The idea arrived over dinner, carried in a friend's pocket. It was 2009, Groupon was filling wallets with paper coupons, and Jason Gardner's friend pulled out a stack. Gardner had already sold one payments company. The friend asked the obvious payment-nerd question: why not put all those offers on one card? Gardner froze. One card for many merchants sounded useful, faintly magical and, according to the industry people he consulted, doomed.
He started Marqeta in Oakland in January 2010 anyway. The first version took about 15 months and a freshly raised $5.2 million Series A to reach customers. The delay was not mainly the card. Gardner's team had decided against using a decades-old issuer processor and built a new one. The consumer product let people prepay for groceries or meals and receive extra value. It was moderately successful in the founder's diplomatic retelling. In plainer English, it did not scale.
Consumer acquisition wanted more capital than the company could sensibly feed it. Gardner also discovered something useful about himself: he was more interested in the machinery than the merchandising. “I like the complexity of building infrastructure,” he later told Forbes. This is the sort of self-knowledge startup folklore usually edits out. Marqeta's first failure was not a moral test. It was a mismatch between a founder, a distribution model and the thing he actually enjoyed building.
Failure number two came with a famous customer
Facebook noticed what Marqeta had built and commissioned a gift card. Launched in January 2013, it could be sent to a friend and used at businesses such as Target and Olive Garden. The social network expected distribution to do what Marqeta's consumer marketing could not. Sales disappointed. Facebook shut the card down roughly a year later.
A second product had vanished, but the pattern around it had changed. Facebook and eBay were not asking Marqeta to become a better consumer brand. They wanted specific payment experiences their legacy infrastructure could not easily produce. Meanwhile, Twilio was demonstrating that complicated communications systems could be offered to developers through clean APIs. Marqeta opened its issuer-processing platform in 2014. The custom machine underneath two unsuccessful cards became the product.
The pivot was not “people finally loved our card.” It was “large companies keep asking to use the part customers never see.”What changed Marqeta's mind
A card that carries permission, not money
Marqeta now sells modern card issuing: APIs and services for building physical, virtual and tokenized debit, prepaid and credit programs. A company can create a card, set where it works, cap an amount, add it to a digital wallet, monitor transactions and manage its life cycle. Marqeta connects that software to issuing banks and the Visa or Mastercard networks, and can supply program management, compliance and risk support around it.
Its defining trick is Just-in-Time Funding. Consider DoorDash's Red Card. A courier arrives at a restaurant that requires payment at the counter. The card need not sit around with a pile of DoorDash money on it. When it is swiped, Marqeta sends the transaction details to DoorDash. DoorDash can check the merchant, the live order and the amount. If the facts match, money is released and the purchase is approved. If they do not, the transaction can be declined. The rules are DoorDash's; Marqeta makes them executable inside a card authorization.
That pattern fits many businesses. Affirm and Klarna can create cards that pay a merchant while the shopper uses a pay-over-time plan. Expense platforms can issue a virtual card for one campaign, vendor or employee, with a limit and expiration date. A crypto service can check a wallet balance and release the fiat needed at the prevailing value. Banks can provision commercial cards without asking corporate clients to wait for plastic.
The product is not merely “issue a card fast.” It is to make the authorization programmable and the data useful. Custom metadata can follow a transaction into a ledger, trimming the miserable work of reconciliation. Webhooks can notify an app the moment a card event happens. Tokenization places credentials in Apple Pay or Google Pay. Marqeta added credit-program capabilities by buying Power Finance in 2023, then acquired TransactPay for about $59.9 million in 2025 to bring regulated e-money and program-management capabilities into the UK and European Economic Area.
The invisible tollbooth
Marqeta's customers are the brands building the financial experience; their users may never encounter the Marqeta name. The roster has included Block's Cash App and Square programs, DoorDash, Uber, Affirm, Klarna, Coinbase, JPMorgan and Western Union. This puts Marqeta between pure software and regulated financial infrastructure. It competes with modern issuers such as Stripe Issuing, Adyen, Galileo, Highnote, Lithic and i2c, with legacy processors such as Fiserv and FIS, and with the expensive option of building directly with banks and networks.
The business earns transaction-linked revenue, including a portion of interchange, along with processing and service fees. From that it pays networks, issuing banks and customer revenue share. More payment volume is useful, but the mix matters: a processing-only program produces less revenue than one where Marqeta provides fuller program management. In 2025, the company processed $383 billion, up 31%, and recorded $625 million in net revenue and $437 million in gross profit.
By the second quarter of 2026, quarterly processing volume had reached $120 billion, up 32% from a year earlier. Net revenue was $176 million, gross profit was $122 million, and GAAP net income was $8 million. It was Marqeta's second straight profitable quarter. Those figures are less cinematic than its June 2021 Nasdaq debut, when shares opened at a valuation above $17 billion, but they tell a better operating story: the engine is processing more, and the company is learning to make money while it runs.
The 45 percent problem
One number refuses to stay backstage. Block accounted for 45% of Marqeta's 2025 net revenue. That is down from 68% in 2023, a meaningful improvement, but still enough for one relationship to shape the company. The Cash App and Square agreements currently run through June 2028 and then renew annually. A renewal on worse terms, a program shift or slower Block volume would be felt immediately.
Block's share of Marqeta net revenue
Marqeta's answer is not subtle: add customers, countries and use cases. TransactPay lets it offer a more complete European package without asking a customer to assemble several local partners. Credit widened the product beyond debit and prepaid. Partnerships announced in 2026 ranged from Expensify's European card program to zerohash stablecoin spending, Riskified fraud decisions and an expanded Google Wallet offering for kids. Diversification is working, but it is unfinished work.
The part worth copying
Marqeta's origin is often told as a perseverance story. The more practical lesson is product archaeology. When a product fails, list the difficult capabilities you had to invent to make it exist. Then ask who values those capabilities enough to buy them directly. Marqeta did not throw away the processor because the loyalty card failed. It watched sophisticated companies ask for the processor, then changed the interface and the customer.
Copy the move, not the mythology
- Separate the failed distribution strategy from the technical asset beneath it.
- Follow repeated requests from demanding customers, especially requests legacy suppliers cannot satisfy.
- Turn the capability into a composable primitive: an API, a workflow or a service with clear boundaries.
- Keep the ugly operational work visible. In payments, the API is only credible because banks, networks, compliance and support sit behind it.
The playbook does not travel everywhere. It fails when the hidden capability is not reusable, when external customers do not control enough volume to justify integration, or when incumbents can copy it cheaply. It is a particularly bad fit for teams unwilling to operate mission-critical systems. Card issuing requires certifications, fraud controls, regulatory work, bank relationships and uptime measured in nines. A clever endpoint without that institutional scaffolding is a demo, not infrastructure.
There is another condition: a platform must avoid becoming a bespoke consultancy for its first big customer. Marqeta's concentration shows the bargain. A giant anchor can validate the technology and drive scale, while also gaining negotiating power and creating a risk visible in every annual report. The best version of the strategy uses early customers to harden a common platform, then deliberately sells beyond them.
Marqeta today is not the card in the photograph. It is the split-second conversation behind the swipe: Is this the right merchant? Is the amount allowed? Is the account sound? Release the money. The company needed two public misses to notice that conversation was what it had built. That is the amusing part of the story, and the useful one. Sometimes the product customers reject is merely the loud casing around a quieter machine they would gladly pay to use.