Forty-six days of negative funding

Between roughly 1 March and 15 April 2026, perpetual funding on bitcoin stayed negative for 46 consecutive days — the longest such stretch since the period after FTX. For anyone running a carry position, that is the income line changing sign. The arithmetic, worked through.

Perpetual futures have no expiry date, so they need a mechanism to stay anchored to spot. Funding is that mechanism: at fixed intervals, usually every eight hours, one side pays the other at a rate proportional to how far the perpetual has drifted from its index. When the perpetual trades above spot, longs pay shorts; when it trades below, shorts pay longs.

The second case is the unusual one. It means the derivative is persistently cheaper than the asset it references — more participants are willing to pay to be short than are willing to be paid to be long. Markets spend most of their time in the opposite condition.

During the 2026 episode, bitcoin traded at $74,287 at the mid-April measurement, shorts were paying on the order of 0.5–1% per week to maintain positions, and open interest rose approximately 12% over the preceding month while they did.¹

Restating the rate at common scale

The September rate of 0.0028% per eight hours is worth roughly 3.1% annualised — thin, unremarkable carry consistent with a market that has stopped being frightened without becoming enthusiastic.² The spring inversion at a 0.75% weekly midpoint was worth close to 39% annualised in the other direction. Funding quoted per interval is easy to underestimate; restated annually it is frequently the largest single line in the economics of a leveraged position.

What an inversion costs a carry book

A cash-and-carry position is long spot and short the perpetual. It carries no net directional exposure, and in a positive-funding regime it collects the rate. When funding turns negative, that same position pays.

Roughly five to six and a half per cent of notional over six and a half weeks, on a position whose entire premise is that it takes no directional risk. For strategies targeting high single-digit annual returns from carry, an inversion of that duration erases most of a year of income.

The more general point concerns correlation. The conditions that invert funding — sustained bearish positioning, elevated hedging demand, falling prices — are the same conditions under which most other exposures are performing badly. Carry's worst periods coincide with everything else's worst periods, which is precisely the relationship a market-neutral description tends to obscure.

The risk in carry is not that it is volatile. It is that its bad periods are correlated with everything else's bad periods.

Leverage compounds this directly. At one times notional a 5% funding cost over six weeks is a poor quarter for that allocation; at four times it is most of the allocation. The decision about leverage has to be made before it is known which regime is coming.

Open interest is the disambiguating variable

Negative funding alone is ambiguous. It can reflect genuine bearish consensus, or a temporary imbalance while a large participant works an order. Open interest resolves the ambiguity.

Falling open interest alongside negative funding describes a market closing positions — risk leaving. Rising open interest alongside negative funding describes something different: new short positions being opened at those levels by participants willing to pay approximately 39% annualised to hold them. In 2026, open interest rose about 12% over the month while funding stayed negative.¹ That is risk accumulating on one side, not leaving.

The consequence concerns the exit. Positions opened in the same window at similar prices with similar leverage share maintenance margin levels. The resulting cluster in April 2026 sat between $80,000 and $85,000.¹ That is not a forecast that price reaches those levels; it is a statement that if it does, the move through them is mechanically amplified by forced buying rather than damped by discretionary selling.

Bitcoin traded above $80,000 in late August 2026. By early September, short liquidations accounted for 65.5% of all futures liquidations.²

The limits of the signal

Two precedents exist for negative funding stretches of comparable length: November 2022, when bitcoin moved from roughly $15,500 to $23,000 for a 48% gain over two months, and June 2021. Both resolved upward, by 48% and 65% respectively, within sixty days.¹

Two observations do not constitute a base rate. A pattern with a perfect record over two instances is exactly the kind of relationship that appears robust in historical testing and fails in application, and the confidence such a record inspires is itself a reason for caution.

Funding is also least informative when it appears most dramatic. During a genuine liquidity event the rate stops describing sentiment and starts describing plumbing — withdrawals suspended somewhere, a venue's price detached from the wider market, one side of the book unable to arbitrage. The number continues to print and no longer means what it usually means. It degrades similarly in thin markets, where a single large account can set the rate, and over long horizons, being a read on days rather than quarters.

What it does reliably measure is crowding: not what participants hold, but what they are willing to pay to keep holding it. That is a statement about the shape of the distribution of outcomes and about how an eventual unwind is likely to behave. It is not a statement about direction, and the distance between those two readings is where most of the mistakes made with this data originate.