ETH Perpetual Open Interest Reached 70.2 Percent of Bitcoin's, Rank 1 of 73 Sessions. Funding Closed on Its 31-Day Median.

By Kresmion Research, September 14, 2026

ETH perpetual open interest closed Saturday at 70.2 percent of Bitcoin's, the highest reading of 73 sessions, and ETH funding closed on its 31-day median. Kresmion's derivatives aggregate marked the September 12 UTC day at 23:59:01 UTC with ETH open interest at $11.910 billion against Bitcoin's $16.971 billion, a ratio of 0.7018, and the open-interest-weighted ETH funding rate closed the next day at 0.005941 percent against a 31-day median of 0.005682 percent, the 53rd percentile (Kresmion crypto_derivatives_aggregate). The largest ETH perpetual book of the window is carried at an ordinary cost. Two sessions before the Fed decides, the ETH spot ETFs took $189.3 million on CPI day while the Bitcoin funds gave back $482.4 million over four sessions (Kresmion crypto_etf_aggregate, computed 2026-09-14 09:36:50 UTC), and crypto's 30-day correlation to US equities sits near the bottom of the 548 to 561 days Kresmion has measured since July 2024.

Key takeaways

The open-interest record

A perpetual futures contract has no expiry, so the book is carried indefinitely and paid for through a funding rate, and open interest is the dollar size of what is outstanding. At the September 12 daily close, 23:59:01 UTC, Kresmion's aggregate read ETH $11.910 billion and Bitcoin $16.971 billion, a ratio of 0.7018. On the constant four-venue window Kresmion holds since July 3, Binance, Bybit, Hyperliquid and OKX, that is rank 1 of 73 complete sessions, against a window median of 60.9 percent and a minimum of 54.4 percent.

The window's three highest readings are its three most recent complete sessions: September 12 at 70.18 percent, September 11 at 69.63 and September 13 at 68.74. Today's 69.22 percent at 12:09 UTC sits outside the ranked window because the UTC day has not closed.

The move is on every venue. Between the September 4 close at 23:55 UTC and the September 12 close at 23:59 UTC, ETH open interest rose on Binance from $5.524bn to $5.861bn (plus 6.1 percent), on Bybit from $1.818bn to $1.946bn (plus 7.0 percent), on Hyperliquid from $2.219bn to $2.488bn (plus 12.1 percent) and on OKX from $1.473bn to $1.615bn (plus 9.6 percent). Over the same eight days Bitcoin fell on Binance from $8.593bn to $7.982bn (minus 7.1 percent), on Bybit from $4.403bn to $4.116bn (minus 6.5 percent), on Hyperliquid from $2.935bn to $2.749bn (minus 6.3 percent) and on OKX from $2.192bn to $2.123bn (minus 3.1 percent) (Kresmion crypto_derivatives_snapshot). Four venues up on one asset and four down on the other rules out a venue-mix artifact.

It is also not the price. ETH/BTC closed at 0.030827 on September 4 and 0.032689 on September 12, up 6.0 percent (Kresmion correlation_close_history), while the open-interest ratio went from 0.6088 to 0.7018, up 15.3 percent. The dollar value of the Bitcoin book fell outright over the same eight days, from $18.123bn to $16.971bn, minus 6.4 percent, while ETH rose from $11.033bn to $11.910bn, plus 7.9 percent. A mark-to-market effect moves both books the same way.

Against the underlying asset, the September 12 close of the ETH book is 3.888 percent of a $306.3 billion market capitalisation read at 12:12:21 UTC today, and Bitcoin's is 1.086 percent of $1,562.2 billion read at the same time (Kresmion crypto_prices), so the ETH book is 3.6 times as large relative to its own asset.

Nobody is paying a premium to hold it

Funding is the rent on a perpetual position: when the book is long and impatient the rate rises and the longs pay the shorts. Open-interest-weighted ETH funding closed September 13 at 0.005941 percent per eight hours, about 6.5 percent a year, against a 31-day median of 0.005682 percent, the 53rd percentile. Bitcoin closed at 0.006507 percent against a 0.005911 percent median, the 60th percentile (Kresmion crypto_derivatives_aggregate). Neither rate is elevated, and the ETH rate is the more ordinary of the two.

On CPI day itself the ETH funding close read 0.000069 percent, the 3rd percentile of 31 days, and that is the session in which the ETH spot ETFs took their fourth-largest print of the window. A book that grows while the rent on it goes to nearly zero is the shape a delta-neutral carry book makes: the carry book holds a spot or fund leg on the other side and has no reason to pay a premium for the perpetual one.

The ETF leg

The five ETH spot ETFs Kresmion aggregates took a net $189.3 million on September 11 (Kresmion crypto_etf_daily flow flag farside, aggregated in crypto_etf_aggregate, computed 2026-09-14 09:36:50 UTC): rank 4 of the 70 sessions in the window that carry a flow print, the 95.7th percentile, and 1.54 percent of the complex's $12.28 billion in assets. By fund, ETHA $148.8M, ETHW $29.1M, FETH $11.4M, ETHE and QETH zero. The day before, the same complex gave back $51.5 million.

The nine Bitcoin funds in Kresmion's aggregate went the other way: minus $17.0 million on September 11 and minus $482.4 million over the four sessions from September 8 to 11, rank 21 most negative of 81 rolling four-session windows. Set against $93.58 billion of assets that four-day total is 0.5 percent of the complex, a trickle in asset terms. Month to date through September 11 the Bitcoin funds are still positive at plus $193.4 million, the ETH funds at plus $188.1 million on roughly one eighth of the asset base. A daily flow print is a creation and redemption count rather than a demand measure.

Third-party weekly tallies run close but not equal, at minus $462.7 million for US spot Bitcoin funds and plus $196.9 million for the Ethereum funds (gncrypto.news); the gap against Kresmion's figures is roster and window. September 12, 13 and 14 carry no flow yet, so Friday September 11 is the last real print.

Equity beta out, dollar beta in

Crypto's correlation to equities has gone to approximately zero and its correlation to the dollar has gone deeply negative, both at once. ETH's 30-day correlation to QQQ reads plus 0.012 against a trailing 252-day baseline of 0.476, the 1.28th percentile of 548 days; ETH to the dollar index reads minus 0.622, the 0.89th percentile of 561 days; Bitcoin to the dollar index reads minus 0.552, the 0.53rd percentile (Kresmion correlation_breaks, snapshot date 2026-09-11, fetched 2026-09-14 02:00 UTC). Bitcoin to the Nasdaq composite moved from a 0.468 baseline to 0.128, and Bitcoin to QQQ from 0.451 to 0.111.

Those four pairs came out of a scan of 969, and a first-percentile reading on roughly ten of them is what chance alone produces at that scan size. The cluster averages do not have that problem: the 132 crypto against US equity pairs average 0.105 and the 22 crypto against dollar pairs average minus 0.402, both the 3.6th percentile of the cluster's own history. Two clusters at the same percentile, pointing opposite ways, is the swap described above.

The same snapshot logged only 19 formal breaks of the 969 pairs, the lowest count in the 15-session window; one is ETH against XLK, the technology sector fund, which flipped sign from a 0.463 baseline to minus 0.028.

What Kresmion's regime engine reads

Kresmion's macro regime engine printed Strong Risk-Off at 100 percent conviction today, computed 11:33:10 UTC, on a smoothed risk score of minus 0.5287. Across the 138 daily rows since April 30 that is the second-lowest on record, the 1.45th percentile, 2.41 standard deviations below a mean of minus 0.0148, and only three rows carry the Strong Risk-Off label at all: June 25, yesterday and today, the first back-to-back pair. The heaviest factor is the least negative of the four: liquidity, weighted 40 percent, reads minus 0.2722, against risk appetite at minus 0.7623, volatility at minus 0.7485 and growth at minus 0.6234. Every input is a five-day momentum z-score, so the reading says everything moved toward risk-off quickly rather than that levels are distressed: the high-yield credit spread sat at 270 basis points on September 10, the 10.46th percentile of 822 sessions since July 2023, while the US 2-year yield at 4.56 percent was its highest since July 2024 and the 2s10s spread 39 basis points. Kresmion's broadcast rule holds a regime flip for five days before announcing it, so yesterday's transition cannot be published as a regime change before September 17, the day after the decision. The engine's last flip, to Risk-Off on September 2, was covered in Kresmion's September 4 brief.

The hike being priced is a hike into an energy shock: core CPI ran 0.3 percent month over month on Friday against a 0.2 percent forecast, the only line of the release that missed, while core year over year printed 2.4 percent, the lowest since March 2021, with gasoline up 3.9 percent on the month and energy up 16.3 percent over the year (CNBC, US Inflation Calculator), and the Saudi East-West pipeline, 7 million barrels a day of design capacity across roughly 750 miles, is shut after drones launched from Iraq, with Brent above $108 and WTI near $103 this morning (CNBC, Euronews). Polymarket's 25 basis point hike leg for Wednesday and Kalshi's contract on the same decision both read 79.5 percent at 12:58 UTC today, with the hold at 19.5 percent on each.

The alternative explanations

Four readings compete for the open-interest record, and the data kills one outright.

The first is a venue-mix artifact: Hyperliquid's ETH book grows faster than its Bitcoin book, and because Kresmion's aggregate sums four venues, a shift in venue weight could move the ratio without anything happening in the positions themselves. The per-venue split above closes this: there is no venue whose behaviour can be removed to recover the old ratio.

The second is basis, or cash-and-carry: a spot or fund position held against a short perpetual, earning the spread between them. This is the reading the data supports best. Funding on its median means nobody is paying a premium to be long the perpetual, which is what a short perpetual leg inside a delta-neutral structure looks like from the outside, and the ETH ETF creations on the same days supply the cash leg. One caveat belongs here rather than in the evidence: $209.3 million of ETH landed at Coinbase-labelled addresses on September 11 across 59 transfers from 40 senders, close to the creation size, but Coinbase custodies ETHA, FETH and ETHW and Kresmion's labels do not separate an exchange address from a custody address, so that figure restates the ETF flow rather than confirming it independently. The independent upstreams counted here are three: the venue open-interest feeds, the fund flows and the daily close series.

The third is the price denominator, which the cross handles: ETH/BTC rose 6.0 percent over the eight days while the ratio rose 15.3 percent, and the Bitcoin book shrank in dollars while ETH's grew.

The fourth is directional leverage: a crowd positioning long ETH into the Fed decision. It cannot be excluded from open-interest data, which reports size and never side. What argues against it is the price of holding the position: a crowd that wanted the exposure badly enough to add 7.9 percent to the dollar size of the book in eight days would normally bid the funding rate above its median, and on CPI day the rate closed in the 3rd percentile of 31 days.

A fifth reading belongs to the correlation section and is partly conceded: the 30-day window opened in late July, inside a stretch in which crypto drew down while US equities did not, so a window artifact contributes to those readings.

The evidence against

Funding cuts both ways. A rate on its median means no premium-paying crowd, which supports the carry reading, and equally means the funding tape shows no positioning pressure to point at.

There is no forced flow on OKX, the one venue Kresmion carries liquidation data for. September 12, the session that set the open-interest record, ran $3.92 million across 511 events, rank 71 of 74 complete days; September 13 ran $21.11 million, rank 53; September 11 ran $44.77 million, rank 19 (Kresmion liquidation_events, read 2026-09-14 12:12 UTC). A book this size that nobody is paying a premium for and nothing is forcing out is a quiet book.

The CPI reaction argues against the dollar-channel framing above. In the 60 minutes after the 12:30 UTC print, Bitcoin rose 0.85 percent and gold rose 1.27 percent while the dollar index fell 0.11 percent (Kresmion macro_event_reactions), a dollar-weakness response on a release that moved the Polymarket hike leg up 19 points in one five-minute snapshot. That table carries a trap worth stating: 12:30 UTC is before the US cash open, so the 30-minute equity reactions have no sample and the 60-minute figures are the opening gap.

The on-chain tape underneath all of this is concentrated: nearly half of the raw ETH exchange-inflow tape over the last four weeks is one address depositing every day, and every figure in this article excludes it. Kresmion's whale table is a biased sample of large transfers and never a census.

Finally, the record itself is a 73-session record. Kresmion's four-venue window starts on July 3, when the venue count went from three to four, and readings before that break are not comparable. A rank 1 of 73 is a real extreme inside a ten-week sample, and ten weeks is the whole claim.

What would change the read

Four conditions in the same tables would move this away from the carry reading, each observable rather than forecast.

ETH spot ETF flows turning negative for consecutive sessions while ETH open interest holds near current levels. The carry reading requires the fund leg and the perpetual leg to move together; redemptions against a book that does not shrink describe a directional position instead.

Open-interest-weighted ETH funding closing above its 31-day median for several sessions in a row. The 53rd percentile close on September 13 is the centre of the distribution; a run above it is the tape of a crowd paying for its position.

The open-interest ratio dropping back under the 60.9 percent window median after Wednesday's decision, which would describe a book placed around the event.

ETH's 30-day correlation to QQQ returning above 0.3. The current plus 0.012 against a 0.476 baseline is what the correlation section rests on, and a move back toward that baseline ends the dollar-channel description regardless of what open interest does.

Frequently asked questions

Does a record open-interest ratio predict the price?

No. Open interest measures the dollar size of the outstanding position and says nothing about which side is crowded, because every contract has a long and a short. Kresmion holds 73 complete sessions on this four-venue window, far too short a sample to fit a relationship between the ratio and any subsequent return, and none is claimed here. The reading is descriptive: the ETH book is at its largest relative size of the window and is carried at its ordinary cost.

What does funding sitting on its median actually mean?

Funding is the periodic payment between longs and shorts that keeps a perpetual contract near spot. A high positive rate means longs are paying to hold, the signature of a one-sided book; a rate at its own 31-day median means the payment is ordinary and neither side is bidding for the exposure. ETH's rate closed September 13 at 0.005941 percent against a 0.005682 percent median, the 53rd percentile, while the book itself sat at a window record.

Are the ETF creations and the on-chain deposits two separate pieces of evidence?

No, and the article treats them as one. Coinbase is the custodian for ETHA, FETH and ETHW, so an in-kind creation arrives at a Coinbase-labelled address and is recorded as an exchange inflow, and Kresmion's labelling does not separate an exchange wallet from a custody wallet. The $209.3 million of September 11 ETH deposits and the $189.3 million of creations are therefore two views of one event.

How much of the correlation extreme survives the size of the scan?

The scan covered 969 pairs, so roughly ten first-percentile readings come from chance alone and no single pair carries much weight. The cluster averages survive that: the 132 crypto against US equity pairs average 0.105 and the 22 crypto against dollar pairs average minus 0.402, both the 3.6th percentile of the cluster's own history. The 30-day window also opened in late July, inside a stretch in which crypto drew down and US equities did not, so part of the reading is a window effect.

What is on the calendar between now and the end of the week?

The Fed decision lands Wednesday September 16 at 18:00 UTC with projections and a press conference, and the calendar consensus carries a target of 4.0 percent against 3.75 percent. US retail sales print the same morning at 12:30 UTC, the Bank of England decides Thursday and the Bank of Japan Friday. The Fed's communications blackout runs to September 17 (Atlanta Fed).

Sources

Kresmion crypto_derivatives_aggregate, total_oi_usd and oi_weighted_funding_pct, ETH and BTC daily closes 2026-07-03 to 2026-09-14, last read 12:09 UTC; September 12 close stamped 23:59:01 UTC.

Kresmion crypto_derivatives_snapshot, per-venue open_interest_usd, four venues, closes 2026-09-04 23:55 UTC and 2026-09-12 23:59 UTC.

Kresmion crypto_prices, BTC and ETH market capitalisation, 2026-09-14 12:12:21 UTC.

Kresmion correlation_close_history, ETH/BTC daily closes, 2026-09-04 and 2026-09-12.

Kresmion crypto_etf_aggregate, total_net_flow_usd and assets by asset and fund, 2026-09-10 and 2026-09-11 prints, computed 2026-09-14 09:36:50 UTC.

Kresmion correlation_breaks, baseline and current_corr across 969 pairs, snapshot date 2026-09-11, fetched 2026-09-14 02:00 UTC.

Kresmion macro_regime_history, regime, conviction, risk_score_smoothed and factor fields, 2026-09-14, computed 11:33:10 UTC.

Kresmion hy_credit_spread and sovereign_yield_snapshots, the high-yield credit spread and US 2-year and 10-year yields, 2026-09-10.

Kresmion polymarket_history, market id 2252245, the September Fed ladder, 2026-09-14 12:58 UTC.

Kresmion kalshi_markets, September 2026 Fed decision contracts, 2026-09-14 12:58 UTC.

Kresmion macro_event_reactions, 60-minute reactions to the 2026-09-11 12:30 UTC CPI release.

Kresmion macro_calendar_events, US CPI actuals and forecasts for 2026-09-11 and the September 16 to 18 central bank calendar, fetched 2026-09-14 11:48 UTC.

Kresmion whale_transactions, clean ETH exchange inflows 2026-08-17 to 2026-09-13, one sweeper address excluded from every figure.

Kresmion liquidation_events, OKX 24-hour totals and events, 2026-09-14.

CNBC, August 2026 CPI report: https://www.cnbc.com/2026/09/11/cpi-inflation-report-august-2026.html

US Inflation Calculator, August 2026 CPI detail: https://www.usinflationcalculator.com/inflation/us-cpi-august-2026/100073342/

CNBC, the Saudi East-West pipeline shutdown: https://www.cnbc.com/2026/09/14/iran-war-saudi-arabia-east-west-pipeline-oil.html

Euronews, oil above $108 on the Hormuz attack and the pipeline shutdown: https://www.euronews.com/business/2026/09/14/oil-surges-past-107-as-hormuz-attack-and-saudi-pipeline-shutdown-rattle-markets

gncrypto.news, weekly US spot Bitcoin and Ethereum fund flows to September 11: https://www.gncrypto.news/news/bitcoin-etfs-463m-outflows-ether-funds-197m/

Federal Reserve Bank of Atlanta, FOMC blackout periods: https://www.atlantafed.org/news-and-events/press-room/blackout-periods