Sep 3, 2026 · 59m · a16z
Why AI Agents Could Finally Reinvent the Credit Card
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Fintech pioneers Max Levchin and Alex Rampell join Erik Torenberg on The a16z Show to analyze payment economics, recount the strategic origin and underwriting architecture of Affirm, and evaluate how artificial intelligence could reinvent digital checkout.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 5.5% of the talking time here. How this is scored →
The host as informed peer 2.3 Guest teaching 4.5 Guest disagreement 1.4 The host pushing back 0.1
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 53:56 Alex challenges Max on agentic shopping utility
Alex openly disputes Max's premise that autonomous shopping will fail, arguing consumers will readily delegate buying exact SKUs to agents to optimize price.
Hardest push from the host ▶ 13:01 Erik checks Max's claim of being late to trendsErik directly interrupts Max's self-deprecating claim that he is habitually late to technological waves by citing PayPal as an obvious counterexample.
Biggest teaching moment ▶ 3:55 Max breaks down card network latency constraintsMax educates the room on the rigid 2.5-second technical rule imposed by Visa and Mastercard and how Apple Pay cleverly engineered around it using local secure enclaves.
The host holds their own ▶ 48:29 Erik links PayPal Mafia density to Thiel FellowshipErik synthesizes the historical discussion of PayPal's unique talent density by offering the Thiel Fellowship as the only modern equivalent with a comparable hit rate.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| The Evolution of Tap-to-Pay and EMV | 3 | 5 | 1 | 0 | Erik opens with a broad prompt asking the guests to reflect on 20+ years in fintech. Alex explains the EMV liability shift and how merchant terminal upgrades accidentally catalyzed tap-to-pay adoption. | |
| The 2.5-Second Rule and Trillion-Dollar Markets | 3 | 6 | 1 | 0 | Erik redirects the question to Max, who delivers a technical masterclass on the card networks' strict 2.5-second authorization window and why payments niches are universally massive. | |
| Payment Volume Versus Take Rates | 1 | 3 | 2 | 0 | Alex and Max engage in a direct back-and-forth about take rates vs transaction volume, invoice factoring, and their early discarded PayMeSooner concept without host involvement. | |
| Biometric Payments and the Law of Critical Mass | 3 | 5 | 1 | 0 | Erik asks why certain intuitive payment ideas fail to emerge. Max explains the law of critical mass in payments and why the physical credit card interface remains unbeatable. | |
| Crypto, Convenience, and the Lessons of PayPal | 4 | 6 | 2 | 2 | When Max claims he is frequently late to tech trends, Erik pushes back by noting PayPal was early. Max clarifies PayPal succeeded precisely by abandoning the cypherpunk obsession with anonymity for practical consumer convenience. | |
| The Pajama Problem and Affirm's Early Spark | 2 | 4 | 1 | 0 | Erik prompts the founders to recount Affirm's genesis. Alex details his TrialPay background, pitching Max in Russian, and formulating the 'pajama problem' around mobile checkout friction. | |
| From Social Credit to Machine Learning Underwriting | 1 | 4 | 1 | 0 | Max shares his transition after Slide, studying informal credit practices in Israel and Japan, and pursuing machine learning underwriting over basic payment processing. | |
| The 1-800-Flowers Pitch and Initial Hurdles | 1 | 3 | 1 | 0 | Alex and Max recount pitching 1-800-Flowers founder Jim McCann with a bespoke PHP prototype, encountering early merchant skepticism over pricing from operator Amit Shah. | |
| Discovering Product-Market Fit with DTC Mattresses | 2 | 5 | 1 | 0 | Erik inquires about reaching product-market fit. Max explains how moving installment options upfunnel at Beautylish yielded a 30% conversion lift, paving the way to DTC mattress brands. | |
| Eliminating Predatory Deferred Interest | 1 | 5 | 2 | 0 | Max details exiting online coding bootcamps due to poor course completion quality, then passionately critiques deceptive deferred-interest credit cards as Affirm's core anti-pattern. | |
| Affirm as a Demand Engine with Negative CAC | 3 | 5 | 1 | 0 | Erik asks what the market misunderstands about Affirm. Max and Alex explain how Affirm acts as an advertising demand engine with negative customer acquisition costs. | |
| The Strategy Behind Long-Term Installment Loans | 4 | 5 | 1 | 0 | Erik draws a sharp comparison between the PayPal Mafia and the Thiel Fellowship. Max details how PayPal deliberately hired future founders and how seeing peers under stress demystifies greatness. | |
| Agentic Payments Versus Agentic Shopping | 2 | 3 | 3 | 0 | Max argues agentic shopping is overhyped compared to agentic payments. Alex pushes back with a nuanced counterexample regarding price-optimization for known SKUs via sites like CamelCamelCamel. | |
| The Adoption Curve of Autonomous Commerce | 2 | 4 | 2 | 0 | Max concedes that high-trust commoditized tasks like Instacart grocery buying are already agentic, though full commerce delegation will follow a bumpy adoption curve before Erik wraps. |
Statements from this episode (28)
Rampell: Tap-to-pay was unused on initial rollout before becoming ubiquitous
“Nobody was using taps when those machines came out, and now it's ubiquitous.”
Levchin: Visa and Mastercard enforce a strict 2.5-second transaction limit
“Visa and MasterCard have a hard two and a half second limit on interaction between the network issuing bank and the merchant and the acquiring bank.”
Levchin: Every niche in payments is at least $100 billion
“There are no niches and payments that are smaller than a hundred billion dollars, and so that's probably the most surprising thing I've seen.”
Rampell: Starbucks Pay was designed to avoid per-transaction payment processing fees
“That's why Starbucks invented Starbucks Pay, because they wanted to stop having to pay. It's like these prepaid, they only have to pay one time. You chart, you put 50 dollars on your Starbucks account, well, they only have to pay Visa, MasterCard, and everybod…”
Rampell: Higher payment transaction sizes yield smaller percentage take rates
“What's surprising about payment area under the curve, if you will, is that, like, the bigger the dollar is, like, the smaller the rake, which I guess makes sense, because you're not gonna get two percent of that 40 trillion dollar payment”
Levchin: Revenue opportunity in B2B financing is lower than consumer lending
“There, there's some healthy businesses these days in both accounts, payable financing and accounts receivable financing. So it's not, This is, it's not a bad idea, but the revenue opportunity, despite being an unregulated lending space, is seemingly lower than…”
Levchin: Card payment interface is the best UI ever created
“The card payment interface is the singular best user interface ever created, and it, it's not something that people haven't tried to do better.”
Levchin: Payments innovations either reach critical mass or fail completely
“There's some unobvious a priori, but clear a posteriori version of critical mass. And if you don't reach it, you're going to fail. Like there's not really a okay outcomes in payments. You either get there and everyone needs to have your widget or your network …”
Levchin: Crypto has proven successful as a store of value, not payments
“And I'm not convinced it's a payment method now, and it's been quite some time. But as currency and commodity and a star value goes, it's certainly proven to be extraordinarily successful.”
Levchin: Convenience trumps all other factors as transaction size shrinks
“The convenience just trumps everything else as the total amount you're trying to send goes down.”
Rampell: Active searchers for emergency credit represent high credit risk
“But if you go on Google and you say, like, I'm out of money, need money, credit, credit, credit, like you're probably a bad credit risk. So kind of going back to the general store concept, People that aren't looking for credit, that might not have their wallet…”
Levchin: Affirm's original goal was building a credit score, not lending
“As a person recovering from doing lots of machine learning in the service of social media, which was, you know, good and well, but wasn't really what I was meant to do, and lusting for the days of doing machine learning to fight fraud, my MO was, I just want t…”
Levchin: Up-funnel Affirm installment options drove a 30% conversion increase at Beautylish
“And the only difference that Beautylish Did as they implemented Expedite, I think by then it may have already, I think it was named the firm by then, was that they told their shoppers up funnel basically as they were selecting their shampoos and perfumes that …”
Rampell: For-profit education institutions accept the highest MDRs, up to 50%
“The highest MDRs are actually for-profit education institutions. So some of them are, like, 50, like, I mean, I don't like these guys in particular, but, like, University of Phoenix.”
Levchin: Affirm exited online education financing due to terrible quality
“A firm briefly flirted with being a payment provider to the online educational facilities, which we ran out of kicking and screaming because the reason they're willing to pay these enormous MDRs is because the quality of education is largely terrible.”
Levchin: Affirm will never charge late fees or use deferred interest
“There will never be an asterisk on a firm zero. That, that is the origin story of why we don't charge late fees, we don't do deferred interest, all the sort of gimmicks and gotchas that the industry loves to throw at people.”
Levchin: Affirm has transacted with over 50 million Americans across four countries
“We've transacted with roughly 50 plus million Americans alone, and we're now live in four countries and growing pretty quickly.”
Rampell: Affirm operates with negative customer acquisition cost
“And like, I think to me, one of the coolest things about a firm to this day is it has negative customer acquisition cost. I mean, it is paid to acquire a customer.”
Rampell: Consumer business models degrade as CAC value flows to Google and Facebook
“I mean, Casper got to scale, but lots of CAC, and that's why a lot of these business models, they tend to degrade over time, because all the economic value goes to Google and Facebook.”
Levchin: Affirm is the only BNPL provider doing multi-year installment loans
“So we are the only company in the industry really, there's plenty of competitors who are trying and some degree of success, but mostly not really, who will go into longer term loans. And by longer, I don't mean mortgage or, you know, 15 plus years, but like th…”
Levchin: FICO scores are insufficient to underwrite multi-year installment loans
“And to do that, you have to underwrite. Like, you can't shortcut the, I'll just look at your FICO score, or I'll just sort of, you know, I'll look at your Facebook friends. Like, none of that works. You actually have to do a real, very sophisticated degree of …”
Levchin: Long-term installment loans serve as touchpoints to upsell financial products
“These long term loans, which are actually quite hard to manage and quite hard to maintain reasonable default rates and delinquency rates on, is the price you pay as a provider of financing to upsell consumers on more and more services.”
Levchin: PayPal deliberately selected and hired future startup founders by design
“As we were interviewing, certainly the early team and even the later team,
We kept on selecting for entrepreneurs. We would ask, like, one of the stock questions I would ask was, so what are you going to do after PayPal? And the answer that I welcomed the most…”
Levchin: Seeing peers' human vulnerabilities inspired PayPal alumni to pursue massive ideas
“So if you know kind of the true base version of who we all are, as you watch one of these people go for something really big, like Elon Taking humanity to Mars, which is extraordinary. And yet, you know, he's a human being. Like, I've seen him sweaty and tired…”
Max Levchin predicts AI agents will handle payments before autonomous shopping
“I'm probably less optimistic about agentic shopping, and I'm very optimistic about agentic payments.”
Levchin: AI agents may finally replace the credit card interface
“This notion of sort of the best user interface ever created is the credit card. This may actually be finally up for renegotiation because agents are in fact smarter than pieces of plastic and even pieces of plastic with rewritable chips.”
Levchin: Adoption curve of autonomous AI commerce will likely be slow
“I think the adoption curve is probably going to be fairly slow, and these things are always sort of easy to predict to be very slow, and sort of the usual adage as well, slowly at first, not always once.”
Levchin: Instacart grocery delivery is already 100% agentic commerce
“Grocery shopping is 100% agentic, like, for anyone who uses Instacart. Like, you tell your Instacart shopper, go bring me milk. You don't think twice when they say, oh yeah, you wanted whole milk from Organic Valley, but I found some other brand, and here it i…”