JPMorgan Gave Bitcoin 10% a Year. Then Told You Not to Own It.

The 2027 long-term assumptions put BTC at 10.0% a year against 7.5% for US large caps, then rule it out of the portfolio on risk. The return was conceded. The risk is an assumption, and at realised volatility the ranking flips against the gold they do recommend. That is the tell in a document whose real answer is private markets.

Every long-term capital market assumption document is a political settlement dressed as a forecast. The numbers are the easy part; the interest is in what the author needs the numbers to say, because the allocation conclusion is usually written before the covariance matrix is estimated.

J.P. Morgan's 2027 edition is worth reading closely for one reason. It is the first year the house has put a ten-year return on bitcoin, and it did so while simultaneously declaring bitcoin uninvestable on risk grounds. Those two statements cannot both be innocent. Either the return forecast is a genuine view, in which case excluding the asset needs a risk argument that survives contact with its own inputs, or the return forecast exists to make the exclusion look generous rather than hostile. This piece tests which.

What the document is actually arguing

Strip the chapter headings and the 2027 assumptions make one argument. Growth is upgraded 20bps to 4.2%. A global balanced portfolio rises 60bps to 7.0%, against 6.4% a year ago. And the diversifier that carried the last four decades is declared structurally impaired, because projected stock-bond correlation is both higher and unstable.

That last point is the engine of the whole book. If bonds no longer hedge equity risk, the 60/40 is not a portfolio, it is two correlated bets. The document's answer is not to shrink risk but to change the diversifier: add 30% in alternatives and the balanced portfolio reaches 7.5% with a Sharpe ratio 21% higher than a public-markets-only version. A blend of private equity, venture and direct lending adds 80bps of return and 11% of Sharpe; a blend of diversified hedge funds, gold, infrastructure and transportation adds 22% of Sharpe while holding return flat.

So read plainly: the 2027 LTCMA is not a return document, it is an allocation document. Its thesis is that the liquid diversifier died and the illiquid one replaces it. Everything else, bitcoin included, is arranged around that conclusion.

Worth noting what work the word “unstable” is doing. The same adjective is applied to stock-bond correlation and to bitcoin's correlation to other assets. For bonds it is a reason to buy something else. For bitcoin it is a reason to buy nothing. One word, two verdicts, and the difference is not analytical.

Debasement is named as the reason, then denied as the conclusion

Here is the sentence that organises the document, from the executive summary: “Higher inflation volatility, ongoing risk of currency debasement and growing opportunities to capture active alpha reinforce the role of alternatives as essential portfolio diversifiers.”

Thirty pages later, on the capital-rewiring chapter, the same driver appears with the opposite conclusion attached: “the suggestion that broader currency debasement will precipitate a rush into crypto assets is also wide of the mark.”

That is the whole thing in two sentences. Debasement risk is the stated reason to allocate to alternatives. Debasement risk is explicitly not a reason to allocate to the one liquid asset whose entire thesis is debasement. Both sentences are defensible individually. Together they describe a preference, not a finding.

The bitcoin chapter then does something stranger. It opens by conceding the mechanism: BTC “mostly trades like a high beta risk asset that tracks global money supply growth.” That is a debasement hedge described in the language of a beta. Having named the transmission channel correctly, the chapter disqualifies the asset on the grounds that the channel is unstable — which is true, and which is also true of the correlation assumptions underpinning every alternative in the book.

Then the contradiction becomes explicit. “Although we think it is premature to create a 10-year forecast for bitcoin or to integrate it into our covariance matrices, examining its potential range of outcomes is instructive.” They decline to forecast, then forecast. They decline to model the covariance, then publish a standalone covariance matrix. The hedging is not caution. Caution would be one or the other.

The return is conceded; the risk is where the argument lives

Take the forecast at face value, because it is more generous than the coverage suggests. Base case 10.0% a year for ten years, bear −11.2%, bull +16.6%, from a spot reference of $83,000 on 30 September 2026. Terminal values on those paths: roughly $215,000 base, $25,000 bear, $386,000 bull.

Set that against US large caps at 7.5% and the scale of the concession is clear. With US inflation assumed at 2.40%, the base case is 7.4% real against 5.0% real for US equities. J.P. Morgan is forecasting bitcoin to beat the S&P by roughly two and a half points a year, in real terms, for a decade.

So the document is not bearish on bitcoin's return. It cannot be accused of that. What it says is that the return is real and the risk is unusable. Which means everything now rests on the risk inputs, and those are the only numbers in the chapter the author actually chooses.

One exhibit deserves more attention than it got. Exhibit B prints arithmetic return 20.9% against compound return 10.0%. That 10.9-point gap is the volatility tax — the amount of average return that compounding takes back. For US large caps the same gap is 1.45 points. Bitcoin's drag is 7.5 times larger. The exhibit is quietly admitting that the distribution is too wide for a point estimate to carry much weight, and then printing a point estimate anyway. That is not dishonesty; it is the standard practice of the genre. But it does mean the headline 10.0% should be read as a centre of mass, not a forecast.

One assumption decides the verdict

Now the number the conclusion hangs on. Exhibit B assigns bitcoin an annualised volatility of 55.0%. Against that, and their own 10.0% return, the Sharpe ratio is 0.124 — the lowest of any asset in the table. Gold at 5.5% return and 17.2% volatility gives 0.134. ACWI gives 0.240. A diversified hedge fund blend gives 0.593 on a 5.9% volatility input.

The problem is that 55% is not a measurement, it is a legacy figure. The same chapter states that “volatility has declined.” Measured rather than assumed: realised volatility over the past year, from daily closes through 20 October, is 43.8%, and the 90-day reading is 37.8%. At 43.8% the Sharpe becomes 0.155. At 37.8% it becomes 0.180.

Look at what that does to the ranking. At their assumed volatility, bitcoin finishes last in their own table, and gold — which they recommend as a diversifier — finishes above it. At realised volatility, bitcoin passes gold on the same return forecast. The asset they exclude outranks the asset they prescribe, using nothing but their own numbers and a volatility input that reflects observable data rather than a prior.

This is where the document stops being a forecast and becomes a position. Not because 55% is indefensible — a long-horizon assumption should probably use a through-cycle figure, and 55% is defensible as that. But because a single input, chosen after the conclusion, is doing all the discriminating work between “diversifier” and “uninvestable.”

The chart is the sensitivity, at their return, volatility and correlation inputs for ACWI, US Aggregate and bitcoin. At 55% volatility the Sharpe-maximising bitcoin weight in a 60/40 is about 2.2%. At the one-year realised figure it is 6.0%. At the 90-day reading it is 9.7%. Nothing in the return forecast changed across those three cases. Only the risk number did.

To be clear about what this does and does not show. It is not an argument for a double-digit allocation; at any of these volatilities the weight is small and the error bars are wide. It is an argument that the distance between “no allocation” and “a small allocation” is set by an assumption, and that the assumption is moving in one direction.

There is a second asymmetry worth naming, because it cuts the other way and should be said rather than buried. Bitcoin is marked daily at 55% volatility. The diversified hedge fund blend that tops their Sharpe table is marked at 5.9%, an appraisal-based figure that is known to understate true risk. Unsmooth it by a typical factor and its Sharpe falls toward 0.30 — still above bitcoin, so the ranking survives. But the two numbers are not measured on the same instrument, and only one of them is priced off a continuous market. The document compares a daily-marked asset to self-reported ones and lets the comparison stand unqualified.

The headline is a choice between 6% and 11%

There is a second editorial decision upstream of all of this. The 10.0% base case is not the output of a model, it is a weighted average of three models — weighted 65:30:5 “on our confidence in the underlying drivers.”

Framework one, supply growth against dollar and global money supply growth: base 10.7%. Framework two, bitcoin as a complement to gold in a portfolio: base 10.7%. Framework three, network adoption measured by actual blockchain usage, with little assumed benefit from tokenisation migrating to other chains: base −3.2%.

Two of the three frameworks produce an identical 10.7%. That is not a triangulation of independent methods, it is one view expressed twice with different labels, and it carries 95% of the weight. The only framework that asks whether anyone actually uses the chain is weighted at 5%, and its entire effect is to drag the answer from 10.7% to 10.0%.

Weight the three equally and the base case becomes 6.1%, the bear −14.0%, the bull 13.5%. Still a respectable forecast — comfortably above cash, roughly a large-cap equity. Nothing in the underlying analysis changes. Only the confidence weights do.

So the honest range for the headline is somewhere between 6% and 11%, and the document publishes the top of it while describing the asset as uninvestable. Both cannot be the considered view. What reconciles them is that the return number is not there to be acted on; it is there to demonstrate open-mindedness so that the allocation verdict reads as analytic rather than editorial.

What would prove this wrong

The claim here is narrow, so it is easy to falsify. It is not that bitcoin deserves a large allocation, nor that J.P. Morgan is wrong about its risk. It is that the exclusion is driven by a volatility assumption that is already stale relative to its own text, and that the document's structure consistently resolves ambiguity toward the intermediated product.

Two things would break it. If bitcoin's realised volatility climbs back above 55% and stays there through 2027, the assumption is vindicated and the exclusion stands on its own numbers without any need for a commercial explanation. And if projected stock-bond correlation falls back toward zero and 60/40 resumes diversifying, then the 30% alternatives push loses its rationale and the entire 2027 stance — not merely the bitcoin chapter — comes apart at the joint.

One item flagged rather than laundered. The same chapter notes tokenised assets, money market funds among them, at roughly $40bn with “scope for rapid growth.” J.P. Morgan operates a tokenisation franchise. That is not evidence of bad faith and this piece does not claim it. It is context: the house arguing that public blockchains are not an investable monetary asset also sells the permissioned version of the same rail. Readers can weigh that themselves, provided it is on the page.