Two numbers carried this week's perp market. Open interest expanded while price fell - $479.65bn globally, up 16.14% in a day to 16 September - and funding stayed flat at +0.0044% per 8 hours. A market that adds leverage into a drawdown and still rallies +5.6% is a market where the marginal buyer is not leveraged.
Bitcoin perp funding opened the week at +0.0044% per 8 hours and finished it at +0.0056%, against a benchmark anchor of 0.01%. Annualised on the 8-hour clock, +0.0044% reads as about 4.82% - a number close enough to the neutral carry that a delta-neutral position earns almost nothing and a directional long pays almost nothing.
Ether funding is the more interesting series this week. It sat at +0.004% per 8 hours on 16 September and had climbed to +0.063% by 17 September, as buying pressure extended through the Fed decision. Even at the higher print it stays far from the levels that signal an over-leveraged long book - it is a reading worth monitoring, not a warning.
The annualised figure matters more than the raw 8-hour print. A +0.0044% funding rate on an 8-hour clock compounds to about 4.82% a year. That is close enough to the risk-free dollar rate that a delta-neutral long-spot short-perp position earns no meaningful excess after borrow costs, and a directional long pays almost nothing for the leverage. When funding sits at this level through a +5.6% rally, the market is telling you the marginal buyer is paying cash, not borrowing it.
| Horizon | Cost | As % of notional |
|---|---|---|
| 1 day | $13.20 | 0.01% |
| 1 week | $92.40 | 0.09% |
| 1 month | $396.00 | 0.40% |
| 1 quarter | $1,204.50 | 1.20% |
The compounding assumption is worth making explicit. The 4.82% annualised figure is the result of compounding the observed 8-hour rate three times a day across 365 days. It ignores the fact that funding rates move with price, that margin calls alter position sizes, and that exchanges sometimes skip or modify settlement prints. Even so, it is a useful benchmark: anything within a few hundred basis points of the risk-free rate is effectively a free carry, which is what this week delivered.
The ETH-BTC funding divergence also tells a cross-asset story. Bitcoin funding held near its anchor while ether funding rose to +0.063%. That divergence widened through the Fed decision and stayed wide into the weekend. In a generic risk-on rally the two would move together; the divergence suggests that ether had a separate demand driver, most likely the ETF-flow and re-staking narratives that re-emerged after the rate decision.
The weekly open-interest story is the inversion of the usual one. Global derivatives open interest stood at $479.65bn on 16 September, up 16.14% over 24 hours, with bitcoin derivatives near $51.96bn. Price, meanwhile, printed $74,944. Rising open interest into falling price is the structure that produces cascades - and it produced one the same session, $224.81m of bitcoin derivatives liquidated, 84.6% of it longs.
On-chain, DeFiLlama's perps cohort ran between $25.80bn and $26.90bn a day and $147bn to $152bn across the week. Hyperliquid's own open interest lifted to $8.10bn from $7.90bn seven days earlier, on $237.00bn of 30-day volume.
The on-chain share is smaller than the CEX book but growing faster. Global derivatives open interest reached $479.65bn on 16 September, with bitcoin derivatives near $51.96bn; the DeFiLlama perps cohort is a low-single-digit fraction of that total. What matters is the direction: Hyperliquid's OI rose while price fell, which means the positions were added into weakness rather than carried through a rally. That is the same structure that produced the $224.81m CLARITY-day liquidation and the subsequent squeeze.
CEX and on-chain OI moved in the same direction but not at the same speed. The global derivatives number rose 16.14% in 24 hours, while Hyperliquid's own OI moved from $7.90bn to $8.10bn. The CEX book is still an order of magnitude larger, but the on-chain book is the one that added into weakness first. That sequence is consistent with the idea that sophisticated traders used the on-chain venue to express directional views while CEX order books were still absorbing spot selling.
Volume-versus-OI is another quality check. Hyperliquid printed $237.00bn of 30-day volume alongside the OI build. High volume with rising OI means positions are turning over; low volume with rising OI means positions are being held. This week was closer to the latter, which points to directional holders rather than day-traders. Directional holders are the cohort most exposed to a failed breakout retest.
Hyperliquid's token was the week's standout instrument. HYPE closed at $92.31, +17.6% on the week from $78.48, and printed a record $94.46. The catalyst was structural: Payward's plan to deploy HIP-3 markets for US clients obliges a 500,000 HYPE stake that must be maintained - $46m at the market price. With only 26% of the 951.6m maximum supply in circulation (251.5m tokens), demand of that size moves the price far more than it would for a large-cap asset.
The float arithmetic is what amplified the move. Only 26% of the eventual 951.6m supply is in circulation - 251.5m tokens. A 500,000 HYPE lock-up is therefore a larger share of the free float than the same dollar stake would be for bitcoin or ether. Add to that the protocol's burn dynamics - 48.76m HYPE burned ($4.50bn) lifetime and 26,310 HYPE burned in the last 24 hours alone - and the supply picture looks tighter than a simple market-cap comparison suggests. HYPE closed at $92.31, +17.6% on the week.
| Metric | Reading | Context |
|---|---|---|
| HYPE, weekly close | $92.31 | from $78.48 a week earlier |
| HYPE, all-time high | $94.46 | 19 September 2026, intraweek |
| Hyperliquid open interest | $8.10bn | from $7.90bn seven days earlier |
| 30-day perpetual volume | $237.00bn | the platform's trailing month |
| Circulating supply | 251.5m of 951.6m max | only 26% of maximum supply is tradable |
| Lifetime buyback-and-burn | 48.76m HYPE | 5.12% of supply, worth $4.50bn |
The validator-revocation clause is the under-discussed part of the HIP-3 design. A deployer must maintain the stake; if validators vote to revoke it, the market is removed. That creates a governance check on top of the regulatory one. For Payward, it means the stake is not a one-time marketing expense but a continuing liability. For the protocol, it means bad behaviour can be punished without waiting for a court. The clause matters because it converts the stake from a sunk cost into an ongoing bond.
The supply dynamics extend beyond the locked stake. 48.76m HYPE has been burned lifetime, equivalent to $4.50bn, and 26,310 HYPE was burned in the last 24 hours alone. Burn reduces the effective circulating supply over time, which amplifies the impact of any large lock-up. The combination of a small float, a large mandated buyer and ongoing burn is why the token moved before any fee revenue was visible.
The structural backdrop for the venue comparison is unchanged and unflattering to onshore wrappers: offshore perpetual volume estimated at $90 trillion in 2025, up from $28 trillion in 2023, is the stock of activity every regulated structure competes against. The HIP-3 plan is the first concrete attempt to move a slice of that activity onshore without giving up the on-chain engine.
The offshore baseline is the key reference. Any regulated wrapper that captures even a small fraction of $90 trillion would be material to Hyperliquid's economics and to US market structure. The HIP-3 plan is the first attempt to build that bridge without forcing users off the on-chain engine, which is why it is being treated as a structural rather than a product-level event. The entities involved - Bitnomial, Payward's CFTC-regulated exchange and NinjaTrader Clearing - also matter: this is not a sandbox experiment but a CFTC-regulated exchange and clearinghouse taking real balance-sheet risk.
The Bitnomial precedent is the closest historical comparison. That structure also kept execution on-chain and wrapped a regulated entity around clearing and custody. The lesson from the Bitnomial case is that the regulatory wrapper moves more slowly than the market wants, and that legal challenges from incumbents can outlast the initial approval. Payward's plan differs in scale: 6.6 million funded accounts and a 500,000 HYPE stake are an order of magnitude larger than prior attempts. If the wrapper succeeds, it becomes the template; if it stalls, it becomes another cautionary tale.
The immediate question for the desk is whether the HYPE re-rating has already priced in most of the stake demand. A 500,000 HYPE obligation is a known, one-time buyer, and the token moved before any volume or fee data existed. The next leg of the trade requires either additional deployers to announce similar plans, or Payward's volume to exceed expectations. Without one of those, the structural case is stronger than the price case.
+0.0044% per 8 hours on 16 September, about 4.82% annualised, against a 0.01% neutral anchor. It finished the week at +0.0056%. Funding stayed flat while price rose +5.6%, which points to spot-led demand rather than leveraged longs.
It expanded into weakness. Global derivatives open interest reached $479.65bn on 16 September, up 16.14% over 24 hours, with bitcoin derivatives near $51.96bn, while price printed $74,944. That structure produced the session's cascade.
Payward's confirmed plan to deploy HIP-3 perpetual markets for US clients obliges a 500,000 HYPE stake, about $46m at the market price, and only 26% of maximum supply is in circulation. HYPE closed at $92.31, +17.6% on the week, and touched $94.46.
DeFiLlama's perps cohort ran between $25.80bn and $26.90bn a day, totalling $147bn to $152bn across the seven days.
The week of September 7 - 13, 2026 report remains published in the Perp Market Intelligence archive. Prior issues are frozen as published and never updated against later data.