Why Bitcoin price forecasts disagree: ARK, CoinShares and the assumptions behind the numbers

Put issuer, original document date, forecast horizon, output type and critical assumptions side by side. Price target, hypothetical upside, market capitalisation and valuation floor are not interchangeable. Nor is a 2028 estimate directly comparable with a 2030 case or a 2035 decade forecast.

ARK, CoinShares and Bitwise manage or provide investment products linked to digital assets. That does not invalidate a model, but it is a reason to read its assumptions and disclosures with care.

ARK’s 2025 Bitcoin study estimates how much capital could accrue to Bitcoin from institutional portfolios, a digital-gold role, emerging-market demand, treasury holdings and on-chain services. Its 2030 cases change the assumed market sizes and penetration rates, leading to approximately $300,000, $710,000 or $1.5 million per coin.

ARK’s 2026 report updates the context: its Bitcoin market-cap outlook is approximately $16 trillion by 2030. It says the assumed size of the digital-gold market rose while the estimated safe-haven penetration in emerging markets fell sharply. Citing the 2025 scenarios alongside this newer report requires two dates and two labels.

CoinShares instead models potential owners, their annual allocation from disposable income, a flow-to-market-cap multiplier and Bitcoin’s issuance. Its updated framework uses approximately 560 million estimated 2025 owners, less than 1% of disposable income allocated annually, and a 3.5× flow multiplier. The output is about $248,000 at the end of 2028 and $317,000 at the end of 2029.

The 3.5× value is an input based on an interpretation of past market behaviour, not a physical law. Real marginal flows can have a different impact in thin, exuberant or stressed markets. CoinShares itself discusses adoption, policy and recession as sources of model error.

CoinShares states that an earlier version of its model put end-2028 Bitcoin near $133,000, while its newer estimate is around $248,000. The change is ($248,000 ÷ $133,000 − 1) × 100 ≈ 86.5%. This is a revision to a forecast for the same future date, not an 86.5% realised investment return.

The practical lesson is to store each version with its publication date. Overwriting an old target with a new one makes it impossible to judge whether the model anticipated the market or merely followed it.

ARK’s 2026 report states a roughly $16 trillion Bitcoin market-cap outlook for 2030. If, purely for illustration, we divide $16 trillion by 20.5 million coins, the result is about $780,000 per coin; dividing the same rounded market cap by 21 million gives about $762,000. The 20.5 million supply assumption appears in ARK’s earlier 2030 modelling.

These are MyProfitFactor’s arithmetic examples using rounded inputs, not new ARK per-coin targets. Market cap itself is a forecast, circulating and economically active supply differ, and no one can trade an entire market capitalisation at its quoted last price.

Bank analyst targets often circulate through interviews and rewrites of client research. Without the dated original, a number may be revised, may describe an intrayear high instead of a year-end close, or may be conditional on a policy event. We do not print it as an authenticated bank forecast until the underlying note or bank publication is available.

A strong institution can publish a poor forecast. Evidence quality depends on the accessibility of the original argument, version history and assumptions, not just the reputation of its logo.

For every new edition, record issuer and analyst, original URL, publication and revision dates, exact wording of the horizon, reported value, output category, key inputs and disclosed product interests. Preserve the previous value when a model changes. Once a target year has passed, compare it with an explicitly chosen observation date and price source.

Avoid declaring a long-run target right or wrong from a price touched years earlier. Write down whether the target was year-end, anytime-before, median scenario or an illustrative ceiling before scoring it.

Treat the forecasts as maps of assumptions. Ask which input drives most of the outcome, what independent data could contradict it, and how far the market could fall while the thesis is being tested. A wide range of published outputs is information about uncertainty; it is not a menu of guaranteed returns.