The same five questions every week, so the series is comparable. Every answer carries its data and its source. Where the desk does not know, it says so.
Adding leverage, in aggregate, while specific corners of the alt complex remove it. Six of the eight largest perpetuals closed the week above a +2 open interest z-score. Bitcoin sat at $20.77bn (z +2.49), ether at $12.19bn (z +2.01), Hyperliquid at $2.70bn (z +2.14), Solana at $2.51bn (z +2.33) and Zcash at $1.15bn (z +2.52, the 99th percentile of its own history). Broader futures open interest, which includes dated contracts, was reported near $136bn, with 24-hour futures volume down 7% - turnover falling while open interest holds, which indicates traders avoiding fresh directional commitment.
Within that aggregate, the alt complex did the opposite. XRP shed 11.8% of open interest on a 6.9% price fall and Dogecoin shed 13.5% on an 8.1% fall. In coin terms bitcoin open interest fell from 331,100 BTC on 21 August to 318,600 BTC on 31 August, a 3.8% decline. So the honest answer is that the market is not uniformly adding leverage: large caps are holding elevated books, and the crowded alt trades are being unwound inside them.
Longs are paying, but barely. Bitcoin's annualised funding printed 4.95% with a z-score of +0.01 - its own average to two decimal places. Ether printed 5.36% at +0.38. On Hyperliquid, the large-cap cluster all printed the same 0.0013% per hour, which annualises to roughly 11.4%, and the open-interest-weighted average across the tracked HL cohort was 12.24%.
In other words: longs pay, everyone pays about the same, and nobody is paying an unusual amount. That has two consequences. For cash-and-carry, the spread over the risk-free rate at these levels is thin enough that execution and borrow cost dominate the trade. For directional holders, carrying a long is close to free, which is precisely why the open interest build in question 1 has been able to happen without resistance.
The exceptions sit in the tails. NEAR, at 24.31% annualised, was the only perp above $150m of open interest beyond two standard deviations. Cross-venue, SKR printed -4,079.6% annualised on Hyperliquid against -4,376.8% on Binance and +11.0% on Bybit. That is a spread of 4,387.7 percentage points on a book of about $5.1m - treat it as a warning about liquidity, not as an opportunity.
On-chain venues lost committed capital and gained fee revenue, which is the interesting combination. The perp DEX category cleared $533.63bn over 30 days while open interest fell -12.74% on the week to $22.44bn. Hyperliquid took $106m of fees in August, up 23% month on month, and now holds about 59.2% of category open interest on $13.28bn.
The competitive picture underneath that headline is a ranking by turnover quality. Hyperliquid turns its book over roughly 15 times a month. Aster, at about 20, and Variational, at about 19, are in the same band. Lighter, at 32, Grvt at 33 and Extended at 42 cycle more of their book; ApeX Protocol, at 291, and Jupiter, at 86, cycle far more. High turnover-to-open-interest ratios usually mean incentives are paying for volume rather than liquidity choosing to stay - and that is the number to watch when points programmes roll off.
The larger structural development was commercial rather than competitive. Bloomberg reported on 31 August that Hyperliquid Labs and Payward are in advanced talks to list selected Hyperliquid-linked perpetuals on Bitnomial, Payward's CFTC-licensed venue. If cleared, it would give an on-chain venue a regulated US front end without becoming a US exchange.
Short-term risk is skewed to the long side; medium-term risk already ran the other way. In the 24 hours to 31 August, $438m was liquidated, split $298m long against $140m short - roughly 68% on longs. Ether led at $138m and bitcoin followed at $95.6m; the largest single liquidation was $6.1292m on ETHUSDT at Aster.
Over the prior two weeks the split inverts: $9.7bn total, of which $6.55bn was shorts and $3.16bn longs. Shorts took about 67% of two-week damage. The market squeezed shorts through the back half of August and is now leaning on longs at a flat tape. Notably, no perpetual above $100m of open interest printed a liquidation z-score above 2 during the week - the leverage build did not produce forced exits, so this is pressure rather than stress.
The desk's concern is not the current level, it is the missing signal. With open interest at +2.5 z-scores and funding at zero z-scores, the usual warning - expensive carry on a crowded side - is absent. Nothing above $100m of open interest showed liquidation stress, and realised volatility collapsed to 27.46 for bitcoin from 60.99. A quiet tape with a large book is where cascades come from, not where they are advertised.
None of the above is a forecast. It is a list of the readings that would change the desk's interpretation, with the reason each one matters.
Both, in different places. Large caps are holding elevated books with six of the eight largest perps above a +2 open interest z-score, while the crowded alt trades are deleveraging - XRP shed 11.8% of open interest and Dogecoin 13.5%.
Longs are paying, but only slightly. Bitcoin annualised funding was 4.95% with a z-score of +0.01 and ether 5.36% at +0.38 - both essentially at their own averages. NEAR, at 24.31% annualised, was the only perp above $150m of open interest beyond two standard deviations.
Hyperliquid by committed capital, with $13.28bn of open interest or about 59.2% of the category. But turnover quality separates the field: Hyperliquid cycles its book about 15 times a month, while ApeX Protocol cycles its own 291 times, which points to incentive-driven volume rather than standing liquidity.
Short-term pressure is on longs, which took about 68% of the $438m liquidated in the 24 hours to 31 August. Over two weeks the split inverts, with shorts absorbing roughly 67% of $9.7bn. No perp above $100m of open interest printed a liquidation z-score above 2 during the week.
Whether open interest rebuilds while funding climbs. Individually neither number is alarming at present; together they would signal leveraged long exposure being rebuilt into a flat tape.