Notes — Bill Miller on Amazon, Bitcoin, and Buying at a Discount to Future Value

Source-grounded literature notes. Own words. Citations by section. Bill Miller, interviewed by William Green for Richer, Wiser, Happier (RWH007), May 2022. Context: sharp tech and crypto declines. ARK Innovation Fund had retraced to March 2020 levels; Kathy Wood, James Anderson, and Dennis Lynch all down 50%+ from November 2021 peaks.

Four questions [Adler frame]

Q1 — What is it about? A full account of Bill Miller‘s investment philosophy at a moment of personal stress — May 2022, with Amazon and Bitcoin both down sharply and margin calls forcing liquidations. The central claim is that the dominant failure mode of investors is anchoring on present-day intrinsic value when the only thing that actually matters is value at a future date. Supporting this is Miller’s pragmatist epistemology (drawn from William James), his extended case study of Amazon across 25 years, his Bitcoin thesis, and his reflections on retiring from active fund management.

Q2 — How is it argued? Biographical and dialectical: William Green asks and Miller recounts, using market episodes (1987 crash, 2002 Amazon lows, 2008–09, May 2022) as evidence for general claims. The argument is self-referential — Miller’s own concentrated and uncomfortable positions are offered as proof of what his philosophy permits. The philosophical sections (William James, Frank Knight, Wittgenstein) provide the theoretical superstructure; the case studies flesh it out. Miller consistently acknowledges counter-cases (when the crowd is right, e.g. LTCM, Amaranth) while maintaining that the crowd is epistemically over-confident about its ability to identify them.

Q3 — Is it true? The future-value refinement of classical value investing is a genuine contribution, though it is also the standard DCF argument restated: any DCF already discounts future cash flows. Miller’s claim is more specific — that most practitioners treat the visible present metrics as proxies for intrinsic value, thereby systematically under-weighting optionality. The Frank Knight distinction between risk and uncertainty is orthodox and well-supported; the question is whether Miller genuinely operates with correct calibration in the uncertainty domain or whether his record is partly explained by running concentrated positions in industries that happened to have compounding tailwinds (tech, crypto). The 30-year threshold argument (from the Times of London journalist) is structurally sound — 30 years is long enough to reduce the false-positive rate from sustained lucky runs — though it does not account for selection bias among the population of fund managers who survive to be written about at all. [?] The Bitcoin supply-demand thesis is logically valid but ignores the possibility of demand collapse; Miller’s analogy to insurance presupposes a catastrophic scenario that may never arrive.

Q4 — What of it? For a non-professional investor, the practical takeaway is the future-value framing: when evaluating a holding, the key question is not ‘is this cheap on current metrics?’ but ‘what will this business be worth in 10 years and what am I paying for that?’ This reframing changes both the holding behaviour (patient, through volatility) and the error-type focus (avoiding selling Amazon at a 20× PE in 2001 is more important than avoiding buying a cigar-butt stock at 0.5× book). The second takeaway is Miller’s explicit comfort with Knightian uncertainty — an honest acknowledgement that the forward-looking investor cannot know the probability distribution of outcomes, only that the odds appear better than the consensus suggests.

Glossary

Future-value investing: Miller’s term for the approach of buying businesses at a discount to what they will be worth, rather than what they are worth today. The distinction matters because present-day intrinsic-value calculations anchor on past earnings, book values, and cash flows, while the entire value of a business depends on the future. A company that looks expensive at 50× earnings may be cheap if earnings will multiply 20×; a company that looks cheap at 5× earnings may be expensive if it is in terminal decline.

Knightian uncertainty: The domain, identified by economist Frank Knight (1921), where no probability distribution of outcomes can be established. Insurance, casinos, and actuarial pricing all operate in the risk domain (known probabilities). Stock investing, particularly in novel companies or assets, operates largely in the uncertainty domain. Miller holds that most institutions systematically avoid uncertainty because their mandates and benchmarks require explainable risk bounds — and that this avoidance is the structural source of opportunity for those who can tolerate it.

The four percent principle: James Anderson’s (Bailey Gifford) empirical observation that all S&P 500 returns over any long period concentrate in roughly 4% of publicly traded companies. The remaining 96% underperform cash. Implication: diversification to control tracking error virtually guarantees underperformance; the correct question for any potential holding is whether it is likely to be one of the 4%.

The Silver Blaze argument: Miller’s application of Sherlock Holmes’s ‘dog that didn’t bark’ reasoning to Bitcoin. In the Holmes story, the significance is that a guard dog failed to bark at the intruder — meaning the intruder was known to the dog. Miller applies this: venture capitalists whose job is to assess every new technology and short what is unlikely to succeed have not systematically bet against Bitcoin. Their silence is informative evidence that they believe Bitcoin will not go to zero.

Bitcoin as insurance: Miller’s core thesis: Bitcoin is an insurance policy against catastrophic failure of the fiat financial system. The expected loss if Bitcoin goes to zero is bounded (the premium paid); the expected gain in a genuine currency collapse is very large. The analogy to personal insurance (health, home, life) is that the policy holder hopes never to collect — hopes the policy is ‘worthless’ in the sense of never being triggered — but values the protection nonetheless.

Pragmatic umpire: From William James’s three-umpires parable. First umpire (naïve realist): ‘I call balls and strikes as they are.’ Second umpire (perspectivist): ‘I call them as I see them.’ Third umpire (pragmatist): ‘They aren’t anything until I call them.’ Miller adopts the third: perception structures reality; there is no purely objective vantage point, so the useful question is not ‘what is the truth?’ but ‘what frame is most useful for navigating the world?’ Applied to Bitcoin: the question is not ‘what is Bitcoin really?’ but ‘what job does it do in my portfolio?’

The 30-year threshold: Informal rule from a Times of London financial journalist: any fund manager who survives and outperforms for 30 years has crossed the threshold where their record can no longer be plausibly attributed to luck alone. The probabilistic reasoning is that even a manager with zero skill can string together impressive multi-year runs; 30 consecutive years of outperformance reduces the false-positive rate from lucky sampling to a negligible level.

Section notes

The 2022 context: regime change

May 2022. The Federal Reserve is tightening into a strong economy; the ARK Innovation Fund and the Bailey Gifford funds are down 50%+; MicroStrategy has retraced to pre-Bitcoin levels; Terra/LUNA has collapsed to near zero. Miller’s historical analogy: the Nifty Fifty collapse of 1973–74 (high-growth stocks fell from 90× PE to 9× PE over two years) and the 1982 environment (30-year Treasuries at 10% made equities unattractive on a relative basis). But the 2022 situation differs from 1929 and 1982 in one crucial way: the economy is strong rather than rolling over. In 1929, industrial production was already falling before the market crashed in October; in 1987, the same underlying dynamic held. In 2022, the economy is being tightened into, not collapsed.

This asymmetry matters for Miller’s posture: when the market falls due to strength being tightened rather than due to economic breakdown, the downside is more bounded and the recovery is faster. He is buying rather than reducing.

The future-value principle

Classical value investing asks: what is this business worth today, and can I buy it at a discount? Miller’s refinement: ‘100% of the information you have to value a business is based on the past, but 100% of the value depends on the future.’ Today’s P/E, P/B, and cash flow metrics are backward-looking constructions. They tell you nothing about whether the future will resemble the past.

The practical implication: buying at a discount to future value sometimes requires buying at what appears expensive on present metrics. Amazon in 2001 at $6 was not cheap relative to then-current earnings (it had no earnings). It was cheap relative to what Amazon would earn in 2015. The investor who sold because of the PE ratio bought the past and sold the future.

Miller explicitly distinguishes his approach from growth investing: growth investors often lack patience with the inevitable hiccups in a company’s growth rate and flee when fundamentals temporarily disappoint. Miller’s patience comes from his conviction about long-run future value — which allows him to hold through volatility that would force a shorter-horizon investor to exit.

Amazon: the 25-year case study in patience

The timeline:

1997 (IPO): Miller bought on the premise that internet-delivered books had structural cost advantages — a first-mover able to bypass physical retail distribution. Bezos recognised that Amazon could deliver any product, not just books; the book business was a beachhead.
1998 ($88 per share): Bought again after the stock had doubled. Already expensive on any present metric.
2001–02 (single digits): Stock fell from $90+ to $6. The bear case was bankruptcy; Ravi Suri at Lehman put a zero target on it. Miller’s countervailing evidence: (1) Amazon had positive free cash flow from its book business, which had a favourable returns policy (unsold books could be returned to publishers); (2) Jeff Bezos in 2001 told Miller his time was consumed by the balance sheet; in 2002, he said his time was consumed by the customer experience. The shift from defence to offence at the bottom of a crisis is the decisive signal. Miller also offered Bezos a $100–200M capital infusion if needed; Bezos didn’t need it, but the conversation revealed the balance sheet was solid.
2012–13 (LEAPS): Stock had fallen and appeared expensive on near-term metrics but underpriced the AWS and Prime optionality. Miller bought long-dated call options (LEAPS) — if Amazon returned to its old highs, the calls returned 5× rather than 2× (the gain from owning the stock). He then exercised the options rather than selling them, deferring tax.
May 2022 ($2,100): Amazon was 40–50% of Miller’s personal portfolio. The stock was down ~30% from its $3,500 peak. Miller’s view: Amazon should return to $2,100–$2,200 on a 1–2 year horizon; a 10-year holder would be dramatically higher.

The Carol Loomis problem: Loomis (Fortune) had warned that even if Amazon succeeded, the gains would be captured by employees rather than shareholders through stock-option dilution. Miller raised this with Bezos directly. Bezos invited him to present to the Amazon board. The board initially resisted — arguing that tech companies could not compete for talent without generous options. Shortly after, Amazon switched to restricted stock. Result: share count moved from approximately 460 million to approximately 480 million over 15–20 years (~5% dilution), then buybacks began. Loomis’s dilution concern was neutralised by the board’s decision, which Miller’s intervention had partly influenced.

Frank Knight and the case for uncertainty tolerance

Miller draws on Frank Knight’s 1921 distinction between risk and uncertainty:

Risk: The probability distribution of outcomes is known. Insurance companies, actuaries, and casinos price risk with actuarial tables and expected-value arithmetic.
Uncertainty: The probability distribution is not known and cannot be established from historical data. Keynes termed this ‘irreducible uncertainty.’ Most of the interesting decisions in investing — especially in novel assets or businesses — fall into this domain.

Most institutional investors operate as if uncertainty can be converted into risk through models. Miller argues this is the dominant failure mode: models built on historical data impose a spurious probability distribution on genuinely unknowable futures, creating false precision. His advantage is accepting the discomfort of not knowing the distribution while still making a directional bet.

The contrast with Buffett: Buffett in March 2020 and in 1987 was paralysed by extreme uncertainty — he didn’t act aggressively in either episode. Miller acted in both. He attributes this partly to temperament and partly to the key analytical move: when the cause of a market decline is not economic breakdown but excess strength being corrected by policy, the downside is bounded and the recovery is predictable.

The 1987 crash: anatomy of a rare correct macro call

In October 1987 the market fell 20% in a single day — twice as deep as the 1929 crash on a one-day basis. The received wisdom was that a 1929-style depression would follow. Miller’s counter: 1929 had seen industrial production already falling through July–October as the market rose — the economy was rolling over into the crash. In 1987, the economy was overheating, not collapsing. The Fed was raising rates because the economy was too strong, not because it was weak. When the crash came and the Fed cut rates, Miller’s read was that the economy would remain strong. His fund had 25% cash going into the crash; he deployed it in the following month. The fund was the best-performing fund in the country in 1988.

This is a rare example of a macro call Miller is willing to defend: not a prediction but an observation that the economic conditions (strong economy + tightening + high PE + bond yields near the implied equity return) had been misread by the market as structurally similar to 1929 when they were structurally different.

Bitcoin: the insurance thesis

Miller’s position: Bitcoin is not productive (Buffett is right about that), but investment is not about owning productive assets — it is about making money. The pragmatic question is not ‘what is Bitcoin’s intrinsic value?’ but ‘what job does Bitcoin do in my portfolio?’

The job it does: insurance against catastrophic financial failure. The supply schedule is fixed (21 million coins, halvings every four years, terminal supply reached in 2140). The demand outlook: Miller believes demand will grow faster than the supply growth rate (~1.7% pa, declining). The conclusion: long-run appreciation is likely, and the loss if Bitcoin goes to zero is bounded by the premium paid.

The Wenceslao Casares insight: Casares (Argentine entrepreneur) explained to Miller that his family had been wiped out multiple times by the Argentine government over 150 years — bank nationalisations, debasement, capital controls. He said: with Bitcoin, the government cannot take it. This gave Miller the frame for Bitcoin as insurance against precisely the kind of government expropriation that is invisible to an investor raised in the US but very salient to investors from Argentina, Estonia, Lebanon, or Ukraine.

The Silver Blaze argument in full: the venture capital community is paid to assess every technology and short what they believe will fail. They have not shorted Bitcoin; they have invested heavily (2021 saw more crypto venture investment than all previous years combined). Their silence on the bear case is as informative as a guard dog that doesn’t bark — it implies the technology is legitimate.

The stablecoin risk: Miller is alert to the Shadow Banking 2.0 argument (law professor Hillary Allen): stablecoins without transparent backing and without a Federal Reserve backstop are structurally analogous to the reserve fund that ‘broke the buck’ in 2008, triggering the commercial paper freeze. This is a risk to the crypto ecosystem, but not to Bitcoin directly — Bitcoin has no backing requirement and no run mechanism.

Pragmatist epistemology in investing

Miller draws on William James’s ‘On a Certain Blindness in Human Beings’ (1899) and the three-umpires parable to argue that most critics of Bitcoin have made up their minds first and then searched for confirming evidence. The checklist:

Buffett and Munger: ‘It’s not a productive asset.’ True, but irrelevant — investment is about making money, not owning productive assets.
Munger: ‘It’s anti-government and encourages behaviour outside the regulatory system.’ This is a values disagreement, not an analytical claim.
Critics in general: ‘Somebody will invent something better.’ Fidelity Digital Assets’s counter: Bitcoin is like the wheel — nobody reinvented the wheel. Ethereum adds features but serves a different function; they are not in competition for the same job.
‘Bitcoin is a climate catastrophe.’ Quantified: Cambridge Energy Institute estimates Bitcoin uses ~0.6% of world electricity. Laundry uses ~7%. Air conditioning uses more.

Miller’s rule: when a claim is normative but stated as if factual, and when it lacks quantification, treat it as an emotional reaction dressed up as analysis. Find the data.

Technical analysis: supply and demand visualisation

Miller does not believe in chart patterns as predictive. He uses technical analysis as a way to visualise supply and demand dynamics: when there are 2,000 new lows and 6 new highs, there are 2,000 sets of shareholders who are unhappy and selling. This is not a mystical pattern — it is a count of motivated sellers. When motivated sellers become scarce (because they have already sold), prices stabilise and reverse. The technician’s signal (‘market bottoming’) is a summary of this supply-demand exhaustion.

The crypto caveat: in crypto, technicals are more predictive than in equities because there are no fundamentals to fall back on. Prices in crypto are determined entirely by psychology, and large populations of unsophisticated participants behave in well-documented ways. Equities have a fundamental floor that limits how far technicals can drive prices below intrinsic value.

Career retrospective: the comeback as the greater achievement

Miller argues his 2009–2020 recovery period — during which Miller Value Partners was simultaneously in the top 1% for 1-, 3-, 5-, and 10-year periods — was more impressive than his famous 15-year consecutive market-beating run (1991–2005). His reason: achieving top-1% performance across all rolling periods simultaneously is harder than beating the index in most years. In some years during the 15-year run, 60% of money managers beat the market; in others, only 10% did. The run reflects partly rising odds in favourable conditions. The post-2009 record demonstrates consistent excellence across varying conditions.

His retirement rationale: compliance constraints made active management increasingly painful. The 7-day trading prohibition (triggered when Samantha McLemore sold from the public fund, creating overlap with his personal positions) meant that during the May 2022 sell-off he was prohibited from transacting precisely when he most needed to manage margin calls. He was forced to liquidate Amazon and Bitcoin at low prices to meet margin requirements. He considers this an institutional failure, not a personal one — but it accelerated his decision to step back.

The 30-year threshold: a Times of London financial journalist told Miller that the statistical threshold at which performance ceases to be explainable by luck is 30 years of outperformance. Miller crossed it. Bill Gross’s famous ‘snake eyes’ dismissal (comparing Miller’s 12-year run to rolling dice) is simply wrong numerically: the odds of rolling snake eyes 12 consecutive times are millions to one.