Deal analysis · estimated

Brex × Capital One: who got paid on a down exit

BRBrexacquired byCACapital One$5.15bn

On January 22, 2026, Capital One agreed to buy Brex for $5.15bn in roughly half cash and half stock; the deal closed on April 7. That is a real outcome — and it is also 42% of what the company was worth in October 2021. Brex raised $1.157bn of equity across seven rounds, and the last two of them went in at $7.4bn and $12.3bn. When the price lands below the marks, the preference stack stops being paperwork and starts deciding who gets paid.

The round history

Four of the seven post-money valuations were publicly disclosed; * marks the three we estimated — the 2017 seed (never disclosed), the 2018 Series B (reported around $220m) and the May 2020 extension (reported just over $3bn). The April 2019 $100m Barclays warehouse line was debt and sits outside this stack.

The waterfall at $5.15bn

With standard 1x non-participating preferences, each class picks the better of two outcomes: take its money back, or convert to common and share the proceeds. The 2021 rounds bought 8.0% of the company between them — worth about $410m at this price against $725m invested. So they don't convert. They take the cash back and walk.

• • $725m of the price — about one dollar in seven — goes back out as preference to investors who make nothing on it.
• • The common still clears roughly $1.99bn, because the whole stack is only 22% of the price. Assuming the founders held 12% between them, that is about $577m; a 15% employee pool is worth about $722m.
• • The 2017 seed is worth about $775m — more than the entire 2021 pref stack got back, on $7m invested.

The founders

Two Brazilians who met arguing on Twitter in high school, sold the payments company they built at 16, dropped out of Stanford eight months in, and were running a unicorn before either turned 23.

PFPedro FranceschiCo-founder & CEOHDHenrique DubugrasCo-founder & former chairman

The investors

Investor	In the deal	Best est. return
RIRibbit Capital	led A and B; C, C-2, C ext., D	~111x on the seed, par on the D
Y Y Combinator	W17 batch; A, B, C, C-2	~111x on the seed
DSDST Global	co-led C; C-2, led C ext., D	~3.7x on the C, par on the D
GRGreenoaks	co-led C; C-2, D, co-led late-stage	~3.7x on the C, par after
IVIVP	C, C-2	~3.7x on the C
KLKleiner Perkins	led C-2	~1.6x
TITiger Global	led D; late-stage	par — cash back, no gain
TCTCV	co-led late-stage	par — cash back, no gain

The part almost everyone missed

The headline on this deal was the discount: $5.15bn against a $12.3bn peak, filed under the 2021 hangover. But look down the cap table and the interesting thing is that both stories are true at once, inside the same firm. Ribbit Capital led the 2017 seed and kept writing cheques through the Series D. The seed dollars come home at something like 111x. The Series D dollars come home at par — five years, no return. Same company, same conviction, same investor; the entry price did all the work.

That is the argument for a preference stack in one picture. The 2021 money was not wrong about Brex — the company grew into a business doing $700m of annualised revenue and turned cash-flow positive in 2025 — it was wrong about the price, and the pref is what turns being wrong about the price into getting your money back instead of losing 60% of it. Founders trading a term sheet with a high headline number against a clean one should read this table as the cost of that headline.

The other thing worth naming: this is what a good outcome looks like when the last round was the wrong round. Airtable is the version where the stack eats almost everything and the common is left with scraps; DroneDeploy is the version where the price clears the whole stack and every class converts. Brex sits between them: the preference bites, but only the top two classes, and the people who built the company still keep the large majority of the price.