The “cash out” that is a quarter not cash — Snowflake’s co-founder has now run the same 66,668-share tranche three times Benoit Dageville · Snowflake Inc. (NYSE: SNOW)
- Who
- Benoit Dageville, co-founder of Snowflake and a sitting director (the form reports him as director only, not officer or 10% owner).
- What they did — 21 October 2026
- Sold 50,000 common shares at $321.74 ($16,087,000) and made a gift of 16,668 shares. Both came out of the same vehicle, The Snow Trust UTA dated 9/10/19, where he is trustee. That holding went 2,368,211 → 2,351,543 on the gift and → 2,301,543 on the sale.
- The document
- Form 4, accession 0001821737-26-000017, period of report 21 October, filed 23 October, signed by Marie Reider as attorney-in-fact. The Rule 10b5-1 box is checked, and footnote F1 states that the gift as well as the sale was made under a plan he adopted on 3 April 2026.
| Tranche date | Sold | Wtd avg | Gifted | Gift % |
|---|---|---|---|---|
| 29 Jul 2026 | 50,000 | $280.00 | 16,668 | 25.00% |
| 7 Oct 2026 | 50,000 | $333.42 | 16,668 | 25.00% |
| 21 Oct 2026 | 50,000 | $321.74 | 16,668 | 25.00% |
Consensus read
A founder-director continues routine programmatic diversification under a pre-set 10b5-1 plan. The figure that gets reported is the combined 66,668 shares “cashed out”, valued at the sale price — roughly $18.7m for the July tranche, roughly $21.5m for this one.
On this reading there is no information content. Plan sales are deliberately uninformative; that is what the plan is for. It is a fair reading and it is how every aggregator will log it.
Contrarian read
This is not a selling plan. It is a distribution schedule with a fixed 75/25 split, and a quarter of every tranche never becomes cash at all.
Pricing the gifted leg at the sale price overstates realised proceeds by exactly the size of the gift: $5.36m on 21 October, $4.67m on 29 July. The headline number is wrong by a quarter, in the same direction, every time.
The evidence that discriminates
- The gift is inside the plan. Footnote F1 puts the gift, not just the sale, under the 3 April 2026 10b5-1 plan. Someone who wants liquidity has no reason to pre-commit the timing of a gift; someone executing a transfer target does. This is the single field that separates the two readings, and it is in a footnote rather than a headline.
- Quantity is invariant to price across three executions. 50,000 / 16,668 at $280.00, $333.42 and $321.74 — a 19.1% price range with zero response in share count. A dollar-denominated plan varies quantity; a share-denominated transfer target does not.
- The ratio is exact, not approximate. 16,668 / 66,668 = 25.0002%. A rounding would not land there three times.
That the purpose is transfer rather than diversification is inference drawn from those three fields. The filing says none of it.
What would falsify me
A subsequent tranche that carries a sale with no accompanying gift, or that shifts the ratio away from 25%. Either kills the fixed-transfer-target reading and leaves plain diversification standing.
So would any disclosure that the plan is denominated in dollars rather than shares — that would make the invariant share counts a coincidence of three similar prices rather than a design.
Pattern vs his own history Sixty-four of his Forms 4 since 2021 mention a gift, so the gifting itself is long-running; what is new in 2026 is the fixed pairing. Separately, the same filing carries four grantor-retained annuity trusts all dated 13 March 2025 — The Selene GRAT No. 1 and No. 2 (he is trustee) and The Thira GRAT No. 1 and No. 2 (his spouse is trustee) — holding 358,087, 358,087, 391,913 and 391,913 shares. That is exactly 1,500,000 shares, split exactly in half between his side and hers. Those four counts are unchanged from his 8 October filing. Nineteen months after funding, no shares have moved out. Two live hypotheses: annuity payments are being satisfied in cash from outside the trusts, or these are longer-dated trusts with back-loaded annuities. The filing does not distinguish them. A future Form 4 showing a transfer out of Selene No. 1 or Thira No. 1 would.