The Card That Doesn't Want You to Spend

In January 2026, Capital One agreed to pay $5.15 billion for Brex, a company once valued at $12.3 billion. Five months later, investors handed Ramp a $44 billion valuation for chasing the exact same customer with a sharper version of the same idea.

Ramp's year-over-year growth in total purchase volume in March 2026

Here’s what you’ll learn: 

Every corporate card issuer makes money the same way: a small cut of every purchase, called interchange, paid by the merchant. That cut only grows if cardholders spend more, so decades of card design pointed toward one goal: richer sign-on bonuses, deeper point multipliers, better lounges, anything that kept the card moving. It worked extraordinarily well. Ramp's own CEO has noted that American Express is worth roughly $230 billion and JPMorgan Chase more than $800 billion, built in part on that exact mechanic. A business that profitable rarely has a reason to reinvent itself.

The reward programs sitting on top of that mechanic ran on a quiet trick. Issuers marketed points as valuable, then made them progressively harder to redeem for their full worth; Ramp's founders have described this as banks talking up how much a point was worth while steadily devaluing it in the background. Businesses tolerated the arrangement because expense software was sold as a completely separate purchase. A finance team bought a card from a bank, then bought Concur or Expensify to police what employees did with it, then spent hours each month reconciling the two by hand.

That reconciliation wasn't a minor annoyance, it carried a documented price tag. Research from the Global Business Travel Association puts the average cost of manually processing one expense report at $58 and 20 minutes of staff time, with nearly one in five reports containing an error that costs another $52 and 18 minutes to fix. Multiply that across a mid-size company's monthly volume, and finance teams were spending real money just to reconstruct how their own money had already gone out the door, for a process software vendors billed per seat to maintain, no matter how badly it worked.

The deepest flaw was personal risk. Most corporate cards required a personal guarantee, meaning a founder's own credit backed the company's spending. Ramp co-founder Eric Glyman experienced the consequence directly: after Capital One's integration team missed a payment on the corporate Amex tied to his prior startup, American Express came after his personal credit for a bill worth several hundred thousand dollars. That one specific failure became the seed of a much bigger question: what if a card judged a company by its cash instead of its founder's credit score, and still made money exactly the same way the old cards did?

Eric Glyman and his co-founder Karim Atiyeh had a front-row seat to the whole arrangement. Capital One had acquired their first company, a price-tracking app called Paribus, in 2016, and the pair spent the next few years working inside a major card issuer before leaving in 2019 to start something new. Before writing a line of code, they did something unusual for founders in a hurry: they spent months talking with more than 150 finance leaders, CFOs, and founders to find out exactly where the pain lived. The answer kept coming back the same way. Nobody wanted better rewards. They wanted a card company that was actually on their side.

So they built one. Instead of tiered points, Ramp offered a flat, unadorned cashback rate on every purchase, a smaller, honest number instead of a complicated one engineered to be hard to redeem. Then it gave away the software that competitors charged for: expense reports, receipt matching, accounting sync, bill pay. The same interchange fee that funds a bank's rewards program instead funded Ramp's entire product suite, so customers never saw a subscription bill for the tools that made the card worth using. Concur, Expensify, and Bill.com charged per seat for exactly the features Ramp bundled in for free, which made switching an easy call for any finance team running the math.

The other fix was underwriting. Ramp sets credit limits by monitoring a company's linked bank balance in real time rather than pulling a founder's personal credit, requiring roughly $25,000 in a US account and no personal guarantee. That single design choice opened the card to a category legacy issuers had mostly ignored: young, well-funded, cash-rich startups with no credit history to speak of. It also happened to be a direct fix for the exact failure that had cost Glyman his own peace of mind a few years earlier.

Ramp then moved faster than any incumbent could react. An early engineering team of roughly eight people built a credit-card competitor to American Express in three months, then an Expensify competitor six months after that, before expanding into accounts payable, procurement, travel, and a treasury product that has already gathered more than $1.5 billion in assets. Internally, the founders reduced the whole strategy to one formula: growth equals purchase volume times interchange rate, minus funding costs. When they tested which lever moved that number the most, it wasn't the interchange rate, it was volume, so the entire company organized around capturing more of each customer's total spend rather than squeezing more out of every dollar already flowing through the card.

By 2025, the same logic extended into artificial intelligence. Ramp's automated policy checks blocked more than 511,000 out-of-policy transactions in a single three-month stretch, saving customers roughly $291 million; a fraud-detection agent caught a $49,000 AI-generated fake invoice; and a treasury agent has been sweeping idle cash into accounts paying 4% without anyone asking it to. None of it required a human on the finance team to notice the problem first, which was always the point of the business: not spending less for its own sake, but never wasting an hour or a dollar finding out what happened after the fact.

The scoreboard settled the argument in under a year. Ramp crossed $1 billion in annualized revenue in September 2025, and by mid-2026 was reportedly running above $1.5 billion, nearly doubling in under a year while turning cash-flow positive for the first time. Payment volume rose from an estimated $22.3 billion in 2023 to roughly $57 billion by mid-2025, and by its June 2026 funding round the company was processing more than $200 billion in purchases annually, with its customer base growing from 50,000 businesses to more than 70,000 in just seven months. Growth like that, sustained at that size, rarely comes from signing new logos, it comes from existing customers routing more of their total spend through a single platform.

Investors priced that shift in almost real time. Ramp's valuation moved from $13 billion in March 2025 to $16 billion in June, $22.5 billion in July, and $32 billion that November, before reaching $44 billion by June 2026: five markups inside fifteen months, each one arriving faster than the last, backed by an increasingly institutional roster that came to include ICONIQ, GIC, Ontario Teachers' Pension Plan, Goldman Sachs Alternatives, and Morgan Stanley Investment Management. That isn't hype capital chasing a story. It's the kind of pre-IPO money that shows up once growth is already verified in the metrics, not promised in a pitch deck.

Brex tells the opposite story. Founded two years before Ramp with a nearly identical pitch, Brex chased the enterprise market aggressively and, in 2022, stopped serving most small and midsize businesses to focus on venture-backed customers with deeper pockets. It raised more money earlier and reached a higher valuation sooner, hitting $12.3 billion in 2021. But the growth never compounded the same way afterward. In January 2026, Capital One agreed to buy Brex for $5.15 billion, a discount of more than 50% to that 2021 peak, in a deal split evenly between cash and stock. Brex's own leadership framed the sale as an accelerant rather than a retreat, but a company absorbed by a bank isn't setting the market's pace anymore. It's being priced by someone who is.

Ramp, meanwhile, behaves like the incumbent it once set out to challenge. Its internal AI tooling now writes more than two-thirds of its own code, it shipped more than 70 product releases in a matter of months, and it closed two acquisitions (Billhop and Juno) while preparing to launch in the UK and Europe. Eric Glyman has said the company intends to go public eventually, without offering a date. The company that started by promising to help businesses spend less is now building its next act around a category nobody budgeted for three years ago: the token spend of artificial intelligence itself.

Key takeaways to consider…

$44 billion - Valuation after the 2026 Series F round

$1.5 billion - Estimated annualized revenue in 2026

70,000+ - Business customers as of June 2026

$200 billion - Annualized purchase volume

3,200+ - Customers generating $100,000+ in annualized revenue

$12 billion - Customer’s cumulative savings since launch