What Bitcoin is worth

Our approach says we avoid anything we cannot value. It also says we look at open networks with a fixed or predictable supply. Bitcoin is the largest of those networks, and it produces no cash flows, so the two statements need reconciling. This essay sets out why we think demand for assets like it will grow, how we think about its value, which methods we reject, and where the argument is weakest.

Figures are as of October 5, 2026, unless stated. On that day one bitcoin traded at about $86,140, for a network value of about $1.73 trillion [1].

No cash flows, no DCF

A business is worth the cash it will hand to its owners over its life, discounted for time and risk. Bitcoin hands nothing to anyone. There is no coupon, no dividend and no claim on an issuer. Any yield offered on it comes from lending it to someone, which turns a bearer asset into a credit exposure. A discounted cash flow model gives zero, and no choice of discount rate changes that.

The same model gives zero for gold, which has been held as a store of value for millennia. Of the roughly 222,600 metric tons above ground, 45% is jewelry, 21% bars and coins and 18% central bank reserves [2]. Industrial use explains little of its price. What holders pay for is a monetary premium: the value of an asset that others will accept later, that no one can print, and that carries no promise someone could break.

That premium is the only source of value Bitcoin has, and it is circular. Bitcoin is worth holding because others hold it and will hold it later. A premium of this kind can last for centuries, as gold's has, or disappear within a year. Economists at the European Central Bank argue that Bitcoin's fair value "is still zero" [3]. On cash flows they are right. By the same reasoning gold's fair value would be zero, and no market in living memory has priced it that way. So the useful question is not whether Bitcoin has intrinsic value in the DCF sense. It has none. The question is how much monetary premium it can hold, for how long, and with what chance of losing it.

The answer depends first on how much demand there is for money that no government issues. That demand rests on what governments do with the money they do issue.

Why fiat money keeps losing value

We expect the major fiat currencies to keep losing purchasing power, faster than their central banks' targets imply and at times abruptly. Public debts of today's size have historically been reduced through inflation rather than repaid, and the pension promises of aging societies sit on top of them. We hold this view with more confidence than any view about Bitcoin's price. The rest of this section sets out why.

The record

The US dollar has lost 97% of its purchasing power since 1913, the year the Federal Reserve was founded [4]. The pound has lost about 95% since 1971, when the last link between the major currencies and gold was cut, and over 98% since 1945 [5]. Euro area prices rose 76% between the euro's launch in 1999 and the end of 2025 [6].

This erosion is policy, not accident. Central banks target inflation of about 2% a year, at which a currency loses half its value in 35 years. Recent years ran well ahead of that. After 2008, and again in 2020, central banks expanded their balance sheets on a scale without peacetime precedent: the Federal Reserve's peaked at $8.97 trillion in April 2022 and the ECB's at €8.84 trillion two months later [7][8]. Inflation followed, peaking at 9.1% in the United States in June 2022 and 10.6% in the euro area that October [9][10], while policy rates stayed near or below zero into 2022. US prices are 30% higher than in January 2020 [4], euro area prices were 24% higher by the end of 2025 [6], and none of that loss comes back.

Nor has inflation returned to target. US consumer prices have risen faster than 2% a year in every month since March 2021, five and a half years without a break; in August 2026 the rate was 3.4% [4]. Euro area inflation rose to 3.8% in September 2026, its highest in three years, with energy up 18.8% [11].

The debt

Global debt, public and private, passed $365 trillion in the first half of 2026, about 310% of world output [12]. The ratio to output has edged down in recent years, but the Institute of International Finance points out why: inflation raised nominal output, while the debt itself kept growing. The IMF puts global public debt at just under 94% of GDP in 2025 and expects it to reach 100% by 2029, a level previously seen only after the Second World War [13]. The Bank for International Settlements says much the same, and adds a finding that matters here: countries with higher debt show a weaker fall in inflation when their central bank tightens [14].

The United States runs deficits once seen only in recessions. It borrowed 6.1%, 6.2% and 5.8% of GDP in fiscal years 2023 to 2025 [15], with unemployment at around 4% [16]. The Congressional Budget Office expects a deficit of $1.9 trillion in fiscal 2026 and debt held by the public to rise from 101% of GDP this year to 120% in 2036 [17]. Over the longer term it projects 175% by 2056 [18]. Net interest is about $1.0 trillion this year and rises to $2.1 trillion by 2036 [17]; in fiscal 2025 it already exceeded spending on national defense, at $970 billion against $917 billion [19]. Federal debt stood at $40.3 trillion on October 1, 2026 [20].

Europe has less room than its lower deficits suggest. Italy's public debt is 139% of GDP and France's 118% [21]. In early October 2026, after a new austerity package, the French ten-year yield reached about 5%, its highest since 2002, and its spread over German debt the widest since the euro crisis [22]. In September 2022 a sell-off in long-dated gilts forced the Bank of England, which was meant to be tightening, to buy £19.3 billion of them within three weeks [23]. In Japan the central bank holds 46.7% of all government bonds [24]. When the central bank is the largest buyer of its government's debt, the line between monetary and fiscal policy is hard to find.

The pension promises

Headline debt figures leave out the largest commitments. Across the OECD there are about 33 people aged 65 and over for every 100 of working age; by 2054 there will be about 55 [25]. Public pensions already cost 15.9% of GDP in Italy and 15.1% in France [26]. Eurostat estimates the pension entitlements already accrued in social insurance schemes at between 200% and 400% of GDP in most EU countries, mostly unfunded and outside the headline debt [27]. In the United States, the reserves behind Social Security run out in 2034, after which incoming payroll taxes cover 83% of scheduled benefits [28].

Promises on this scale are rarely broken openly. Cutting pensions loses elections; raising taxes on a shrinking workforce slows growth. Paying the promised amount in money that buys less is the path of least resistance, and it needs no vote.

How such debts end

Over-indebted states have three exits: growth, default or inflation. Growth is welcome but rarely enough on its own. Reinhart and Rogoff's history of eight centuries of sovereign debt shows defaults coming together with "inflation, exchange rate crashes, banking crises, and currency debasements" [29]. Advanced economies usually prefer a quieter route. After 1945 the United States and the United Kingdom reduced their war debts largely by holding interest rates below inflation. Reinhart and Sbrancia estimate that this "financial repression" liquidated debt worth 3 to 4% of GDP a year, with real rates negative about half the time between 1945 and 1980 [30]. Savers paid for it.

Sargent and Wallace showed in 1981 that when fiscal policy dominates, a central bank cannot control inflation for good: tight money today only defers it [31]. Today's combination of record debt, unfunded pensions and central banks that already own large shares of government bonds is the setting they described.

At the extreme, states that lose fiscal control lose their currency. Prices in Germany rose about 29,500% in the single month of October 1923, and Hungary in 1946 and Zimbabwe in 2008 went further [32]. More recently, annual inflation reached 85.5% in Turkey in October 2022 [33], 192% in Lebanon over 2023 [34] and 289% in Argentina in April 2024 [35]. We do not expect this in the dollar or the euro. We expect the slower version, which is harder to see and does more damage over a lifetime.

Who is already acting on it

Central banks themselves have been buying the one reserve asset no government issues. They bought more than 1,000 metric tons of gold a year from 2022 to 2024 and 863 metric tons in 2025 [36]. In the World Gold Council's 2026 survey, 84% of emerging-market central banks and 61% of those in advanced economies named inflation concerns as relevant to holding gold [37].

This is the demand side of the valuation that follows. If savers and reserve managers look for assets outside the reach of fiscal need, the pool Bitcoin competes for grows. Whether Bitcoin earns a share of it depends on what a holder actually owns, and on whether it behaves as a store of value when that is tested.

What a holder actually owns

Before valuing a premium, it helps to be precise about the asset.

A supply schedule anyone can check. No more than 21 million bitcoin will exist. New coins are issued with each block, and the issuance halves every 210,000 blocks, roughly every four years. Since April 2024 each block issues 3.125 BTC: about 450 a day and 164,000 a year, or 0.82% of the 20.09 million in circulation [38]. The 20 millionth coin was mined on March 9, 2026 [39]. After the next halving, expected in April 2028, annual issuance falls to about 0.4%. Gold's above-ground stock grows by about 1.6% a year from mining [40]. No committee can change either schedule to meet a budget.

Scarcity alone is worth little; many scarce things have no price. What is unusual is that the holder can verify the supply rather than take it on trust. A full node, ordinary software on ordinary hardware with about 773 GB of storage [41], checks every block against the rules and can sum every unspent coin in existence with a single command [42]. No outsider can audit a gold vault or a central bank's balance sheet to that standard.

The rule has been tested. In August 2010 a bug let one transaction create 184 billion bitcoin. A fixed version was out within about five hours, and the valid chain overtook the faulty one [43]. In 2018 developers found a flaw that would have let a miner inflate the supply, and patched it before anyone used it [44]. Both cases show that the supply rule depends on software and on the people who write and run it. It is enforced, not guaranteed.

A bearer asset, if held directly. A coin held with its own keys is no one's liability. A share in an exchange-traded product that holds bitcoin is a claim on an issuer, a custodian and a trustee. The price may be the same; the asset is not. We avoid assets that exist only as a claim on an intermediary, and that applies here as much as anywhere.

A smaller float than the headline supply. Some coins will never move again. Chainalysis counted about 3.7 million bitcoin, around 20% of the supply at the time, that had not moved in five years, and treated them as lost [45]. About 1.1 million of the oldest coins are attributed, on circumstantial evidence, to Bitcoin's pseudonymous creator, and almost none have ever been spent [46]. None of these coins is provably destroyed. The effective supply today lies somewhere between about 16.4 and 20.1 million, and that range is itself a source of uncertainty.

Methods that do not hold up

Several valuation models circulate. We have looked at each and rely on none.

Stock-to-flow. This model fits price to scarcity, measured as the existing stock divided by annual issuance [47]. It called a price of $55,000 after the 2020 halving, and the market did pass that level in 2021. It has failed since. At today's stock-to-flow of about 122, the 2019 formula implies a price near $835,000, almost ten times the market. The deeper problem is that the model contains only supply. Halvings are known decades in advance, and a market that ignored them would be strange. Demand, the variable that matters, is missing.

Cost of production. Here the price is anchored to the marginal cost of mining a coin [48]. But mining difficulty adjusts every two weeks so that blocks keep arriving every ten minutes, whatever the computing power. When the price rises, miners add machines until the margin is competed away; when it falls, they switch them off. Costs follow the price rather than lead it [49]. Mining cost is a fair gauge of miner stress. It is not a floor.

Network effects. Models based on Metcalfe's law fit price to the square of the number of users, approximated by active addresses [50]. The fit has been reasonable over long periods. But an address is not a user, and a growing share of holders never appear on chain, because they hold through custodians and exchange-traded products. The model describes the past without explaining it.

Equation of exchange. This approach treats Bitcoin as the currency of a payment economy and solves MV = PQ for the money value needed to settle a given volume of transactions [51]. Bitcoin is little used for payments, and the velocity of coins held as savings is unknown. The answer depends on assumptions chosen to produce it.

None of these gives a number we would act on. That is the honest starting point.

A frame we can use: what the price assumes

A value that cannot be estimated with confidence can still be tested. Rappaport and Mauboussin suggest reversing the discounted cash flow: start from the market price, ask what it assumes, and judge whether the assumption is reasonable [52]. For Bitcoin the assumption to recover is adoption, meaning the share of the world's demand for a non-sovereign reserve asset that it will hold.

Gold is the obvious yardstick. The World Gold Council valued the above-ground stock at $29.0 trillion at the end of June 2026 [2]. Bitcoin's network value of $1.73 trillion is about 6% of that. The table shows what other shares of gold's value would mean per coin, using today's supply and the smaller float that excludes coins counted as lost. It is not a forecast. It turns assumptions into prices, so that a price can be read as an assumption.

Two adjustments turn the table into a valuation. The first is time: a share reached in ten years is worth less today, and at a required rate of 10% a year a dollar ten years out is worth about 39 cents now. The second is failure. Bitcoin's monetary premium could fall to zero through a protocol flaw, a lasting loss of security, or demand moving to something else. Together:

value today = probability of survival × share of gold's value × gold's value ÷ supply ÷ (1 + required rate)^years

Run in reverse from today's price, holding gold's value flat, the formula gives the product of survival probability and eventual share that the market is paying for:

Take the middle of the table. At a required rate of 10% over ten years, today's price is consistent with an even chance that Bitcoin reaches about a third of gold's value, and nothing otherwise. It is equally consistent with a near-certain path to a sixth. Whether that is cheap or dear depends on what one believes about those two numbers, and reasonable people differ by multiples. The frame does not settle the question. It makes the question precise.

Holding gold's value flat is conservative if our view of fiat money is right. The required rate is a nominal one, and if currencies lose value faster than they have, gold's value in those currencies rises with it, and the adoption share the price requires falls.

The frame has other limits. Gold is a yardstick, not a ceiling or a floor: the demand for reserve assets can grow or shrink, and Bitcoin could draw on bank deposits and government bonds as well as on gold. The survival probability is a judgment, not a measurement. And a monetary premium is reflexive. A falling price can itself lower the chance of survival, by cutting the miners' revenue that secures the network.

Where the margin of safety comes from

Graham's margin of safety is the gap between price and a conservatively estimated value, wide enough to absorb errors in the estimate [53]. For Bitcoin the estimate is too uncertain for that gap to be measured, and one possible outcome is zero. The margin has to come from elsewhere.

From the price. Buying when the price implies a modest adoption share leaves room to be wrong; buying when it implies near-certainty does not. The implied product can be recalculated at any price, and it moves a great deal. Measured against today's gold value, at 10% over ten years, it was about 0.23 at the high of October 2025 and about 0.10 at the low of June 2026.

From size. An asset that can go to zero should be held only in an amount whose loss would be survivable. That is a rule about the holder, not the asset, and it applies to any holding with a failure state.

From custody. The losses that have actually destroyed holders' coins came mostly from intermediaries, from Mt. Gox in 2014 to FTX in 2022, not from the protocol. Holding the keys removes that risk and replaces it with operational risk, which procedure can manage.

From time, and the absence of leverage. Bitcoin fell 84% from its 2017 peak, 78% from its 2021 peak, and 54% from the October 2025 high to the June 2026 low [54]. Only a holder who cannot be forced to sell can wait through drawdowns of that size.

The risks we watch

The security budget. Miners receive about 450 new coins a day, roughly $39 million at today's price, or about 0.8% of network value a year. Transaction fees added only 1.0% to miner revenue in 2025 and 0.65% so far in 2026 [55]. Each halving cuts the subsidy in two. Budish shows that the recurring payments to miners must stay large relative to the one-off gain from attacking the chain, so the cost of security rises in line with the value secured [56]. Carlsten and co-authors show that a chain paid by fees alone can become unstable [57]. If fees do not grow, and the price does not double with each halving, security per dollar secured falls. This is the most serious long-term question about Bitcoin, and the answer is not yet in.

Custody returning to intermediaries. US spot exchange-traded products, approved in January 2024 [58], held about 1.29 million bitcoin on October 5, 2026 [59]. Listed companies held about 1.28 million, of which Strategy alone held 848,000 [60][61]. Together that is about 13% of the supply. Governments hold some 650,000 more; the US order of March 2025 set up a reserve of forfeited coins that "shall not be sold" [62]. An asset designed to remove intermediaries is increasingly held through a few of them. Concentrated custody is also where a state under fiscal pressure can act most easily: in 1933 the United States ordered private holders to deliver their gold coin, bullion and certificates to the Federal Reserve [63].

Mining concentration. Over the past year two mining pools produced 46% of blocks, and three produced 59% [64]. Pools coordinate computing power rather than own it, and miners can switch pools in minutes. It is still a point of pressure.

Quantum computing. Coins whose public keys are already visible on chain could be taken by a large enough quantum computer. Estimates of the exposed amount range from 6.0 to 6.9 million bitcoin, about 30% of the supply, most of it in reused addresses that owners can fix by moving their coins [65][66]. About 1.9 million sit in older script types, including the early coins attributed to Bitcoin's creator, which no owner is likely to move. Proposals for quantum-resistant outputs and for retiring old signatures exist as drafts (BIP-360 and BIP-361) and are not activated [67]. Project Eleven's central scenario places a capable machine in 2033, within a range of 2030 to 2042 [66]. In a system without a central authority, migration is slow, and what to do about coins that never migrate is a governance question with no comfortable answer.

Behavior in a crisis. Bitcoin is often described as a hedge against monetary disorder, and the case for it rests on that role. Its record does not yet support it. The IMF found that its daily correlation with the S&P 500 rose from 0.01 in 2017 to 2019 to 0.36 in 2020 and 2021 [68], and it has stayed positive since. During the inflation of 2022 it fell 78% while gold held its value. In liquidity squeezes it has traded as a risk asset. A store of value has to prove itself in exactly the conditions we expect, and Bitcoin has not done so yet.

What we conclude

• We expect fiat currencies to keep losing purchasing power, driven by public debt, deficits and pension promises that cannot be met in money of today's value. That is the case for holding scarce, verifiable assets at all. It is not a case for paying any particular price.
• Bitcoin has no cash flows and cannot be valued like a business. We do not pretend otherwise.
• Its value is a monetary premium, best described as an option on adoption, weighted by the chance of survival.
• Every price implies an adoption share and a survival probability. Those implied numbers, not a price target, are what we judge.
• A margin of safety comes from the price relative to what it implies, from size, from custody and from time. It does not come from precision.
• What we are most confident about is the network's verifiability. What we are least confident about is the price at which anyone should own it.

We would change our view on Bitcoin if fee revenue fails to grow as the subsidy falls, if custody concentrates to the point where the rules become negotiable, if a flaw appears that consensus cannot fix, or if the move to quantum-resistant signatures stalls. We would change our view on fiat money if governments ran sustained primary surpluses and brought their debts down without help from inflation. If either happens, we will revise this essay and date the change.

Sources

1. CoinGecko, Bitcoin price and market capitalization, October 5, 2026.
2. World Gold Council, Above-ground stocks, end of Q2 2026.
3. U. Bindseil and J. Schaaf, ETF approval for bitcoin: the naked emperor's new clothes, ECB Blog, February 22, 2024. See also Bitcoin's last stand, November 30, 2022.
4. US Bureau of Labor Statistics, CPI-U and PCE price index, via FRED (CPIAUCNS, PCEPI), 1913 to August 2026; our calculation.
5. Office for National Statistics, long-run price index (CDKO), to August 2026; our calculation. Based on RPI, which runs slightly above CPI.
6. Eurostat, HICP monthly index (prc_hicp_midx), January 1999 to December 2025; our calculation.
7. Federal Reserve, total assets (WALCL), via FRED.
8. European Central Bank, total assets (ECBASSETSW), via FRED.
9. US Bureau of Labor Statistics, CPI-U.
10. Eurostat, HICP annual rate of change.
11. Eurostat, euro area flash estimate, September 2026, October 2, 2026.
12. Institute of International Finance, Global Debt Monitor, September 23, 2026, as reported by The Star.
13. IMF, Fiscal Monitor: Fiscal Policy under Pressure, April 2026.
14. Bank for International Settlements, Annual Economic Report 2026, chapter II: High public debt and shifting financial markets, June 2026.
15. US Office of Management and Budget, federal surplus or deficit as % of GDP (FYFSGDA188S), via FRED.
16. US Bureau of Labor Statistics, unemployment rate (UNRATE), via FRED.
17. Congressional Budget Office, The Budget and Economic Outlook: 2026 to 2036, February 2026.
18. Congressional Budget Office, The Long-Term Budget Outlook: 2026 to 2056, March 2026, as summarized by the Committee for a Responsible Federal Budget.
19. US Treasury, Monthly Treasury Statement, fiscal year 2025, table 9.
20. US Treasury, Debt to the Penny, October 1, 2026.
21. Eurostat, quarterly government debt (gov_10q_ggdebt), Q1 2026.
22. Fortune, France's debt crisis, October 2, 2026.
23. Bank of England, Financial stability buy/sell tools: a gilt market case study, Quarterly Bulletin, 2023.
24. Ministry of Finance Japan, JGB and T-bill holdings by type of holder, June 2026 (preliminary).
25. OECD, Pensions at a Glance 2025, November 2025.
26. Eurostat, pension expenditure (tps00103), 2024.
27. Eurostat, Pensions in national accounts: statistics, supplementary table 29, 2021.
28. Social Security Board of Trustees, 2026 Annual Report, June 2026.
29. C. M. Reinhart and K. S. Rogoff, This Time Is Different: A Panoramic View of Eight Centuries of Financial Crises, NBER Working Paper 13882, 2008; published as This Time Is Different, 2009.
30. C. M. Reinhart and M. B. Sbrancia, The Liquidation of Government Debt, NBER Working Paper 16893, 2011; Economic Policy 30(82), 2015.
31. T. J. Sargent and N. Wallace, Some Unpleasant Monetarist Arithmetic, Federal Reserve Bank of Minneapolis Quarterly Review 5(3), 1981.
32. S. H. Hanke and N. Krus, World Hyperinflations, Cato Working Paper, 2012.
33. TurkStat, Consumer price index, October 2022.
34. Central Administration of Statistics, Lebanon, December 2023 on December 2022, as reported by Crédit Libanais Economic Research.
35. INDEC, Índice de precios al consumidor, April 2024.
36. World Gold Council, Gold Demand Trends, central banks: full year 2024 and full year 2025. Figures are revised between editions.
37. World Gold Council, Central Bank Gold Reserves Survey 2026, June 2026.
38. Block height and supply from mempool.space and blockchain.com, October 5, 2026.
39. mempool.space, block 940,000, March 9, 2026.
40. World Gold Council, Gold Demand Trends, full year 2025 and Q2 2026.
41. blockchain.com, Blockchain size, October 4, 2026.
42. Bitcoin Core, gettxoutsetinfo, RPC reference.
43. Bitcoin Wiki, Value overflow incident (CVE-2010-5139).
44. Bitcoin Core, notice on CVE-2018-17144, September 20, 2018.
45. Chainalysis, bitcoin market data report, June 2020.
46. S. D. Lerner, The return of the deniers and the revenge of Patoshi, April 16, 2019.
47. PlanB, Modeling Bitcoin's Value with Scarcity, March 22, 2019.
48. A. S. Hayes, Bitcoin price and its marginal cost of production: support for a fundamental value, Applied Economics Letters 26(7), 2019.
49. J. E. Marthinsen and S. R. Gordon, The Price and Cost of Bitcoin, Quarterly Review of Economics and Finance 85, 2022.
50. T. Peterson, Metcalfe's Law as a Model for Bitcoin's Value, Alternative Investment Analyst Review 7(2), 2018.
51. C. Burniske and J. Tatar, Cryptoassets, 2017.
52. A. Rappaport and M. J. Mauboussin, Expectations Investing, 2001.
53. B. Graham, The Intelligent Investor, chapter 20.
54. Kraken, XBT/USD weekly prices.
55. mempool.space, block fees and block rewards, 2023 to 2026; our calculation.
56. E. Budish, Trust at Scale: The Economic Limits of Cryptocurrencies and Blockchains, Quarterly Journal of Economics 140(1), 2025.
57. M. Carlsten, H. Kalodner, S. M. Weinberg and A. Narayanan, On the Instability of Bitcoin Without the Block Reward, ACM CCS 2016.
58. US Securities and Exchange Commission, Release No. 34-99306, January 10, 2024.
59. bitbo.io, US spot ETF holdings, October 5, 2026.
60. bitcointreasuries.net, October 5, 2026.
61. Strategy Inc., Form 8-K, October 5, 2026.
62. The White House, Establishment of the Strategic Bitcoin Reserve and United States Digital Asset Stockpile, March 6, 2025.
63. Executive Order 6102, April 5, 1933.
64. mempool.space, mining pools, last 12 months.
65. Glassnode, Measuring Bitcoin's quantum-exposed supply, May 20, 2026.
66. Project Eleven, 2026 report, as summarized by PostQuantum.com, May 2026.
67. BIP-360 and BIP-361, bitcoin/bips repository, both drafts.
68. IMF, Crypto Prices Move More in Sync With Stocks, Posing New Risks, January 11, 2022.