Bitcoin +22% weekly: how much was squeeze and how much was new money

Bitcoin closed Friday, August 21, 2026, near $78,335, up 22% from the start of that week—the largest weekly gain since 2024. Within those five days, there were two short squeezes—forced liquidations where the exchange's risk engine forcibly repurchases on behalf of those betting on a decline: one of approximately $2.77 billion on Wednesday and another of about $1.06 billion on Friday. The second was roughly one-third the size of the first, yet the price rose more in the final stretch than in the initial one. This asymmetry allows us to separate the two factors that blend in any violent rebound: the purchases no one chose to make and the money that decided to enter. Spot exchange-traded funds added $1.92 billion across those five sessions, and Wednesday's session alone removed about 7,400 Bitcoin from available supply—more than sixteen times the network's daily issuance.

Why did Bitcoin rise 22% during the week of August 17-21, 2026?

The 22% surge is supported by three distinct events occurring within 72 hours: a U.S. Treasury announcement, two waves of forced liquidations, and five consecutive sessions of inflows into spot exchange-traded funds. The timeline, ordered by date and without attributing causality where there is only succession:

The institutional package during those days was not unanimously favorable: the SEC delayed the innovation exemption for tokenized securities, and the CLARITY Act remains stuck in Congress. None of these acts force anyone to buy Bitcoin. An exchange's risk engine does, however, and that part leaves a record.

How much of the August 2026 Bitcoin rally was a short squeeze and how much was ETF flow?

The two liquidation cascades of the week of August 17-21, 2026, are measured in 24-hour windows ending at the same time each day, making them comparable; net flows for spot funds are published per session. With the two series side-by-side, the split between forced buying and discretionary buying changes clearly within 48 hours:

The forced component on Friday, August 21, 2026, was roughly 38% of what occurred on Wednesday the 19th: $1.06 billion in shorts compared to $2.77 billion. Discretionary demand held steady: five consecutive sessions of net inflows, $1.92 billion into Bitcoin funds and $697.47 million into Ether funds—the best combined week since October. And the price, which had gained 8.3% on the day of the large cascade, added another 12.6% from that close until Friday's close.

In the Bitcoin rally of August 17-21, 2026, the liquidation cascade ignited the move and the exchange-traded fund flows sustained it: the forced component dropped from $2.77 billion to $1.06 billion between Wednesday and Friday while the price rose another 12.6%. A purely mechanical rebound gives back the gains once the inventory of vulnerable shorts is exhausted, because it leaves no new buyers behind; here, that inventory was partially exhausted on Wednesday, the second wave was much smaller, and the price continued to rise with fund inflows as the only persistent buying. August 2026 has accumulated about $2.38 billion in Bitcoin funds, a monthly high for the year surpassing the April record ($1.97 billion), despite four sessions of net outflows between August 10 and 14 that subtracted $394.6 million: the five sessions of the week of the 17th reversed that exit and added $1.92 billion.

The final stretch includes $1.06 billion of forced buying: the second cascade shrank but did not disappear. And three sessions are three sessions. For the balance to remain persistent, the created shares must stay created, which will be determined by the redemption series in the following weeks.

How does a Treasury debt buyback reach the price of Bitcoin?

A Treasury buyback is debt management, not monetary policy: the issuer buys its own bonds already in circulation on the secondary market, without an interest rate decision, without the creation of bank reserves, and without Federal Reserve intervention. The August 19, 2026, statement increased the maximum size of these operations from $2 billion to at least $4 billion in the 10-20 and 20-30 year sectors, effective from September 9 to November 4.

The transmission from that operation to the price of a token has four links, and none are automatic:

1. Duration. A recurring buyer in the long and illiquid part of the curve reduces the interest rate risk the market must absorb, compressing the term premium at the margin.
2. Dollar. Lower long-term yields, with short-term rates remaining where they were, narrow the yield differential that makes holding dollars attractive compared to other currencies.
3. Discount Rate. Discounting at a lower rate raises the present value of any asset whose appeal lies in distant expectations and makes financing risk positions cheaper.
4. Beta. The returning appetite for risk hits assets higher on the risk scale with more amplitude—this amplification is beta—and the crypto market occupies that position.

Doubling an operation from $2 billion to $4 billion is small compared to the size of the U.S. debt market, so magnitude alone does not explain the move; what traders read was the signal of an issuer willing to put a floor under long-term liquidity. The legal nature of each move that week in Washington, who signed them, and their binding force is detailed in the five levers of Washington. For what follows, it is enough to remember that the fuse existed and its effect arrives through pricing, not an obligation to buy. It is the logic of fiscal dominance we explored in the Bitcoin cycle under fiscal dominance, with debt management conditioning financial conditions.

What is a forced liquidation and why does it trigger cascades?

Anyone opening a leveraged short in a perpetual—a future without an expiration date—deposits collateral and commits to repurchasing the contract. If the price rises enough that the collateral no longer covers the loss, the exchange's risk engine closes the position for them and buys the contract at the prevailing price. This purchase has a property that distinguishes it from any other: no one chose it. Its origin is accounting-based; a seller ran out of collateral.

Many short positions share liquidation price ranges because popular leverage clusters at round multiples and the same technical levels. Forced purchases push the price toward the next group, which also liquidates, and the cascade feeds back on itself until the inventory of vulnerable shorts is exhausted. The full mechanism is in our guide on how liquidations work.

The figures from the two cascades clearly illustrate this pattern:

• First cascade, 24h until August 20 at 7:30 UTC: ~$3.02 billion liquidated, of which ~$2.77 billion were shorts and ~$257 million were longs. Nearly eleven dollars on the bearish side for every dollar on the bullish side, a distribution indicating leverage concentrated on a single side.
• ~$1.31 billion in a single hour, 43% of everything liquidated in the 24h until August 20, 2026; of that, 1.23 billion were shorts. More than four out of every ten dollars that day came from sixty minutes.
• ~$1.10 billion corresponded to Bitcoin perpetuals (K33 Research, a Norwegian crypto-asset market analysis firm), the first time Bitcoin perpetual shorts exceeded $1 billion in a day. The rest were forcibly closed in Ether and the rest of the market.
• Second cascade, 24h until the morning of Friday, August 21, 2026: ~$1.24 billion total, with $1.06 billion in shorts and $174.41 million in longs, spread across 141,191 accounts. On the short side, Bitcoin contributed $789.68 million, Ether $206.88 million, and XRP $41.94 million; the rest was distributed among the altcoin market.

CoinGlass records begin in 2021, so the first cascade is the absolute all-time high for its category in the entire series. The scale is set by the other extreme: the largest total deleveraging recorded remains October 10, 2025, at about $19 billion, but that was on the long side.

Can derivative liquidations be compared to spot ETF flows?

Derivative liquidations and spot exchange-traded fund flows are not directly comparable because they measure different things in different units. The $2.77 billion in shorts liquidated on August 19, 2026, and the $517.19 million in net inflows that same day illustrate the trap.

The $2.77 billion is derivative notional. It is the value of the forcibly closed exposure, not the actual money at stake: a short opened with 10x leverage on $100,000 of collateral represents $1 million in notional, and upon liquidation, the record notes the million while the destroyed capital is the $100,000. Furthermore, a perpetual liquidates against the contract itself: the forced purchase occurs in derivatives and reaches the spot market indirectly via market makers hedging their books and arbitrage against the index.

The $517.19 million is spot cash. It is money that entered fund shares and obligates the authorized participant—the entity that creates and redeems those shares—to acquire real Bitcoin to back them: at $69,564 per unit, about 7,400 Bitcoin removed from available supply in one session. The network produces 450 new Bitcoin per day—3.125 per block and about 144 blocks daily—so one day absorbed more than sixteen times the daily issuance. This disproportion between regulated vehicle demand and new supply is what we analyze in the supply deficit between ETFs, mining, and treasuries.

Each magnitude is compared against its own base. Liquidations are weighed against open interest and the day's derivative volume, indicating what proportion of existing leverage was wiped out. Fund flow is weighed against spot volume and issuance, indicating what proportion of available supply changed hands persistently. The 3.92x ratio on Wednesday serves to compare it with the 2.15x on Friday, not to claim that mechanical factors weighed four times more in price points.

The concentration in a single fund is a data point in itself: on Thursday the 20th, IBIT contributed 502.99 of the ~$606 million for the day, or 83%. A flow that depends on a single vehicle can reverse through the same door it entered, as shown by the record redemptions of June 2026. The series also has a publication lag: it comes out after the close, so it serves to verify the next day's movement and never to trade on the same day.

Why did Ether and HYPE rise more than double Bitcoin?

The distribution of the August 19, 2026, movement between Bitcoin, Ether, and HYPE is the most direct evidence of where positioning was, because short leverage leaves its mark exactly where it accumulated.

The two multiples are almost identical and have different origins. In Ether, the explanation is positioning: a good portion of the liquidated shorts were outside Bitcoin perpetuals, and Ether is the deepest non-Bitcoin derivatives market, so it absorbed the large portion of that cascade. Lower capitalization with an equally active derivatives book converts the same forced buying pressure into more percentage points.

HYPE adds a specific catalyst to that amplification, dated to the August 19 White House meeting, detailed in the piece on Washington levers; the token was trading around $69.78 after the move, in line with what Hyperliquid was already building with pre-IPO price discovery. For the decomposition in this article, HYPE is the case where short positioning and specific news combined, which is why its movement does not serve to measure the mechanical component.

The higher on the risk scale, the more short leverage had accumulated and the more violent the reversal was: the convexity structure we described in the crypto market fragility pyramid, working this time in a bullish direction.

How to tell if a crypto rally is a short squeeze or real buying?

The reading method that separates forced buying from flow applies to any double-digit movement within hours, and the order matters because each series answers a question that the following ones can no longer address:

1. The long/short split of liquidations in the first hour. An imbalance greater than 4 to 1 toward one side is the signature of directional deleveraging; a split near 50/50 describes volatility without positioning bias and suggests another explanation.
2. Hourly concentration. When a single hour contains one-third or more of the daily total—40.7% on Wednesday—the movement contained a cascade within it. If distributed flatly throughout the day, there was sustained buying or selling, which is different.
3. Liquidations against previous open interest—the total of live contracts just before the move. This avoids double counting: the dollar notional only says something compared to the leverage that existed before. Without that denominator, a large figure in a large market distinguishes nothing.
4. The size of the second cascade relative to the first. Measure them in 24-hour windows ending at the same time of day to ensure they are comparable. If the second wave drops to less than half of the first—38% between August 19 and 21, 2026—and the price continues to rise in that interval—another 12.6%—the buying sustaining the stretch is no longer forced. If it equals or exceeds the first, the stretch remains mechanical.
5. The funding rate reset. After a short sweep, it moves from negative or neutral to positive; if it stays very positive for several days, the marginal position is now long and leveraged, and fragility has shifted sides. How this rate works is in our explanation of funding rates in perpetual DEXs.
6. The next day's ETF flow, not the same day's. The flow for the day of the movement is published when everything has already happened. The next day's flow separates confirmation from reversal: it is money that decided to enter knowing the price had already risen.

None of the six steps is a buy or sell signal. They are descriptive criteria: they serve to classify the movement before deciding what to do with it.

Which series should be followed in the coming weeks?

The three forces of this week fade over different timeframes, and each has its own tracking series. The accelerant has already been measured: two cascades, the second at one-third of the first. Flow is checked session by session, and what determines if the balance is persistent is not the inflows but the absence of redemptions in the following weeks: August 2026 has not yet recorded a single session of net outflows, and that is the data point that could break. The fiscal fuse has its own date: expanded buybacks begin on September 9, so the August episode moved on expectation, with the operation not yet executed. The full institutional calendar for the coming weeks—Jackson Hole, the Warsh speech, and the September FOMC meeting—is in the piece on Washington levers.

On the side that could break the floor, the most cited risk points to the structure of the sector itself: if corporate digital asset treasuries were to sell Bitcoin in size to manage their balance sheets, that sale would appear first in spot flow and not in liquidations, because it is not a forced closure of a leveraged position but a balance sheet decision. Distinguishing both things using the series in this article is exactly the exercise, and the financial constraints of those treasuries are in the doctrine of never selling.

At the close of this extraction, Saturday, August 22, Bitcoin is trading above $78,000. Two useful things remain from the week beyond the price: a documented episode of how three forces of different natures combine and at what speed each fades, and a repeatable check—the size of the second cascade versus the first—for the next double-digit candle.

Sources and links: The Block — $517M in Bitcoin ETFs · KuCoin — $606M on August 20 · Cointelegraph — largest Ether ETF inflow since October · crypto.news — $2.6B in the week · CoinDesk — second $1B wave in shorts · BeInCrypto — second cascade distribution · Bloomberg — the rebound before the White House meeting · CNBC — the +22% week · U.S. Treasury — expansion of buybacks