94% of Bitcoin's $1.28 Billion ETF Exit Came From One Fund — in the Same Week the AI Trade Crashed

The headline treats eleven tracked funds as one seller. The underlying data shows one dominant seller and a short tail of noise. IBIT shed $1.21 billion of net assets in the week — more than seven times the $165.3 million net outflow of the nine remaining funds combined. Fidelity's FBTC was the only material buyer, taking in $93.1 million of net creations, while Bitwise's BITB, Invesco's BTCO and CoinShares's BRRR together added $23.9 million.

IBIT's own quarterly filing explains why fund-level concentration is not a footnote. The iShares Bitcoin Trust holds its bitcoin with Coinbase Custody Trust Company, names Anchorage Digital Bank as an additional custodian, and lets authorized participants redeem in cash. For those cash creations and redemptions, the trust converts cash into bitcoin and back again, either with third-party trading counterparties or routed through Coinbase as prime execution agent. Fidelity's creations ran the other way. Published fund comparisons note that FBTC custodies its bitcoin in house at Fidelity Digital Assets, so $93.1 million of buying never had to appear in a public spot order book.

The standard explanation for crypto ETF redemptions is that the money went into AI stocks. The same five sessions rule that out. On Nov. 5 the Nasdaq Composite fell 5.52% to 24,315.00 and the S&P 500 fell 4.36% to 7,353.73, on investor doubts that the AI data-center buildout would slow after the midterm election. A dollar cannot rotate from crypto into AI equity in a week when AI equity itself loses 5.5%. The destination was cash, and bitcoin held above its 200-day average while the index broke.

Bitcoin has been widely described as trading below its 200-day moving average. The levels published on Nov. 4 do not support that. The 200-day average sat at $73,188, with spot at $75,880 on Nov. 7 — 3.5% above it — while the 30-day average at $81,319 had flipped to overhead resistance. That distinction decides the playbook. Above $73,188 the week's redemptions read as a rotation inside an intact uptrend; below it they become confirmation of a trend break, and the next map stops near $62,000.

The week's redemption sounds structural because the year-to-date number turned negative: 2026 net flows swung from +$368 million to -$914 million. The balance sheet says something different. IBIT's net assets fell from $67.03 billion at Sept. 30 to $60.3 billion on Nov. 6, and only $1.21 billion of that $6.73 billion decline came from investors leaving. The remaining $5.52 billion came from bitcoin's price falling on the mark. IBIT returned 41.52% at net asset value in the September quarter, which is why one redemption lands so heavily on a single fund's balance sheet.

Bitcoin's weakness reaches the miners through a margin that has already closed. CoinShares data put the average cost of producing one bitcoin at about $75,000 at the end of the June quarter, with hashprice near post-halving lows; bitcoin itself was $75,880 on Nov. 7. The average cost of production now sits at the bitcoin price, leaving the average producer with roughly zero margin. Because that figure is an average, the highest-cost producers are already underwater. The miners' answer has been to rent their power to AI, and the economics are real — AI and high-performance computing earn roughly $1.5 million per megawatt per year against about $0.5 million for mining. The problem is that this makes a miner's earnings a levered bet on the same AI capital spending that broke the Nasdaq on Nov. 5.

Bitcoin's corporate buyers used to be the shock absorber under ETF outflows. That buffer has thinned. Strategy bought 811 bitcoin for $65.5 million between Oct. 26 and Nov. 1 at an average of $80,846, taking holdings to 849,327 coins, and repurchased $208.2 million of its own preferred stock in the same week. But published models now assume the stock trades at about 1.1 times the value of the bitcoin it owns, so every new coin has to be funded rather than bought at a premium.

Over the next few weeks, three dated events matter more than any forecast.

Over one to three years, the channel's role in the price changes more than the price itself does.