Perpetual funding rates: a week of divergent signals^*

01 / THE TAKEAWAY

One market, very different funding readings

The same underlying asset can carry materially different funding rates across perpetual-market feeds. At the end of this window, ETH funding was negative on Hyperliquid (−21.8% APR), Lighter (−7.0%) and Nado (−6.5%), but positive on GRVT (+3.2%), Variational (+6.5%) and Pacifica (+11.0%).

Those endpoint readings also differ from the seven-day averages: all six selected ETH markets had positive average funding in the supplied report. A single snapshot and a weekly average answer different questions.

02 / OPEN INTEREST

BTC and ETH contracted while HYPE expanded

On the Hyperliquid feed, BTC open interest fell from $3,581.64M to $3,365.29M, and ETH fell from $2,619.29M to $2,360.05M. Over the same endpoint comparison, HYPE rose from $1,620.89M to $2,039.88M — approximately 26% — while ZEC rose approximately 21%.

HYPE’s window-end funding was +35.6% APR, versus a seven-day average of +13.7%. XMR, outside the chart, combined approximately 37% OI growth with average funding of +39.9% APR. These are positive funding premiums alongside higher open interest, not proof of a squeeze, directional positions or persistent crowding.

03 / BEYOND CRYPTO

Large changes in equity and commodity perps

The report contains striking endpoint changes outside crypto. GOOGL on hyperliquid-xyz grew from $97.71M to $282.46M OI, approximately 189%. On Aster, XAU grew from $5.93M to $88.30M — about 14.9 times its starting value — and CL grew from $1.90M to $24.50M.

Large percentage increases from small starting bases deserve context. These figures do not establish listing dates, the cause of growth or order-book depth. For Aster CL, −53.6% annualized funding at the window end means shorts would pay longs under the usual convention; it does not mean shorts are being paid.

04 / DATA QUALITY

Extreme APRs are a starting point for investigation

UNITREE on hyperliquid-para had a reported seven-day average of −389.3% APR and an endpoint reading of −1,406.9%. PRL on hyperliquid-io averaged +146.0%. These are annualized observations, not evidence that such rates were achievable continuously or can be captured profitably.

Several rows contain identical minimum and maximum readings, while other feeds reach repeated extremes. Flat readings can reflect genuine stable funding, caps, stale data or collection behavior. The summary alone cannot determine which explanation applies. Those cases need timestamp, coverage and venue-methodology checks before being used for a trading decision.

A funding spread is not automatically an arbitrage opportunity. Fees, slippage, borrowing costs, collateral requirements, liquidation risk and mismatched contracts can outweigh the apparent spread.

05 / METHODOLOGY & LIMITATIONS

What this study measures — and what it does not

• The supplied report covers 23 October 2026, 05:45 UTC to 30 October 2026, 05:45 UTC, and describes roughly five-minute snapshots from Tensorsoft’s data pipeline. This article uses the supplied summaries, not independently checked raw snapshots.
• There are 21 feeds and 2,731 feed-market listings. Multiple hyperliquid-* endpoints belong to the same underlying venue; these are not 21 independent exchanges or necessarily 2,731 unique instruments. We do not aggregate them into unique-venue market share.
• Detailed data covers up to ten markets per feed by endpoint OI (193 rows supplied), not every listed market. Cross-platform charts use selected exact-symbol matches; contract specifications may differ.
• Funding APR is the supplied annualized percentage. The report notes a ×8,760 annualization convention; that assumes an hourly rate. Interval normalization, weighting of weekly averages, missing samples and coverage cannot be checked from this file.
• Minimum and maximum APR values are extremes, not durations. OI charts compare two endpoints; no intraday or daily path is reconstructed.
• OI and 24-hour volume are in USD millions. Missing volume, including GMX and Nado entries, remains unavailable — not zero. Rounded zero values may represent small balances.

Research for informational purposes only. Not investment advice or a recommendation to trade on any platform.

06 / REPORTED FIGURES

Feed overview & market appendix

The supplied figures below preserve separate feed labels. Funding columns are annualized percentages; OI and volume are USD millions.

Phoenix, lighter-robinhood, polymarket-perps and aftermath links include Tensorsoft referral codes; we may receive compensation. Other links use the supplied platform addresses. Builder feeds without a supplied front end remain unlinked.

ΔOI percentages are the rounded changes reported in the study, not recalculated from rounded OI figures.