The market de-risked into the Federal Reserve rather than positioned for it. The five questions below are the same ones every week; the answers are what changed.
For a hike, not against one - and by leaving, not hedging. Bitcoin closed the week at $77,238.98, -3.4%, inside a $76,393-$80,494 range it never exited cleanly in either direction. Binance BTC futures OI fell $529m over seven days to about $8.0bn. Spot bitcoin ETFs, the unlevered marginal buyer of the past month, posted four straight days of outflows totalling about $462.7m - the largest weekly withdrawal in ten weeks. Long-term-holder SOPR crossed below 1.0 for the first time in the recovery, ending the week at 0.9945 (BGeometrics). Total crypto market cap: $2.67 trillion, bitcoin dominance 60%.
The rest of the complex tracked the same path at different betas. Solana closed at $101 (-5% on the week), XRP at $1.36 (-4.3%), BNB at $721 (-4%) and ZEC at $1,111 (-5.5%). The cleaner cross-asset tell sat outside crypto: gold's -1.72% week means the same rates repricing that knocked crypto also lifted the classical hedge - the signature of a macro-driven drawdown, not a crypto-native one. On-chain positioning indicators agree with the de-risking read: NUPL at 0.3123 and MVRV at 1.454 place the market mid-recovery, neither at euphoria where leveraged longs cluster nor at capitulation where forced selling has already cleared.
Almost nothing, which is the answer. Perp funding printed +0.005% per 8 hours on Binance and OKX and +0.01% on Bybit - at or just below the 0.01% neutral anchor, roughly 5.47% annualised. The mechanics matter: Binance anchors at the neutral rate while its premium index sits inside the clamp, so the print describes a perpetual trading at a slight discount to spot. This is a flat basis, not leveraged long demand (Velo via Zerocap). The week's one deviation was directional information: ether funding flipped to -0.0056% per 8 hours on the 10 September flush, and the largest bitcoin long on Hyperliquid - $233m - went from profit to a $-3.39m unrealised loss the same day.
Priced at the actual print rather than a round number, the carry on a $100,000 leveraged long is the cheapest the recovery has offered. The table below compounds the week's Binance/OKX print into holding-period costs; it is the arithmetic behind the 'de-risked, not positioned' reading, because a book paying this little to hold longs had no funding-driven reason to unwind - the de-leveraging was discretionary.
| Holding period | Funding cost | Share of notional |
|---|---|---|
| 1 day | $15.00 | 0.01% |
| 1 week | $105.00 | 0.10% |
| 1 month | $450.00 | 0.45% |
| 1 quarter | $1,368.75 | 1.37% |
The week's most consequential flow deserves its own table. Spot bitcoin ETFs printed four consecutive outflow days from 8 to 11 September, cumulating toward the $462.7m week - the largest weekly withdrawal in ten weeks - while ether ETFs took $216m on the 11 September session alone, the rotation trade InteractiveCrypto flagged running at full speed. The unlevered marginal buyer stepping back is what made both liquidation windows possible: with spot demand absent, the derivatives book set the price, and the derivatives book had already de-risked.
| Date | Net flow | Note |
|---|---|---|
| 8 September 2026 | -589.32 BTC | - |
| 9 September 2026 | -1,532.66 BTC | - |
| 10 September 2026 | -3,612.12 BTC | ARKB -2,099.30 BTC, the largest single-fund redemption of the week |
| 11 September 2026 | -172.47 BTC | IBIT -250.86 BTC |
| Four-day total | -5,906.57 BTC | about $462.7m for the week in dollar terms |
Hyperliquid remains the centre of the on-chain book, but the week's honest answer is that its OI depends on who is counting. On the venue's own API (10 September): $10.60bn across 234 markets, $6.85bn traded in 24 hours. On the aggregator cohort (CoinGlass coverage via AInvest): $14.67bn, a 76% share of the $19.20bn tracked perp DEX cohort, ahead of Aster at $2.50bn and Lighter at $1.10bn. Both prints are published because they measure different things, and neither is carried as the single truth. The fee ledger is less ambiguous: $13.5m in trailing-7-day fees and $10.42m in protocol revenue (DefiLlama via AInvest), with HLP at $185.1m TVL and a +7.1% APR.
| Venue / basis | Open interest | Note |
|---|---|---|
| Hyperliquid (DeFiLlama basis) | $14.67bn | 76% of the $19.20bn tracked cohort |
| Aster | $2.50bn | second-largest perp DEX by open interest |
| Lighter | $1.10bn | third-largest perp DEX by open interest |
| Hyperliquid (exchange API basis) | $10.60bn | the venue's own measure, 10 September - a different denominator, printed side by side |
Both sides, in sequence. The 10 September PPI session flushed $562m, 86% of it longs, as hike odds jumped from 64% to 70%. The 11 September CPI session ran the film in reverse: price slipped toward the weekly low, then reversed violently - $250m+ of shorts liquidated within about an hour, over half in ether, about $660m for the day across 100,000 traders. The options market had priced the event rather than the surprise: DVOL near near 37%, the post-FOMC 18 September expiry at about 35% ATM, and the 1-week 25-delta risk reversal moving from +1.14 vol to -0.37 over the week - puts marginally richer, not panic.
| Window | Total | Composition | Detail | Source |
|---|---|---|---|---|
| 24h around 10 September (PPI flush) | $562.0m | $484m longs (86%) | bitcoin-only reading: $85.64m liquidated, 92% longs | CoinGlass via AInvest; PAPER |
| 11 September (CPI whipsaw) | $660.0m | 100,000 traders | $250m+ shorts in about an hour, over half ETH | CoinGlass via CryptoPotato |
Leverage is priced against a collateral base, and that base kept growing through the de-risking week. The Big-3 stablecoins (USDT, USDC, DAI) rose +$742.7m over the seven days to 9 September, from $261.79bn to $262.53bn (DeFiLlama via NeoDrop) - accelerating from the prior week's expansion. USDC added +$661.6m and USDT +$80.6m; the sharpest single-jurisdiction move was TRON, where USDT supply grew +$504.2m to $92.27bn - settlement float that ends up posted against transfers, not trades. Hyperliquid's own USDT margin float stood at $85.4m. For scale: offshore perpetual venues cleared an estimated $$90 trillion trillion of volume in 2025 against $$28 trillion trillion two years earlier, and that float is the depth every liquidation cascade in this issue propagated through.
Neither, in size. Funding at or below the neutral anchor, falling open interest, put-skewed options and ETF outflows describe a book that reduced exposure into the event rather than expressing a direction. Zerocap's derivatives desk called it precisely: de-risked into the meeting rather than positioned for it.
Because they measure different things. The venue's own API reports $10.60bn on 10 September; the aggregator cohort reports $14.67bn, a 76% share of $19.20bn tracked across perp DEXes. Where published figures differ, both are shown with the cohort named - a difference in definition is not a data error.
Headline 0.4% month on month and 3.4% year on year, both in line. Core ran 0.3% on the month against a 0.2% forecast - the hawkish surprise - while core year-on-year eased to 2.4%, the lowest since March 2021. The market traded the core monthly print and the yields that repriced after it, not the in-line headline.
The desk prints it and declines to rank it. The 50-day EMA crossed above the 200-day on 11 September, the first such cross since mid-2024 (InteractiveCrypto) - and price was rejected at $80,000 within hours. A signal that resolves nothing inside a session is context, not a trade.
The ETF ledger. Whether spot bitcoin ETF flows resume accumulation after the $462.7m weekly drawdown determines whether any post-FOMC move is absorbed or amplified - Zerocap's desk flagged the same variable, and it is the one number in this issue that the FOMC decision itself cannot print.
Offshore perpetual venues cleared an estimated $$90 trillion trillion of volume in 2025, against $$28 trillion trillion in 2023 - roughly a threefold expansion across the cycle. That is the market whose funding, open interest and liquidation mechanics this column tracks weekly, and why the collateral float behind it (the Big-3 base at $262.53bn) is printed alongside the derivatives data rather than treated as a separate beat.
TRON: USDT supply there grew +$504.2m to $92.27bn in the week to 9 September - the largest single-jurisdiction change in the Big-3 ledger. It matters indirectly: TRON float is settlement and transfer collateral rather than margin, so it deepens the payment rail the trading float ultimately settles against, without itself entering a liquidation cascade. The margin-relevant floats are Hyperliquid's $85.4m USDT base and the exchange-held balances behind the CEX book.