HYPE Market Intelligence — Week of September 14–20, 2026
The week in one line: HYPE closed $93.75, +21.3%, set a fourth all-time high at $94.60, and — for the first time in a month — grew protocol fees faster than it grew its own market cap, compressing the price-to-fee multiple from 26.4x to 20.8x. This report works through the price ladder, the valuation arithmetic, the buyback ledger, relative strength on identical clocks, and the four regime changers published here last week, graded after the fact.
1. The price ladder
Four records in four weeks, with the gap between each successive record widening:
| Record | Date | Intraday high | Close that day | Interval since prior record |
|---|---|---|---|---|
| 1st of the cycle | Aug 23 | $83.46 | $82.37 | First close above the May high of $67 |
| 2nd | Aug 27 | $86.73 | — | 4 days |
| 3rd | Sep 6 | $89.60 | $87.94 | 10 days |
| 4th | Sep 19 | $94.60 | $92.22 | 13 days |
The path between the third and fourth records is what makes this week interesting. The Sep 6 record was followed by a close at $87.94, then a −12.1% week that closed at $77.26 on Sep 13 — a −13.8% retreat from the $89.60 intraday print. The $76.59 level acted as support into the next week, where it was briefly broken to a $75.15 intraday low on Sep 15 before the token reversed. Thirteen days after the Sep 6 record, HYPE printed $94.60 and closed $93.75, 6.6% above the prior record close and 21.3% above the Sep 13 close. Monday Sep 21 has so far held $93.09, 1.6% below the record.
One measurement discipline note, because it matters for anyone comparing coverage: the record print is venue-specific. Hyperliquid's own candle series shows $94.60; CoinGecko $94.48; Coinbase $94.46; and same-day reporting also carried $93.04, $92.56 and $90.92 for prints made at other hours. The spread is 4.0% — larger than many moves traders would call significant — so HYPE's all-time high should be quoted with its venue or as a range.
2. Valuation arithmetic on verified inputs
From verified inputs only. Market cap at the weekly close ($93.75 × 222.446M circulating) is $20.85B. The trailing-7-day gross fee run-rate is $1.00B annualized ($19.25M × 52), and the 30-day base annualizes to $0.91B ($75.8M × 52 ÷ 4.33). On fully diluted supply of 955.3M the same market value is $89.5B.
| Inputs | Value | Multiple on circulating | Multiple on FDV |
|---|---|---|---|
| Market cap / 7d fee run-rate ($1.00B) | $20.85B | ~20.8x | ~89x |
| Market cap / 30d fee base ($0.91B) | $20.85B | ~22.9x | ~98x |
| Market cap / W4 run-rate ($0.67B, prior week) | $20.85B | ~31.2x | ~134x |
| Market cap / W1 run-rate ($1.28B, late-August peak) | $20.85B | ~16.3x | ~70x |
The bottom two rows exist to make one point: the same $20.85B market cap produces multiples between 16.3x and 31.1x depending on which seven days you annualize. Any single multiple quoted without its fee window is unfalsifiable. Two rows, however, are directly comparable week over week, and this is the week's genuinely unusual fact: a week ago, market cap was $17.7B against a $0.67B run-rate — 26.4x. This week market cap rose 17.8% while the run-rate rose 49.3%, so the comparable multiple fell to 20.8x even as the token made a new record. Price and the multiple moved in opposite directions, which is the arithmetic form of "the business grew faster than the token this week."
These are restatements, not targets. The multiple is set by the market; the inputs are set by the APIs; the FDV rows assume all 955.3M outstanding tokens trade at the same price, which is not how unlocks work.
3. The buyback ledger, stated carefully
What is verifiable: gross protocol fees of $19.25M last week and $75.8M over 30 days, and the documented mechanism that virtually all protocol revenue routes to the Assistance Fund for HYPE buyback-burn, supplemented by the USDC aligned-quote-asset reserve-yield share (estimated ~$160M/yr; under the AQA standard Circle and Coinbase each stake 500K HYPE). What is not verifiable weekly: the AF's actual USD purchases — the ledger is on-chain but has no clean API, so this report derives capacity rather than claiming executed amounts.
| Line | Value | Basis |
|---|---|---|
| Gross fees, 7d | $19.25M | DefiLlama dailyFees, summed |
| Implied weekly buyback capacity | ~$19.1M | ~99% pass-through (third-party estimate) × fees |
| Implied capacity at prior week's fees | ~$12.7M | Same construction, $12.85M week |
| Cumulative HYPE burned | ~48.7M (~4.9% of max supply) | Third-party tally; no single public ledger API |
| 2026 year-to-date revenue | ~$429M, first among crypto projects | Third-party tally; narrower measure than gross fees |
| Reserve-yield contribution | ~$160M/yr estimated | Modeled; first payment scheduled Oct 3 |
| Weekly executed USD purchases | Unavailable | No public API series |
Read the first and second rows together: capacity rose about 50% week over week, tracking fees almost one-for-one. That is the mechanical consequence of near-total pass-through — there is very little buffer between a fee print and a bid. October 3 is the date that converts the largest modeled line in this table into a measured one, and it is the single most informative scheduled event on the calendar after the September 29 unlock.
4. Relative strength, same-source
| Week | HYPE | BTC | ETH |
|---|---|---|---|
| Aug 17–23 | +44.3% | +23.7% | +31.3% |
| Aug 24–30 | −2.9% | −0.2% | −1.9% |
| Aug 31–Sep 6 | +9.9% | +3.4% | +4.0% |
| Sep 7–13 | −12.1% | −4.4% | −1.6% |
| Sep 14–20 | +21.3% | +5.7% | +6.9% |
| Five-week cumulative (Aug 16 → Sep 20) | +64.2% | +29.2% | +41.0% |
The same API, the same clocks — every cell is Sunday close over Sunday close, so the comparison is clean. Two readings matter. First, the amplitude: over five weeks HYPE compounded 64.2% against BTC's 29.2% and ETH's 41.0%, which means the token has run about 2.2x the market and 1.6x ETH — and it has done so while both drawdown weeks (−12.1%) and advance weeks (+21.3%) were roughly three times the majors' moves in the same direction. This is a high-beta instrument that has, so far, been compensated for the beta. Second, the level: BTC remains about 36% below its October 2025 record while HYPE prints records, so this is not a rising-tide week being read as alpha. The five-week excess is token-specific, and so is the risk that comes with it.
5. What would change the regime
Four regime changers were published here last week. Grading them is the most useful thing this section can do:
| Regime changer (set Sep 14) | What happened | Status |
|---|---|---|
| "A fee print above $16.5M would show the contraction was volatility-driven, not demand-driven" | $19.25M printed, 16.7% above the line | Triggered — the contraction was volatility-driven |
| "A close below $76.59 hands the trend to the bears" | Lowest close was $76.91 (Sep 15); $75.15 traded intraday but never closed below | Did not trigger — and the distinction between an intraday break and a close matters |
| "An OI break below ~$9.5B would confirm the stock is following the flow down" | Platform OI rose to $16.87B; core book +25.0% | Did not trigger — the opposite happened |
| "Hard AF/USDC-yield data would convert the largest estimate into a line item" | No disclosure; Oct 3 remains the scheduled first payment | Unresolved |
For next week, the changers:
- A fee print that holds above $16.5M — the same line, now tested in the other direction. A second consecutive print above it makes the structural case arguable; a slide back under $13M makes this week look like a volatility spike.
- The September 29 unlock (14,175,778 HYPE, ~$1.33B at the week's close) — the first tranche to arrive into rising fees rather than falling ones. The relevant reference is now $75.15 (the week's low), not the $76.59 line that failed.
- Funding above 0.05%/8h while price stalls — funding at 0.039% is already meaningfully positive. Crowding shows up here before it shows up in the tape; this is the earliest warning available and it is free to watch.
- A disclosed AQAv2 payment on October 3 — the only scheduled event that transfers a line from "Modeled" to "Confirmed" in this report's ledger.
- Lending utilization decay — a day-one draw of $269M that decays toward a fraction of itself says the launch captured arbitrage; a draw that holds says it captured need.
No price targets here — the levels above are descriptions of the map, not predictions of the route.
6. The week, day by day
| Date | Close | Intraday range | Δ | Fees (day) |
|---|---|---|---|---|
| Mon Sep 14 | $80.33 | $76.91 – $82.48 | +4.0% | $2.92M |
| Tue Sep 15 | $76.91 | $75.15 – $80.56 | −4.3% | $3.03M |
| Wed Sep 16 | $78.23 | $76.62 – $80.36 | +1.7% | $2.83M |
| Thu Sep 17 | $85.12 | $78.06 – $85.25 | +8.8% | $2.50M |
| Fri Sep 18 | $92.61 | $85.07 – $93.17 | +8.8% | $3.96M |
| Sat Sep 19 | $92.22 | $91.01 – $94.60 | −0.4% | $1.90M |
| Sun Sep 20 | $93.75 | $89.60 – $94.09 | +1.7% | $2.13M |
The intraday column is where the week's structure lives. Tuesday's $75.15 low and Friday's $93.17 high are 23.9% apart in a single week, and the record ($94.60) printed on Saturday — a thin-book session that also produced the week's lowest fee print ($1.90M). That is worth flagging before anyone treats the record as a liquidity-confirmed level: the highest print of the week came from the quietest session of the week. Wednesday's range ($76.62 – $80.36) and Thursday's ($78.06 – $85.25) show the transition: Thursday opened near Wednesday's low and closed 8.8% higher, which is the shape of a re-pricing rather than a drift.
The fee line tells its own version of the same story. Summed, the seven daily prints total ~$19.25M (each shown rounded to $0.01M; the unrounded DefiLlama sum is $19.25M). Friday's $3.96M was the week's peak and the largest single-day print in three weeks (since Aug 25's $4.30M); Thursday's equal-sized price move produced the week's lowest weekday fee. Fees track how much volume a move generates, and a session that gaps and holds generates less volume than one that travels the same distance in both directions. That is the mechanism behind last week's fee decline and this week's rebound, and it is why the fee series should be read as a volatility proxy at the weekly horizon.
Frequently asked questions
How can the price-to-fee multiple fall while the token makes a record?
Because the denominator grew faster. Market cap rose 17.8% week over week ($17.7B → $20.85B) while the annualized 7-day fee run-rate rose 49.3% ($0.67B → $1.00B). The multiple is a ratio, and this week the business grew about 2.8x faster than the token. Whether that persists depends entirely on whether the fee print holds.
Which all-time-high figure should I use?
The one belonging to the venue you are quoting. Hyperliquid's own candles printed $94.60; CoinGecko $94.48; Coinbase $94.46; and same-day coverage also cited $93.04, $92.56 and $90.92 for prints at other hours. The 4.0% spread is real venue dispersion, not an error. This report uses the exchange's own series and prints the spread.
Why was the $76.59 line published if a break below it did not matter?
Because a level that is never checked is decoration. Price traded to $75.15 intraday on Sep 15 — 1.9% below the line — and then rose 21.3% into a new record. The lesson is not that levels are useless but that an intraday break and a close below a level are different events, and that support levels describe where sellers previously appeared rather than switching a market into a new regime. The next reference level is $75.15, and it is already labeled as the week's low rather than as an invalidation point.
Does the buyback capacity figure mean $19.1M was actually spent?
No. It is capacity: gross fees times the estimated ~99% pass-through rate. The AF's executed purchases live on-chain with no clean public API, so this report computes what the mechanism could absorb and labels it as capacity. When a disclosure endpoint exists — the Oct 3 AQAv2 payment is the closest scheduled thing — this table switches from capacity to actuals.
Is this financial advice?
No. This is data analysis for informational purposes. Nothing here is a recommendation to buy, sell or hold HYPE or any other asset.