Hyperliquid Weekly News Review — Week of September 14–20, 2026

This review covers the Hyperliquid ecosystem and the HYPE token for the week of September 14–20, 2026, in sixteen stories. Every number below is pulled from a named primary source — the Hyperliquid API, DefiLlama, published release parameters — and each entry names where it can be re-verified. The week in one line: HYPE gained 21.3% to close at $93.75, erasing the prior week's entire 12.1% drawdown and setting a new all-time high at $94.60 — a move that arrived alongside two structural additions to the platform rather than a narrative alone.

1. The week in sixteen stories

Sixteen items, each with the verified number first and an editorial take second. The takes are interpretation only — no take introduces a new figure; every fact above it is re-verifiable at the named source.

1. HYPE closes the week at $93.75, +21.3% Price

HYPE opened the week at $77.26 and closed Sunday Sep 20 at $93.75 — a +21.3% week measured close-over-close, the convention this site uses everywhere. The five-week closing series now reads $82.37, $79.99, $87.94, $77.26, $93.75, and this week's close sits 6.6% above the previous record close of $87.94. The week's range was unusually wide in both directions: a $75.15 low printed Tuesday and a $94.60 high printed Saturday.

Our take: a single week that erases a prior −12.1% week and then clears the old record is a statement about who was selling and who was waiting. Last week's review called that drawdown a "local de-risking" on the evidence of flat open interest, neutral funding and range-bound deposits. This week validates the diagnosis rather than the forecast: the sellers were positioning, not exiting, and they were replaced at higher prices. The 6.6% overshoot above the prior record close is the part nobody could have sized in advance.

2. Two sessions did the work Price

Sep 17 closed $85.12 (+8.8%) and Sep 18 closed $92.61 (+8.8%) — two consecutive sessions that together account for essentially the entire week's gain, pushing the token from $78.23 on Wednesday night to $92.61 on Friday. Monday and Tuesday had gone the other way: Sep 14 closed $80.33 (+4.0%) and Sep 15 closed $76.91 (−4.3%) with the week's $75.15 low. Wednesday was flat-to-up at $78.23 (+1.7%), and the weekend was quiet — Sep 19 $92.22 (−0.4%), Sep 20 $93.75 (+1.7%).

Our take: when two adjacent sessions carry a +21% week, the honest question is whether the tape moved on news or on flow. Here the sequencing is legible: Wednesday Sep 16 was flattish into the Payward announcement, Thursday ripped +8.8%, and Friday ripped another +8.8% into the lending launch. That is news-then-flow ordering, which is the pattern that tends to hold rather than revert. It also means the week's $75.15 low is the number longs now have to defend — everything above it was built in 48 hours.

3. The record falls again: $94.60 Price

The Sep 19 candle printed a $94.60 high — a fresh all-time high, above the Sep 6 print of $89.60 that had stood for thirteen days. The Sep 18 session had already reached $93.17, and Sunday Sep 20 tapped $94.09 before closing $93.75. Measured from the prior record close ($87.94 on Sep 6), the new close is +6.6% higher; measured from the Sep 13 close of the drawdown week ($77.26), it is +21.3%.

Our take: the relevant fact about a record is not the level, it is the interval. Records set 13 days apart, twice, in a token whose float is roughly a fifth of max supply, tell you the marginal supply is thin and the marginal buyer is not. That cuts in both directions, which is why story 4 on this list matters as much as this one.

4. Six venues, five different record prints Method

The week's all-time-high number depends on who you ask. The Hyperliquid daily candle high printed $94.60 (Sep 19). CoinGecko's ticker record was $94.48, and Coinbase's was $94.46, both reported for Sep 19. Three news outlets the same day reported $93.04, and two others reported $92.56 and $90.92 for a print that had happened the previous session. All six describe the same event.

Our take: this is not a data error, it is a rounding of reality across venues — different order books, different last-trade stamps, different snapshot hours. The spread from $90.92 to $94.60 is 4.0%, which is larger than the move most traders would call significant. Our convention is to use the exchange's own candle series and to print the spread rather than pick the most flattering number; anyone comparing HYPE coverage across outlets this week should assume the level is accurate to roughly ±2% and the direction is not in dispute.

5. Sep 16: Payward plans permissioned markets for US clients Policy

On September 16, Payward — the parent company of Kraken — announced plans to build permissioned Hyperliquid HIP-3 markets for US clients, using the HIP-3 framework with Bitnomial handling regulated market operations. The division of labour as disclosed: Bitnomial Exchange establishes and manages the markets, Bitnomial Clearinghouse handles settlement, and NinjaTrader Clearing oversees customer accounts, with participation restricted to traders approved by both Bitnomial and NinjaTrader. The plan rests on permits Payward acquired with Bitnomial in May 2026 and remains subject to regulatory approval; no launch date, fee schedule, expected volume or initial contract list has been disclosed. Payward also framed the effort as creating new markets rather than opening existing ones to US traders.

Our take: read the disclosure language carefully, because it is unusually precise: new markets, permissioned access, named clearing entities, no timeline. That is the shape of a regulated venue being built alongside an unregulated one, not a bridge into it. The strategic significance is that a US-listed corporate parent is now willing to put its name on a Hyperliquid deployment — which converts a vague on-shore narrative into a counterparty with a compliance department. It also converts it into a filing process, and nothing about this week produced a filing.

6. The 500,000 HYPE stake behind that plan Tokenomics

HIP-3 requires any entity deploying a market to stake 500,000 HYPE and maintain that stake, revocable by validators if the market is poorly run. At the week's close of $93.75, that single requirement is worth roughly $46.9M; at the record print it was about $47.3M. That is a mechanical demand link between a corporate product decision and the token, and it lands before a single contract lists.

Our take: this is the most under-discussed mechanism of the week. Most "institutional adoption" stories are about flows arriving later; this one is a bond posted before launch. It also means the builder market's economics are gated by a real capital cost — which reframes story 14 below: if the entry ticket is ~$47M of staked HYPE and the median builder has earned a few hundred thousand dollars in fees, the entry price is the deterrent, not the competition.

7. Sep 18: native lending goes live on HyperCore Product

On September 18, manual borrowing and lending went live: users can supply HYPE or Bitcoin as collateral and borrow USDC or USDT directly on HyperCore, with interest rates floating on utilization. Hyperliquid's own announcement described the feature as sharing the same underlying infrastructure as the portfolio-margin system, and specified that each borrowed asset is sourced from a supplier rather than created through platform-level margin accounting. Co-founder Jeff Yan described the build as modular — a standalone lending protocol constructed on HyperCore first, then wired into the cross-margin system that already governs perps, spot and outcome markets.

Our take: the sequencing Yan describes is the substance here, not the marketing. Building lending as a separate module and then connecting it means lending risk stays isolable and the utilization curve becomes a live gauge of leverage appetite. It also answers a standing criticism of a perp-first venue — that stablecoin balances sit idle — and does it without forcing holders to sell collateral into the tape to raise cash. The same modularity is the risk: once a loan sits inside the same margin account as a perp position, stress in one propagates across the book, which is a design choice, not an oversight.

8. $269M borrowed on day one Product

Hyperliquid reported approximately $269M in assets borrowed within the first 24 hours of the lending launch, while supplied liquidity was reported above $400M. That demand arrived in a single session, on the same day the token closed +8.8% and printed the largest single-day fee run of the week ($3.96M, DefiLlama).

Our take: a $269M first day is a real number and a fragile one — first-day demand measures novelty and rate arbitrage as much as it measures need. The number to watch is not the day-one draw but whether borrowing persists when rates normalize and when the collateral asset is falling rather than rising. Lending against an asset that just rose 21% is the easy half of the product's lifecycle.

9. The lending parameters are deliberately conservative Product

The published parameters: HYPE collateral supports up to 65% loan-to-value with an 82.5% liquidation threshold; Bitcoin is capped at 50% LTV with a 75% threshold. The protocol retains 10% of interest paid as a liquidation reserve, with the remainder routed to suppliers, and rates float with utilization. For scale: $1,000 of HYPE collateral supports up to $650 of borrowing versus $500 against the same value of BTC.

Our take: a 65% LTV on a token with a fifth of its supply circulating is a more aggressive haircut than the venue's own BTC parameter, and it is worth saying plainly that this is a bet on Hyperliquid's own liquidity depth in its own token. The 82.5% threshold gives roughly a 27% buffer between an account's starting position and partial liquidation at max draw — wide enough to be defensible, narrow enough that a −25% day in HYPE is an event rather than a scenario. The 10% reserve is a sensible retention, but it is sized off interest, not off bad debt.

10. Protocol fees rose for the first time in five weeks Fees

Trailing-7-day protocol fees printed $19.25M, up 49.8% from $12.85M — ending a four-week slide that had taken the series from $24.6M to $21.2M to $16.5M to $12.9M (DefiLlama). The daily path inside the week: $2.92M, $3.03M, $2.83M, $2.50M, $3.96M, $1.90M, $2.13M, summing to $19.25M. Every single day cleared the prior week's $1.83M opening print except Saturday's $1.90M weekend trough.

Our take: this is the story that separates a sentiment rally from a usage rally. Price rose 21.3% and gross fees rose 49.8% — revenue grew faster than the token, which is the opposite of last week's pattern and the first time in a month that price and fundamentals moved in the same direction. A single week does not make a trend; it does falsify the "price caught up to a weakening business" reading that last week's review treated as the central fact.

11. Gross fees versus revenue: $429M and counting Fees

Press tallies this week put Hyperliquid at approximately $429M of 2026 year-to-date revenue, first among all crypto projects, with cumulative all-time revenue of $1.26B–$1.31B and roughly 48.7M HYPE removed from circulation — about 4.9% of supply. The same reporting cites weekly revenue near $13.5M and 30-day revenue near $64M, annualizing the recent weekly pace above $700M. Those figures sit on a narrower measure than the fee series printed above: DefiLlama separates gross fees from protocol revenue, and a $19.25M gross-fee week is not a $19.25M revenue week.

Our take: the two numbers being true at once is exactly the point. Gross fees include amounts shared with counterparties; protocol revenue is what the venue keeps. Any comparison between "Hyperliquid's revenue" and another protocol's revenue — or between revenue and a market cap — is meaningless until both sides use the same line. The milestone is real: leading all crypto projects on year-to-date income, nine months in, is not a rounding artifact of methodology. But the $429M figure should never be dropped into a multiple without naming which of the two lines it is.

12. NEAR routes perpetuals through Hyperliquid Integration

NEAR integrated perpetual futures trading through Hyperliquid, giving its users access to more than 50 markets with leverage up to 40x. The main-book census run for this review shows why the integration lands where it does: NEAR's own perp market carries $0.33B of open interest and $0.29B of 24-hour volume — the fifth-largest book on the venue by volume that session, between Solana ($0.28B) and XRP ($0.09B).

Our take: the interesting part is the direction of the pipe. This is not Hyperliquid borrowing a chain's users — it is a Layer 1 sending its order flow into a venue that already prices its token. That is the strongest version of the "Hyperliquid as derivatives layer" thesis, and it is measurable: the second-order effect shows up in NEAR's own OI and volume, not in a press release. Integrations of this kind compound quietly, which is why they belong in a weekly list even when they do not move the token.

13. HIP-3's share fell to 23.5% — because the core book grew Structure

The census run for this review measured platform open interest at $16.87B: $12.90B on the core order book across 178 live markets, plus $3.97B in HIP-3 builder markets. Builder markets are therefore 23.5% of platform OI, down from the 28.5% ($4.12B of $14.44B) measured in last week's census. The builder total fell 3.6%; the core book grew 25.0%. Live builder markets numbered 147 across five venues, with xyz running 108 of them.

Our take: a falling share that is caused by the denominator growing is not decay, and the distinction matters more this week than usual — a separate third-party census of HIP-3 published during the week reports the same effect, with the builder share of volume dropping from 57.1% to 25.8% over 30 days against a core book whose volume more than doubled. Anyone reading "HIP-3 is losing share" as builders failing has the causality backwards. The correct reading is that a rising market lifted the core book faster than it lifted the satellite venues, which is what you would expect when the token itself is the week's best-performing asset.

14. One builder takes 97.8%; the rest earned $747K Structure

A third-party builder census published during the week, covering complete UTC days from Aug 15 to Sep 13, found Trade[XYZ] at 97.8% of HIP-3 perpetual volume with $64.60B traded over 30 days — down 44.2% month-on-month — while the entire non-Trade[XYZ] builder segment had generated roughly $747K in lifetime fees against around $167M of collectively staked HYPE. The same study reports fees at the protocol-enforced floor (measured at 0.427bp for the leader and 0.400bp for its closest challenger), lists all six suspended builders as having used non-USDC settlement assets, and notes listing costs near $40K per asset.

Our take: this is the most useful piece of analysis published on Hyperliquid this week, and it points the opposite way from the price. A venue where the challengers collectively earn less than $1M in lifetime fees while bonding nine figures of the platform token has a competitive problem that a +21% week does not solve. Note also the settlement-asset correlation: every suspended builder used a non-USDC stable, every live one uses USDC. That is a single-variable story with six observations on one side, which is not proof — but it is the kind of base rate that should shape how the next cohort of builders is evaluated.

15. TVL: $7.10B platform-wide, and a parity claim to unpack Deposits

Hyperliquid L1 TVL closed the week at $6.50B, up 3.6% from $6.279B, with the Arbitrum bridge at $0.60B for a platform-wide total of $7.10B (DefiLlama) — within 0.1% of the $7.095B platform-wide figure independently reported for Sep 19. Separately, a widely shared chain-ranking claim this week put Hyperliquid's Layer 1 chain TVL at roughly $1.39B, level with Arbitrum and seventh among blockchains. That figure measures DeFi protocols deployed on the chain, a different construction from the protocol-level TVL printed above; both are DeFiLlama-tracked and they are tracked separately.

Our take: two TVL numbers about the same platform, differing by five-fold, both correct, is a perfect illustration of why every figure in this review carries its source and its definition. The chain-ranking milestone is genuine and worth reporting. So is the caveat: "Hyperliquid TVL" is not one number, and a headline that swaps between the two definitions between weeks is not tracking anything.

16. Two dates on the calendar Calendar

Two scheduled events frame the next two weeks. First, a core-contributor unlock of 14,175,778 HYPE — about 1.4% of max supply, worth roughly $1.33B at the week's close — is dated September 29. Second, the first AQAv2 distribution is dated October 3, the date on which the estimated reserve-yield contribution to the Assistance Fund becomes a measured payment rather than a projection. Neither event is news in the sense of something that happened; both are on the calendar and both were cited across the week's coverage as the next tests.

Our take: the market has now absorbed two unlocks — 14.18M HYPE on Aug 29 and 9.92M on Sep 6 — without either producing sustained downside, which is the base rate against which Sep 29 should be judged. But note the difference: the prior two arrived into a market at $79–88 with fees declining. This one arrives into a market at $93.75 with fees rising 50%. The composition of the buyer side has changed more than the size of the tranche has.

Three stories that mattered most

Sixteen items make a long list, so here is where we would spend attention if we could only follow three:

1. The two product additions (items 5–9). A regulated US market framework and a native credit layer landed four days apart. Neither has proved its economics; both widen what the platform can do, and one of them carries a ~$47M token bond as its entry ticket.
2. The fee reversal (items 10–11). Gross fees up 49.8% against a +21.3% token move is the first week in a month in which usage outran price. It is one week of evidence against a four-week pattern, which is enough to matter and not enough to conclude.
3. The builder census (items 13–14). The week's most important structural fact is not that HIP-3's share fell, but why — and what the challenger economics look like underneath it. That analysis points the opposite direction from the tape.

2. The week the product stack widened

Two of the week's stories are additions to what the platform does, rather than to how it is priced. Laying them side by side clarifies what each actually changes:

The asymmetry is the point: one of these three is a plan, two are running code with measurable output. The lending layer has a day-one number ($269M) and a daily observable (utilization-driven rates). The NEAR integration has a second-order print ($0.33B of OI in its own market). The Payward plan has a structure and no figure at all — which is not a criticism of the plan, only a statement that it cannot yet be measured.

3. What did not break

• Leverage got more expensive, not more crowded: HYPE funding printed 0.0000494/hour, roughly 0.039% per 8-hour interval — a material positive rate, up sharply from the ~0.010% neutral print in last week's review. Longs are now paying to hold.
• The HYPE book grew with the price: HYPE-perp open interest rose to $2.02B from $1.67B, a 21% increase that roughly matches the token's weekly move — the book scaled with the rally rather than levering into it.
• Deposits kept pace: L1 TVL rose 3.6% to $6.50B while BTC and ETH also rose, so the week's inflows were not HYPE-specific flight capital.
• Breadth held: 178 live core perp markets plus 147 live builder markets, with the core book's top books unchanged in composition — BTC $3.54B, ETH $3.03B, HYPE $2.02B of OI.

The week in seven sessions. The daily decomposition backs the week's central claim — that this was news-then-flow rather than a drift. Session by session, with the day's protocol fee print: Sep 14 $80.33 (+4.0%, $2.92M), Sep 15 $76.91 (−4.3%, $3.03M), Sep 16 $78.23 (+1.7%, $2.83M), Sep 17 $85.12 (+8.8%, $2.50M), Sep 18 $92.61 (+8.8%, $3.96M), Sep 19 $92.22 (−0.4%, $1.90M), Sep 20 $93.75 (+1.7%, $2.13M). The seven daily fees sum to $19.25M, matching the weekly print exactly. Note the shape: the two largest price sessions were not the two largest fee sessions — the fee peak came on the second of them ($3.96M on Sep 18), and Thursday's +8.8% printed the week's lowest weekday fee ($2.50M). Fees track volatility, not direction.

Cross-asset, same candles. On the same Sunday-close clock, BTC rose +5.7% ($76,810 → $81,221) and ETH rose +6.9% ($2,476 → $2,646). HYPE's +21.3% was therefore 3.7x BTC's move and 3.1x ETH's — a large divergence, and the mirror image of last week, when HYPE fell three times as fast as BTC. BTC itself remains roughly 36% below its October 2025 record, which is the context for reading a token-level record as a token-level event.

Measured at the start of next week (Sep 21, ~03:15 UTC). Two structural measurements were taken after the Sunday close and are dated as such. First, the platform derivatives stock: $16.87B of open interest — $12.90B core plus $3.97B across builder markets — with 24-hour platform volume at $6.76B. Second, the builder census: 10 registered builder DEXs, five with live markets, xyz running 108 live markets and roughly 98% of builder open interest, and the top builder books led by SP500 ($429.4M), SKHX ($340.0M), GOLD ($298.0M) and CL ($202.3M). These are post-week archive measurements, not week-of prints, and they are labeled accordingly.

4. Sources & method

Prices, candles, open interest, volume and funding: Hyperliquid official API (api.hyperliquid.xyz). Builder-market census: the same API's perpDexs and metaAndAssetCtxs endpoints, summed across all registered dexes. Fees and TVL: DefiLlama. Supply, market cap and 24-hour token volume: CoinGecko and published venue tallies. Lending parameters (loan-to-value, liquidation thresholds, reserve retention) are taken from the release parameters as published by the venue and corroborated across same-day coverage. HIP-3 builder economics are attributed to the third-party builder census published during the week and labeled as such; the 500,000 HYPE deployment stake is from the HIP-3 specification. Raw API pulls for this week are archived in this site's build directory. Where a figure cannot be independently verified, it is not printed.

Frequently asked questions

Related reading

• Hyperliquid Weekly Data Report — Week of September 14–20, 2026
• Weekly Intelligence — Week of September 14–20, 2026
• HYPE Market Intelligence — Week of September 14–20, 2026
• The Collateral Turn: What Hyperliquid's Lending Layer Changes