Hyperliquid Weekly News Review — Week of August 31–September 6, 2026

Revised Sep 14, 2026: every figure below has been re-verified against primary sources (Hyperliquid API, DefiLlama); estimated figures were replaced with measured values or explicitly marked unavailable.

This review covers the Hyperliquid ecosystem and the HYPE token for the week of August 31–September 6, 2026, in sixteen stories. Every number below is pulled from a named primary source — the Hyperliquid API, DefiLlama — or attributed to a dated news report. The week in one line: HYPE rode to its $89.60 all-time high and closed $87.94, +9.9%, absorbing a second unlock in nine days — while the protocol's fee line fell for a third straight week.

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. Monday strength, mid-week surge Price

HYPE opened the week at $79.99 and closed Monday Aug 31 +5.2% at $84.17 (Hyperliquid candles). Tuesday and Wednesday drifted lower — $82.74, $81.90 — and then Thursday Sep 3 jumped +6.8% to $87.49 with an $88.20 intraday high. Friday gave back −3.5% to $84.45, Saturday held $85.43, and the Sunday candle delivered the week's record: $89.60 high, $87.94 close.

Our take: opening a week +5.2% — two days after absorbing the largest unlock in the protocol's history — is the follow-through the unlock bears were missing. Sideways sessions between strength are not dead tape; they are the market deciding whether to keep the level. Thursday's answer came with a surge, which makes the week's shape: strength, digestion, strength.

2. The $88 print Price

Sep 3–4 candles peaked at $88.20 and $87.88; news reported the "$88 all-time high" on Sep 4 with +5.2% in 24 hours and +51% on 30 days (CoinPost). The candle record is more complicated than the headline: the Sep 3 session made the $88.20 high, and the Sep 4 session that carried the "$88" coverage actually closed −3.5% at $84.45. Both facts are true; they measure different things.

Our take: the candles and the news cycle disagreed by cents and a day — as they usually do. We anchor to the candle series ($88.20) and print the "$88" as the social fact it also was: round numbers become narrative anchors the moment they print, and narrative anchors attract orders. That feedback loop is not nothing; it is just not data.

3. Sunday: the record, properly Price

The Sep 6 candle printed a $89.60 high and closed $87.94, +2.9% on the day — a fresh all-time high on more than $1B of reported session volume (ETHNews). It was the third record print in ten sessions ($86.73, $88.20, $89.60) and it landed on the same day as the second unlock (story 5), which makes the Sunday candle the month's cleanest supply-meets-demand observation.

Our take: a record set on a Sunday, on real volume, is worth pausing on. Weekend records happen in 24/7 markets, but they usually happen on thin books; this one printed nine figures. The close below the high ($87.94 vs $89.60) is also the honest part of the story — the record was an intraday event, and the following week's review is largely about the gap between those two numbers.

4. Price discovery Price

With every prior high cleared — $83.46, $86.73 and now $89.60 inside three weeks — $90 and then $100 stood as the round-number anchors traders cited (ETHNews). The candle data adds the caution: the $89.60 print came with an $87.94 close, so the market touched the anchor without holding it. The following week's review starts exactly there.

Our take: above every prior high there is no resistance history — only round numbers, which are psychology, not supply. The honest description of price discovery is that nobody knows where the sellers are; that is what the phrase means. What we can say from the data: the fee curve below (story 13) did not confirm the price discovery, and divergences like that one are how moves end before anyone has a narrative for why.

5. Unlock #2 absorbed Supply

Sep 6: 9.92M HYPE (~$820M) released — the second large unlock in nine days, after Aug 29's 14,175,778 (~$1.2B). The Sunday candle ran to the record anyway: $85.43 open, $89.60 high, $87.94 close (MoneyCheck, CoinMarketCap; Hyperliquid candles). Combined, the two releases put roughly $2.0B of scheduled supply into the market inside nine days, and both weeks closed green.

Our take: two absorptions in nine days, roughly $2B of scheduled supply combined, both weeks closing green. The May release that preceded −14.1% is now the outlier in the sample rather than the rule. Base rates updated, with the caveat that both absorptions happened in strong tape — the real test of unlock resilience is what happens when supply meets a falling market, and that test has not run yet.

6. The on-shore report got concrete Policy

Sep 4: reporting (The Block, via CoinMarketCap) described a framework involving Kraken's parent company and the regulated platform Bitnomial to offer select crypto perps to U.S. users — no license granted, no timeline stated. The report landed the day after HYPE's $88.20 session and in the same session the token closed −3.5% at $84.45; the tape's non-reaction to a US-access headline is itself worth logging.

Our take: named institutions move a story from rumor to architecture, but a framework without a license or a date is still a sketch. We track it as a standing item and will say nothing further until something is filed — which is a discipline, not indifference. If US-compliant perps access materializes, it lands on a venue that already trades ~$14.4B of open interest; the strategic stakes are the reason the sketch is worth logging at all.

7. Multicoin moved $21.7M to Coinbase Prime On-chain

Sep 2: 261,555 HYPE transferred to Coinbase Prime over 12 hours — ~$21.7M at the prevailing price, a neutral-to-bearish supply signal (CoinMarketCap, on-chain). The transfer landed between Monday's +5.2% session and Thursday's $88.20 run — a quiet window on the tape, which is exactly why it is measurable: there was no noise to hide in.

Our take: a transfer to prime custody is neither a sale nor a buy — it is positioning for optionality, and it deserves neutral language. The reason it makes the list is methodological: large on-chain prints are among the few supply observations that are measured rather than modeled. We will report it again if the coins move to an exchange book; until then, "watched" is the correct status.

8. HyENA shut down Structure

The perps platform (announced Aug 29) wound down Aug 31–Sep 2 after the USDC integration made its USDe-margin model obsolete; it had processed $4B+ in cumulative volume (CoinMarketCap). This site's API census confirms zero active HyENA markets as of Sep 14 — the census, not the press cycle, is what establishes the exit as complete.

Our take: the first full deployer exit matters precisely because it is unremarkable. The standing design question for HIP-3 builder markets was whether departure would be possible without breaking the venue; it was, and the platform kept running. There is also a method here worth keeping: the API census (zero active markets) is what turned a press narrative into a measured fact — this is exactly the kind of claim a data site should verify rather than relay.

9. The burn ledger crossed 48M Tokenomics

Cumulative HYPE burns reached 48.45M — 4.84% of the 1B maximum supply (MoneyCheck). The Assistance Fund has spent ~$1.1B buying back 45M+ HYPE since launch (ETHNews), and the two series are two views of the same mechanism: the dollars in, and the tokens permanently removed. Nothing here moves a weekly tape — that is what makes it a baseline.

Our take: the burn is the sink side of the flywheel this site tracks weekly, and its virtues are the boring ones: slow, mechanical and permanent. Nothing about 48.45M moves a week's tape — that is the point. On a 1B maximum, a sink that only grows is the one line in the tokenomics that has never had a down week, which makes it the right baseline against which to read the noisier stories.

10. 2026 buyback pace Tokenomics

$379M of HYPE repurchased year-to-date per CoinPost's Sep 4 tally; a sampled 24-hour window saw 9,730 HYPE bought at an $85.27 average (~$830K). The sample annualizes near $300M — the same neighborhood as the YTD tally — so the flow checks against itself from two independent constructions. That agreement is why both numbers are printable.

Our take: the sample window is the useful part of this story. Roughly $830K in a day annualizes near $300M — in the same neighborhood as the $379M YTD tally, which is what a cross-check is for: two independent constructions of the same flow landing close enough to trust both. The buyback narrative has many numbers attached to it; these two at least agree with each other.

11. AQAv2 got its first third-party size Tokenomics

An external estimate built on ~$6.21B of USDC in the AQA v2 wallet and ~3.10% indicative yield: with ~90% routed back to the protocol, roughly $527K/day, ~$192.5M/yr — explicitly an estimate (McKenna via ETHNews). The first official payment is scheduled for Oct 3, the date the model becomes checkable; every input above is published, so the check will be arithmetic, not opinion.

Our take: $527K/day is a model, not a ledger — but it is now a specific model with stated inputs, which means it can be falsified on Oct 3 when the first payment lands. That is what makes it useful: "somewhere between zero and meaningful" is unfalsifiable and therefore worthless; a number with published arithmetic is accountable. We will hold both the figure and its author to the Oct 3 print.

12. Institutional holdings on the record Flows

13F filings showed thirty firms holding close to $75M in Hyperliquid ETF exposure (ETHNews). The print is quarterly, lagging, and small against a ~$19.6B market cap at the week's close — and it is the measured version of institutional demand, as opposed to the reported version (the prior week's staking item) or the modeled one. The firm count, thirty, is the number to track across future filings.

Our take: $75M across thirty firms is small against a $17.7B market cap — and the size is not the story. Filings are the measured, quarterly, lagging record of institutional entry; this is the early print of a flow that will either build across future 13Fs or not. The comparison to watch is not the dollar figure but the firm count, which doubled-digit growth would make a trend rather than an event.

13. The fee divergence widened Fees

Protocol fees printed $16.5M for the week, −22.2% — the third consecutive decline (DefiLlama). The daily path shows the fade inside the record week: $2.49M–2.92M through mid-week, a $3.47M peak on the $88.20 Thursday session, then $1.09M and $1.65M into the record weekend. Price made all-time highs while the fee curve declined for a third straight week — the divergence is a daily fact, not just a weekly one.

Our take: records in price and a −22.2% fee week is the cleanest expression of the month's central theme: the token and the fee curve are different series, and they have been diverging for three weeks. One can lead the other for a while. This site's job when they disagree is to say which one is telling the truth — and on the evidence of stories 9 through 11, the sink was filling while the flow cooled, which is a more balanced divergence than the price chart alone suggests.

14. Open interest context Derivatives

Platform-wide open interest stood near $14.3–14.4B (main book plus builder dexes) — roughly 3% below the $14.7B peak set just before the Oct 10, 2025 correction, which erased 56% of OI in 24 hours (MoneyCheck; this site's API archive). This site's Sep 14 census splits the current number: $10.32B on the core DEX, $4.12B across HIP-3 builder markets. The Oct 2025 book's split was never published, so the comparison is a level comparison, not a composition one.

Our take: the historical comparison is context, not comfort — that peak preceded a one-day wipeout, and being 3% below it is not a safety margin so much as a reminder that OI levels carry tail risk. What is different from last October is composition: 28.5% of the current book sits in builder markets with different margin structures (see our HIP-3 deep dive). Same headline number, different machine underneath it.

15. HYPE outran the market again Cross-asset

Same week, same daily candles: BTC +3.4% ($77,659 → $80,314) and ETH +4.0% ($2,418 → $2,515), against HYPE's +9.9% ($79.99 → $87.94). The week's biggest single-day gap came Thursday: HYPE +6.8% against BTC's +5.1% and ETH's +4.8% — the $88.20 day. Modest weekly outperformance on a much larger base move reads as consolidation of gains, not fresh aggression.

Our take: a third outperformance week in four — but modest this time, roughly 2.5x rather than the launch week's 1.9x on a much larger base move. Modest outperformance into records reads as consolidation of gains rather than fresh aggression, and that read was testable: the following week returned HYPE to triple-beta on the downside. Outperformance in both directions is the same mechanism.

16. TVL held the plateau Deposits

L1 TVL closed at $6.41B, −0.3% week-over-week — range-bound through the record week (DefiLlama). After +10.2% and +4.4% in the prior two weeks, a plateau is digestion: deposits absorbed the unlock, the record and the fee decline without moving, and the $6.4B level became the line the following week would test.

Our take: flat deposits during a record price week is not a contradiction — perps activity and deposit decisions are separate choices, made by separate actors on separate clocks. After +10.2% and +4.4% in the prior two weeks, a plateau is digestion. The question this story set up was whether the $6.4B band would survive the first real risk-off week; the Sep 7–13 review answers it — the band held.

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 second absorption (item 5). Two unlock tests passed in nine days converts "absorption" from anecdote to pattern — with the honest caveat that both tests ran in strong tape. The falling-market unlock test remains unrun.
2. HyENA's exit (item 8). The first full deployer departure, verified to zero active markets by API census, answers the builder-market exit question affirmatively: the venue permits departure and survives it.
3. The fee divergence (items 13, 9, 11). Price at records while fees fell 22% is only alarming if you ignore stories 9 and 11: the burn ledger grew and the buyback engine gained a sized, falsifiable estimate. The divergence was real; the cushion was too.

2. The divergence is the story

Price made a new all-time high. Fees fell for a third straight week — $21.2M → $16.5M, and the weekly chain now reads $24.6M → $21.2M → $16.5M. Both series are real; the question is which one is telling you about the future.

The fee line tracks realized trading activity — and it is normalizing off a squeeze spike. The price line is discounting structural demand: two unlocks absorbed in nine days, a documented buyback (48.45M HYPE burned to date, ~$1.1B spent by the Assistance Fund), a second revenue stream (AQAv2) activating Oct 3, and a reported U.S. on-shore path. Whether that discounting is early or wrong is what the next fee prints will start to answer.

3. What did not break

• Supply absorption, twice: 14.18M HYPE (Aug 29) and 9.92M HYPE (Sep 6) — ~$2B of scheduled supply in nine days — and the second unlock week closed +9.9% at the all-time high.
• TVL: $6.41B, within a dollar-billion of its monthly high — deposits did not fade during a record week (DefiLlama).
• Market structure: platform-wide open interest holding ~$14.4B — within ~3% of the pre-October-2025 peak — says positioning rebuilt, not unwound (MoneyCheck; API archive).

4. Sources & method

Prices and candles: Hyperliquid official API. Fees and TVL: DefiLlama. Burn and buyback figures: MoneyCheck and CoinPost tallies (third-party, on-chain-derived), ETHNews Assistance Fund coverage. AQAv2 estimate: McKenna via ETHNews — labeled modeled, not measured. Institutional figures: ETHNews (13F), CoinMarketCap. 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 August 31–September 6, 2026
• Weekly Intelligence — Week of August 31–September 6, 2026
• HYPE Market Intelligence — Week of August 31–September 6, 2026
• The Pullback: Anatomy of HYPE's First Decline From the Record