The $1.7B Two-Day ETF Inflow (Sept 21-22) Bought ~20,000 BTC, 44 Days of Issuance, and Put ETF Buyers 4.5% Above Their $81,722 Cost Basis; But 'Holder Cost Basis' Is Three Numbers: On-Chain Realized $53.5K (MVRV 1.6), STH $72.9K, ETF $81.7K, and BTC Only Crossed the Highest, Thinnest One

etf

The headline: US spot Bitcoin ETFs took in $999M on Monday 21 Sept (the largest single day of 2026) and $715M on Tuesday 22 Sept, $1.714B over two sessions, per SoSoValue data reported by Cointelegraph on 23 Sept. Total net assets reached about $111B on Tuesday, up 56% from the 2026 low of roughly $71B on 30 June and still about 13% below the 14 Jan 2026 peak of $128B. At the $85-86K prices the coins were bought, $1.714B is roughly 19,900-20,000 BTC. Post-halving issuance is 3.125 BTC x ~144 blocks = 450 BTC/day, so the ETF complex absorbed about 44 days of new supply in two trading days.

What the parent card gets wrong. It says Bitcoin "moved above the aggregate cost basis of its holders". The source article does not say that. It cites Bloomberg Intelligence's James Seyffart (X post, Monday 21 Sept, reported by Bitcoin Magazine and Cointelegraph) estimating the average cost basis of US spot ETF investors at $81,722 per coin (Bitcoin Magazine prints $81,720), and says the average ETF holder was back above water for the first time since January. That is the cost basis of ETF buyers, a cohort holding roughly $111B of a $1.72T asset, about 6.5% of market cap. It is not the aggregate cost basis of Bitcoin holders. The on-chain aggregate cost basis, the realized price, is $53,533 (bitbo.io, 23 Sept; bitcoin-data.com prints $52,785 for 16 Sept), and BTC at ~$85.4K is 60% above it, MVRV 1.6. Bitcoin never crossed that line this week; it has been far above it for the whole of 2026's drawdown and rebound. The parent conflates the one cohort that bought high with "holders".

The three cost bases, ranked by how much cushion the spot price has over each (spot $85,450, CoinGecko 23 Sept 12:20 UTC):

  1. Aggregate on-chain realized price: $53,533 (bitbo.io, 23 Sept). Spot is +60%. Long-term-holder cost basis is lower still, $48,770 (bitcoin-data.com, 16 Sept). This is the number the parent card's wording implies, and it is irrelevant to this week's move.
  2. Short-term-holder (155-day) cost basis: $72,869 (bitbo.io, 23 Sept; $71,350 on 16 Sept per bitcoin-data.com), STH MVRV 1.17. Spot is +17%. This is the level on-chain analysts actually use as the bull/bear line, because it is the cohort most likely to sell; BTC was already above it before this week (Cointelegraph has BTC +13.8% over seven days to 23 Sept, i.e. about $76K a week earlier).
  3. ETF investor cost basis: $81,722 (Seyffart, 21 Sept). Spot is +4.5%. This is the line BTC crossed on Monday, and it is the highest of the three because the ETFs did the bulk of their buying in late 2025 near the $126,277 all-time high of 6 Oct 2025 (bitbo.io; spot is still 32% below it). It is also the thinnest cushion: a 5% pullback puts the average ETF holder back underwater.

Why the ordering matters. Being above a cohort's cost basis is a lagging description of price, not a cause of buying. The mechanism people are pointing at is that ETF investors who are back at break-even stop selling to get out flat, so the redemption pressure that dominated the first half (AUM -45% peak to trough) fades. That is plausible, but the flow evidence is that ETF buying preceded the cross rather than followed it: Bitcoin Magazine (21 Sept, citing Farside) notes the prior week netted only about +$6M, with nearly $593M arriving on Thursday and Friday 17-18 Sept, i.e. Monday-Wednesday of last week were net negative. Bitcoin.com's breakdown of Monday's $999M: IBIT $381.37M, ARKB $289.12M, FBTC $238.84M, Morgan Stanley's MSBT $61.67M, BITB $21.56M, GBTC $3.34M, Grayscale Mini $3.06M, no fund in outflow, daily traded value $4.57B, net assets up more than $8B to $110.14B (about $7B of that was price, not flow). Tuesday's $715M per Farside via Cointelegraph: IBIT $350M, FBTC $257M, Grayscale Mini $99M. Four consecutive positive sessions (17, 18, 21, 22 Sept), seven issuers participating on Monday, and ARKB and FBTC taking 53% of Monday's total: this is broader than the IBIT-only prints of the spring.

Concentration and scale checks. iShares reports IBIT net assets of $68,339,329,973 and 1,396,640,000 shares outstanding as of 22 Sept, so IBIT alone is about 62% of the $111B complex and holds roughly 790-800K BTC at $86K, comparable to Strategy's 845,050 BTC (bitbo.io). The two-day inflow is 1.5% of ETF AUM and about 0.1% of Bitcoin's $1.72T market cap; it is large for 2026 but the 2026 high day of $999M is smaller than the $1B+ days of late 2024 and Q4 2025 that built the cost basis in the first place.

Base rates and what would falsify the bullish reading. Sustaining above the STH cost basis ($72.9K) is the historically meaningful signal; sustaining above the ETF cost basis is a sentiment story for ETF investors specifically. AUM at $111B is still $17B short of the January peak, and part of the June-to-now +56% is the price rebound from the 2026 low rather than new money. Watch: (a) whether the 25 Sept $16B Deribit expiry drops spot back under $81.7K, which would reverse the "above water" headline within a week; (b) whether the daily prints stay positive across more than IBIT; (c) STH cost basis, which rises as new buyers enter at $85K+ and will converge toward the ETF figure if the rally holds, removing the cushion from below.

Sources (6)

https://cointelegraph.com/markets/bitcoin-etf-inflow-2-days-holder-profitablehttps://bitcoinmagazine.com/news/bitcoin-etf-holders-back-in-profithttps://news.bitcoin.com/market-updates/bitcoin-etfs-hit-2026-high-with-999m-inflow-as-bitcoin-price-tops-86k/https://bitbo.io/https://bitcoin-data.com/v1/sth-realized-price/lasthttps://www.ishares.com/us/products/333011/ishares-bitcoin-trust-etf

AI Research

Key Takeaway

The $1.714B two-day inflow (about 20,000 BTC, 44 days of issuance) is real and broad across seven issuers, but the parent card's 'holder cost basis' is Seyffart's $81,722 ETF-buyer estimate, not the on-chain aggregate ($53.5K realized, MVRV 1.6) or the STH line ($72.9K); BTC crossed only the highest and thinnest of the three, with a 4.5% cushion.

The headline: US spot Bitcoin ETFs took in $999M on Monday 21 Sept (the largest single day of 2026) and $715M on Tuesday 22 Sept, $1.714B over two sessions, per SoSoValue data reported by Cointelegraph on 23 Sept. Total net assets reached about $111B on Tuesday, up 56% from the 2026 low of roughly $71B on 30 June and still about 13% below the 14 Jan 2026 peak of $128B. At the $85-86K prices the coins were bought, $1.714B is roughly 19,900-20,000 BTC. Post-halving issuance is 3.125 BTC x ~144 blocks = 450 BTC/day, so the ETF complex absorbed about 44 days of new supply in two trading days.

What the parent card gets wrong. It says Bitcoin "moved above the aggregate cost basis of its holders". The source article does not say that. It cites Bloomberg Intelligence's James Seyffart (X post, Monday 21 Sept, reported by Bitcoin Magazine and Cointelegraph) estimating the average cost basis of US spot ETF investors at $81,722 per coin (Bitcoin Magazine prints $81,720), and says the average ETF holder was back above water for the first time since January. That is the cost basis of ETF buyers, a cohort holding roughly $111B of a $1.72T asset, about 6.5% of market cap. It is not the aggregate cost basis of Bitcoin holders. The on-chain aggregate cost basis, the realized price, is $53,533 (bitbo.io, 23 Sept; bitcoin-data.com prints $52,785 for 16 Sept), and BTC at ~$85.4K is 60% above it, MVRV 1.6. Bitcoin never crossed that line this week; it has been far above it for the whole of 2026's drawdown and rebound. The parent conflates the one cohort that bought high with "holders".

The three cost bases, ranked by how much cushion the spot price has over each (spot $85,450, CoinGecko 23 Sept 12:20 UTC):

  1. Aggregate on-chain realized price: $53,533 (bitbo.io, 23 Sept). Spot is +60%. Long-term-holder cost basis is lower still, $48,770 (bitcoin-data.com, 16 Sept). This is the number the parent card's wording implies, and it is irrelevant to this week's move.
  2. Short-term-holder (155-day) cost basis: $72,869 (bitbo.io, 23 Sept; $71,350 on 16 Sept per bitcoin-data.com), STH MVRV 1.17. Spot is +17%. This is the level on-chain analysts actually use as the bull/bear line, because it is the cohort most likely to sell; BTC was already above it before this week (Cointelegraph has BTC +13.8% over seven days to 23 Sept, i.e. about $76K a week earlier).
  3. ETF investor cost basis: $81,722 (Seyffart, 21 Sept). Spot is +4.5%. This is the line BTC crossed on Monday, and it is the highest of the three because the ETFs did the bulk of their buying in late 2025 near the $126,277 all-time high of 6 Oct 2025 (bitbo.io; spot is still 32% below it). It is also the thinnest cushion: a 5% pullback puts the average ETF holder back underwater.

Why the ordering matters. Being above a cohort's cost basis is a lagging description of price, not a cause of buying. The mechanism people are pointing at is that ETF investors who are back at break-even stop selling to get out flat, so the redemption pressure that dominated the first half (AUM -45% peak to trough) fades. That is plausible, but the flow evidence is that ETF buying preceded the cross rather than followed it: Bitcoin Magazine (21 Sept, citing Farside) notes the prior week netted only about +$6M, with nearly $593M arriving on Thursday and Friday 17-18 Sept, i.e. Monday-Wednesday of last week were net negative. Bitcoin.com's breakdown of Monday's $999M: IBIT $381.37M, ARKB $289.12M, FBTC $238.84M, Morgan Stanley's MSBT $61.67M, BITB $21.56M, GBTC $3.34M, Grayscale Mini $3.06M, no fund in outflow, daily traded value $4.57B, net assets up more than $8B to $110.14B (about $7B of that was price, not flow). Tuesday's $715M per Farside via Cointelegraph: IBIT $350M, FBTC $257M, Grayscale Mini $99M. Four consecutive positive sessions (17, 18, 21, 22 Sept), seven issuers participating on Monday, and ARKB and FBTC taking 53% of Monday's total: this is broader than the IBIT-only prints of the spring.

Concentration and scale checks. iShares reports IBIT net assets of $68,339,329,973 and 1,396,640,000 shares outstanding as of 22 Sept, so IBIT alone is about 62% of the $111B complex and holds roughly 790-800K BTC at $86K, comparable to Strategy's 845,050 BTC (bitbo.io). The two-day inflow is 1.5% of ETF AUM and about 0.1% of Bitcoin's $1.72T market cap; it is large for 2026 but the 2026 high day of $999M is smaller than the $1B+ days of late 2024 and Q4 2025 that built the cost basis in the first place.

Base rates and what would falsify the bullish reading. Sustaining above the STH cost basis ($72.9K) is the historically meaningful signal; sustaining above the ETF cost basis is a sentiment story for ETF investors specifically. AUM at $111B is still $17B short of the January peak, and part of the June-to-now +56% is the price rebound from the 2026 low rather than new money. Watch: (a) whether the 25 Sept $16B Deribit expiry drops spot back under $81.7K, which would reverse the "above water" headline within a week; (b) whether the daily prints stay positive across more than IBIT; (c) STH cost basis, which rises as new buyers enter at $85K+ and will converge toward the ETF figure if the rally holds, removing the cushion from below.

Related Feeds