Treating float as free money without accounting for the cost. Float is only valuable when its cost — the underwriting loss in insurance, the operational cost of delayed payouts in marketplaces — is below the return it generates. An insurer that loses 10% on underwriting to generate float it invests at 5% is destroying value. The float's investment return must exceed its acquisition cost, or the business is paying more for the capital than it earns on it.
A second misapplication is treating float as permanent when the underlying business is cyclical. Insurance float can evaporate during catastrophic loss years. Marketplace float shrinks when transaction volumes decline. Float is self-replenishing in stable businesses and self-depleting in volatile ones. The investor who values float as a perpetuity in a cyclical business will overpay.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
Float rewards patience, scale, and underwriting discipline in equal measure. The leaders below didn't stumble into float. They engineered business architectures around it — recognising that holding other people's money for free is the most efficient form of financing ever devised, provided you never forget that the money is not yours and will eventually be claimed.
Buffett's entire investment career has been a float strategy. He acquired National Indemnity in 1967 for $8.6 million because he understood that insurance premiums were a source of investable capital that no other industry matched. Over the next five decades, he built Berkshire's insurance operations into the largest float generator in the world. GEICO, acquired in stages between 1976 and 1995, provides auto insurance to over 17 million policyholders. General Re, acquired in 1998 for $22 billion, added reinsurance float at global scale. Berkshire Hathaway Reinsurance writes policies on catastrophic risks — hurricanes, earthquakes, industrial disasters — where premiums are collected years before potential claims. The float grew from $39 million in 1970 to $167 billion by 2023. Buffett deployed that float into equities (Apple, Coca-Cola, American Express), operating businesses (BNSF Railway, Precision Castparts, Dairy Queen), and Treasury bills ($189 billion in cash reserves by late 2024 when nothing met his hurdle rate). Every dollar of Berkshire's investment portfolio has been funded, in part or in full, by other people's money. The cost of that capital has been zero or negative in the majority of years since 1967. No other investor in history has had access to a comparable funding advantage.
Bezos built Amazon's working capital model around negative cash conversion — customers pay immediately, Amazon pays suppliers and sellers later. The gap between collection and disbursement is float. On the marketplace, Amazon collects buyer payments at checkout and disburses to third-party sellers on a rolling 14-day cycle. With marketplace gross merchandise value exceeding $700 billion annually by 2023, even a 14-day delay generates an average float pool measured in tens of billions. On the retail side, Amazon's inventory turns are fast enough that products are sold and cash is collected before Amazon pays its suppliers — a negative cash conversion cycle that means suppliers are financing Amazon's inventory, not the other way around. Bezos used this float to fund Amazon's relentless expansion: AWS infrastructure, Prime benefits, fulfilment centre buildouts, and R&D spending that exceeded $85 billion in 2023. The float didn't appear on the income statement as revenue. It appeared on the balance sheet as a cash position that perpetually exceeded what the core retail business alone would justify. Bezos never described it as float, but the architecture was identical to Buffett's: collect money early, pay money late, invest the difference.
Section 6
Visual Explanation
The diagram shows the three-stage flow of float: money in (premiums, payments, deposits), the float pool (cash held during the timing gap), and money out (claims, payouts, withdrawals). The downward arrow from the float pool to the investment block captures the mechanism: while the cash sits between collection and disbursement, it can be deployed. The three variables below — volume, duration, and cost — determine whether float is a superpower or a liability. The spectrum at the bottom maps the critical distinction: negative-cost float (Berkshire in most years) is the most powerful financing source in capitalism. Positive-cost float (insurers who underwrite poorly) can destroy the very capital it was meant to enhance.
Section 7
Connected Models
Float sits at the intersection of time value, compounding, and capital structure. Its power comes not from the money itself but from the interaction between temporarily held capital and the investment opportunities available during the holding period. The connections below map how float generates returns, how it creates competitive advantages, and where its logic breaks down.
Reinforces
[Compounding](/mental-models/compounding)
Float is the fuel that compounding feeds on. Buffett's genius was recognising that the insurance float, renewed year after year as new premiums replaced old claims, created a permanent capital pool that compounded indefinitely. A $167 billion float pool compounding at even 10% generates $16.7 billion in annual investment returns — returns that are reinvested into the float-funded portfolio, generating returns on returns. The reinforcement is recursive: float enables compounding, and compounding increases the value of the float. Over fifty-seven years, Berkshire's float-funded investments compounded from a few million dollars into hundreds of billions. Remove the float, and the compounding engine loses its fuel.
Reinforces
Capital Allocation Options
Float expands the capital allocation option set by providing capital at below-market cost. A CEO with access to float has a seventh option beyond the standard six (reinvest, acquire, pay down debt, dividends, buybacks, hold cash): deploy other people's money at your own return rate. Buffett used Berkshire's float to fund acquisitions, equity investments, and cash reserves without issuing shares or taking on debt. The float gave him a structural cost-of-capital advantage over every other investor — he was deploying capital at zero cost while competitors raised equity at 10% or debt at 6%. Capital allocation decisions that are marginal for a CEO funding with expensive capital become clearly positive for one funding with free float.
Reinforces
Time Value of Money
Float is the time value of money made tangible. The entire concept depends on the principle that a dollar today is worth more than a dollar tomorrow — because today's dollar can be invested. The policyholder who pays a $1,000 premium today for a claim that might occur in three years has transferred three years of time value to the insurer. At a 5% rate, three years of time value on $1,000 is roughly $158. Multiply that by millions of policyholders and billions in premiums, and the aggregate time value transfer is the source of float's economic power. Float does not create value from nothing. It captures the time value that the counterparty surrenders when they pay early and receive late.
Section 8
One Key Quote
"If our premiums exceed the total of our expenses and eventual losses, we register an underwriting profit that adds to the investment income the float produces. When such a profit is earned, we enjoy the use of free money — and, better yet, get paid for holding it."
— Warren Buffett, 2004 Berkshire Hathaway Annual Letter
The sentence "get paid for holding it" is the line that separates float from every other form of capital. Equity has a cost — shareholders demand returns. Debt has a cost — lenders charge interest. Retained earnings have an opportunity cost — the returns shareholders could earn elsewhere. Float with a negative cost has no cost at all. The company is being paid to hold the capital. The economic equivalent does not exist anywhere else in corporate finance. No bank offers negative interest rates on loans. No investor accepts negative returns voluntarily. Yet insurance policyholders effectively pay Berkshire to hold their money, because the transaction is structured as risk transfer, not as a financing arrangement. The policyholder perceives a premium payment. Buffett perceives a free deposit with a bonus attached.
The quote also reveals Buffett's hierarchy of preferences. He does not celebrate float volume. He celebrates float profitability. The distinction is everything: an insurer that maximises float volume by writing cheap premiums generates float at high cost. An insurer that writes disciplined premiums at profitable levels generates float at negative cost — less float, but infinitely better float. Buffett would rather hold $100 billion in free float than $200 billion in float that costs 3%. The discipline of maximising float quality over float quantity is the operational expression of the margin of safety applied to capital structure.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Float is the closest thing to a cheat code in capitalism. Every other source of capital costs money. Equity is expensive — shareholders want 10–15% annual returns. Debt costs interest. Retained earnings carry the opportunity cost of what shareholders could earn elsewhere. Float, obtained through disciplined underwriting or natural payment-timing gaps, costs nothing. In the best years, it pays you.
The pattern I track most closely: float disguised as liabilities. The balance sheet is designed to mislead on float. Insurance reserves, customer deposits, deferred revenue, prepaid subscriptions, gift card balances, merchant funds in transit — all appear as liabilities. A surface reading suggests the company owes money. The deeper reading recognises that many of these "liabilities" are permanent, self-replenishing capital pools that fund operations and investments at zero marginal cost. Airbnb's balance sheet shows billions in "funds payable to hosts." That line item is float. Apple's balance sheet shows billions in gift card liabilities. That is float too — and roughly 10% of gift cards are never redeemed, converting the float into permanent profit.
The biggest mistake I see: founders who race to instant payouts. Fintech companies have spent the last decade competing to make payments faster. Instant seller payouts. Real-time settlements. Same-day ACH. Each innovation is genuinely good for the recipient — but each one surrenders float that the platform could have used. The winners will be the companies that offer instant payouts as a premium feature (earning a fee) while maintaining standard payout cycles (earning float) as the default. Shopify charges sellers for accelerated access to their funds. That is monetising both sides of the float equation — earning investment income on the standard cycle and earning fees from sellers who want faster access.
Float compounds asymmetrically with scale. A small insurer with $100 million in float invests in bonds and earns 5%. Berkshire, with $167 billion in float, acquires entire companies and earns 15–20% on invested capital. The same concept — hold money, invest it — produces fundamentally different outcomes at different scales because the investment opportunity set expands with the pool size. This is why float-based business models tend toward winner-take-most outcomes: the largest float pool has access to the best investment opportunities, which generates the highest returns, which funds more insurance writing, which grows the float pool further.
The AI-era float opportunity is in prepaid compute. Cloud providers collect annual commitments from enterprise customers — AWS, Azure, and GCP all offer significant discounts for one- to three-year prepaid commitments. Those prepayments are float: the customer pays now, the compute is delivered over the contract term. With hyperscalers collecting hundreds of billions in committed revenue, the float embedded in cloud computing contracts is arguably the largest pool of technology float ever assembled. The companies that recognise this are using the float to fund the infrastructure capex that generates more committed revenue — a flywheel powered by prepaid customer capital.
Section 10
Test Yourself
Float is present in more business models than most people recognise. The scenarios below test whether you can identify float in contexts beyond insurance — and whether you can distinguish valuable float from costly float that masquerades as an advantage while destroying value underneath.
Is this mental model at work here?
Scenario 1
A subscription software company bills enterprise customers annually in advance. The company collects $500M in January for services delivered evenly across the year. The CFO invests the unearned balance in Treasury bills yielding 5%. By December, the company has earned $12.5M in interest income on customer prepayments.
Scenario 2
An insurance startup offers ultra-low premiums to gain market share. In its first three years, it writes $2B in premiums and builds a float pool of $800M. The startup's underwriting loss ratio is 115% — for every $1 in premiums, it pays $1.15 in claims and expenses. The $800M in float is invested in a portfolio returning 6% annually.
Scenario 3
A marketplace for home services collects full payment from homeowners at booking and pays service providers 5 business days after job completion. A typical job is booked 10 days before the appointment. The marketplace processes $3B in annual transactions. A fintech competitor launches with instant payouts to providers.
Section 11
Top Resources
The float literature is concentrated in Buffett's annual letters, which contain the most detailed public record of float-based capital management ever produced. Beyond Berkshire, the academic insurance literature and marketplace economics research provide the analytical tools for evaluating float in non-insurance contexts.
The definitive primary source on float as a business strategy. Buffett discusses float in nearly every letter from 1967 onward, with the most detailed treatments in 2004, 2009, and 2016. The 2004 letter provides the clearest articulation of float economics. The float tables — tracking Berkshire's float volume and cost from 1967 to the present — are the single most valuable data set for understanding how float compounds over decades. Free, and more educational than any MBA course on insurance economics.
Thorndike's profile of Buffett and the other Outsider CEOs explains how capital allocation discipline — including the deployment of float — separates great companies from good ones. The Berkshire chapter is the most concise external analysis of how Buffett built the float machine, but the Tom Murphy and Henry Singleton chapters show how similar principles (holding cheap capital and deploying it at high returns) operate in media and conglomerates without the insurance structure.
Kilpatrick's encyclopaedic treatment of Berkshire Hathaway traces the float strategy from National Indemnity's $17 million in 1967 float to the $167 billion pool of 2023. The detailed acquisition histories of GEICO, General Re, and Berkshire Hathaway Reinsurance document how Buffett built the largest float machine in corporate history — including the underwriting disasters (General Re's asbestos liabilities, Berkshire's super-cat losses) that tested the strategy's limits.
Stone documents how Amazon's marketplace and working capital structure generate float that funds the company's expansion. The book details the negative cash conversion cycle, the marketplace payment architecture, and the strategic deployment of seller funds — making explicit the float dynamics that Amazon's financial statements obscure through standard accounting treatment. Essential for understanding float outside the insurance context.
Cunningham organises Buffett's letters thematically, with a dedicated section on insurance and float that collects the most important passages from five decades of letters. The thematic organisation makes it the fastest path to Buffett's float philosophy for readers who don't want to read every letter chronologically. The insurance section alone justifies the book — it distils the float strategy into roughly fifty pages of edited wisdom.
Leaders who apply this model
Playbooks and public thinking from people closely associated with this idea.
Float — The timing gap between collecting money and paying it out, and the investment returns earned on that gap. The three variables that determine float's value: volume, duration, and cost.
Leads-to
Economies of [Scale](/mental-models/scale)
Float scales superlinearly with volume. An insurer writing $10 billion in annual premiums generates a float pool roughly ten times larger than one writing $1 billion — but the investment management cost of deploying $100 billion is not ten times the cost of deploying $10 billion. The fixed costs of investment infrastructure (analysts, systems, relationships) spread across a larger float pool, improving the net return per dollar of float. Berkshire's scale advantage is compounding: its $167 billion float pool can access investments — entire companies, billion-dollar equity positions — that smaller float pools cannot reach. Scale begets better investment opportunities, which beget higher returns on float, which beget more capital to deploy.
Leads-to
[Flywheel](/mental-models/flywheel)
Float feeds the flywheel by providing low-cost capital for reinvestment. Amazon's negative cash conversion cycle generates float that funds fulfilment centre expansion, which increases delivery speed, which attracts more customers, which increases transaction volume, which generates more float. The flywheel effect converts a financial structure (float) into an operational advantage (faster delivery) into a competitive advantage (more customers) into a financial structure (more float). Each revolution of the flywheel increases the float pool, and each dollar of float funds the next revolution. Bezos never used the word "float" in shareholder letters, but the Amazon flywheel is, at its financial core, a float-powered compounding machine.
Tension
Margin of Safety
The tension between float and margin of safety emerges in underwriting discipline. Generating more float requires writing more insurance policies — but writing more policies at competitive premiums increases the risk of underwriting losses that exceed the investment returns the float generates. The margin of safety demands underwriting only at premiums that ensure the float's cost remains below its investment return — even in catastrophic years. Buffett resolved this tension through disciplined underwriting: Berkshire walks away from business where premiums are inadequate, even if that means the float pool shrinks. The willingness to accept smaller float rather than cheaper float is the margin of safety applied to capital structure. Most insurers resolve the tension in float's favour — writing aggressive premiums to maximise volume — and pay the price in catastrophic loss years.
My operational rule: wherever money changes hands with a time delay, someone is earning float — and it should be you. The exercise for any business is to map every payment flow and identify the timing gaps. Customer pays at order → you pay supplier at shipment: gap. Annual subscription collected → service delivered monthly: gap. Deposit collected → event occurs: gap. Each gap is float. The question is whether you're investing it or letting it sit in an operating account earning nothing. The difference, compounded over decades, is the difference between Berkshire Hathaway and every other insurance holding company.
Scenario 4
A property and casualty insurer reports $50B in float, $3B in annual investment income, and an underwriting profit of $200M. The CEO announces plans to grow float by 20% over three years by entering new insurance lines with less pricing discipline. Analysts project float will reach $60B.
