Funding Mechanics
How each venue's risk engine bounds the funding rate — and why "capped" almost never means what it says.
Mapping the mechanics across the ten major perpetual venues reveals three distinct architectural approaches to bounding the funding rate: Margin-Anchored Caps, Frequency Compression and Decentralised Elasticity. The distinction matters — it determines whether the ceiling holds under stress or whether it silently relocates into the time domain.
1. Margin-Anchored Engines
Bybit · Binance · BingX · Bitget · Blofin
These centralised venues tie the maximum allowable funding tax directly to the contract's risk parameters rather than a static percentage.
- Clamp formula. All these engines suppress minor deviations with an identical base clamp that pulls the rate back toward the base interest rate when the spot/perp premium tightens.
- Dynamic limits. The absolute ceiling is a function of the Initial Margin Rate (IMR) and Maintenance Margin Rate (MMR). Binance and Bybit both bound the rate by
min((IMR − MMR) × 0.75, MMR). - Volatility overrides. During extreme structural imbalances, Bybit removes the 0.75 coefficient (adjusting up to 1.0). Binance overrides the margin-based limits entirely, hard-capping at fixed tiers like 2% or 3%. Both systems force settlement frequency down from 8h to 1h to accelerate the margin drain on offside flow. Bitget, BingX and Blofin largely mirror this architecture.
2. Frequency Compression Engines
OKX · KuCoin · Gate
Instead of expanding the mathematical ceiling per print, these exchanges use strict per-interval limits but compress the settlement window to achieve geometric expansion.
- Static thresholds. The cap per settlement tick is hard-coded by asset tier (e.g. 0.75% max per interval).
- Interval escalation. When the rate sustains its absolute maximum, OKX and KuCoin algorithmically downshift the funding cycle: 8h → 4h → 2h → 1h.
- Net effect. The absolute rate looks capped per print, but the daily extraction rate multiplies up to 8×. Shorts bleed out through frequency rather than magnitude. Gate natively runs on an 8-hour cycle with fixed limits, creating slight arbitrage discrepancies against its frequency-escalating peers.
3. Decentralised Elasticity
Hyperliquid
The on-chain clearinghouses operate with massive, aggressive upper bounds — explicitly designed to punish structural dislocations harder than centralised venues.
- Hyperliquid. Settles continuously using an hourly average. It uses the standard clamp formula but caps the rate at 4% per hour — a 96% daily maximum. The most frictionless environment for harvesting trapped flow.
Mechanical Comparison Matrix
| Exchange | Standard Interval | Volatility Interval | Cap Architecture | Max Limit Driver |
|---|---|---|---|---|
| Hyperliquid | Continuous (1h) | Continuous (1h) | Fixed high-band | 4% per hour |
| Binance | 8-hour | 4h / 1h | Margin / fixed tier | ±0.75 × MMR or static % |
| Bybit | 8-hour | 1-hour | Dynamic margin | min((IMR − MMR) × 0.75, MMR) |
| OKX | 8-hour | 4h / 2h / 1h | Tiered fixed cap | Interval compression |
| KuCoin | 8-hour | 4h / 1h | Tiered fixed cap | Interval compression |
| Gate | 8-hour | 8-hour | Tiered fixed cap | Static asset tier |
| Bitget | 8-hour | 1-hour | Dynamic margin | Clamp limit escalation |
| BingX | 8-hour | 1-hour | Margin function | ±0.75 × MMR |
| Blofin | 8-hour | 8-hour | Margin function | Base clamp limits |
Deep dive — Bybit vs OKX at the ceiling
Neither venue is truly "free-floating." What looks uncapped is automated interval compression taking over when the premium snaps.
Bybit — dynamic margin peg
Bybit anchors its absolute cap to the contract's risk parameters rather than a static tier: the upper and lower limits are computed as min((IMR − MMR) × 0.75, MMR). Under extreme divergence Bybit lifts the 0.75 coefficient to 1.0 and, if the rate continuously hits the limit at settlement, mechanically switches the settlement frequency to 1 hour to reflect real-time order-book pressure.
OKX — tiered cap & frequency escalation
OKX defines hard-coded absolute caps per settlement interval by asset tier — e.g. ±0.375% for majors like BTC, ±0.75% for lower-tier alts. What makes OKX feel free-floating is explicit frequency escalation: when the rate hits the tier cap, OKX does not expand the percentage cap — it compresses the settlement window from 8h → 4h → 2h → 1h. The per-print cap is static; the daily aggregated limit multiplies up to 8×.
Shared core
Despite the distinct boundary logic, both engines use the same clamp buffer to pull the base rate back toward the base interest rate whenever the spot/perp premium tightens. An absolute per-print limit exists on both venues — interval compression is what unleashes the geometric bleed on offside shorts and creates the illusion of an uncapped, free-floating rate.