🧠 Generative AI Word Salads + Robinhood's Gold Card

It's contrarian to believe Generative AI will disrupt Fintech, which is wild to me. I Ranted + Robinhood’s new Gold card & Why V/MC are fine but consumers aren't.

Welcome to Fintech Brainfood, the weekly deep dive into Fintech news, events, and analysis. Join the 32,264 others by clicking below. 

Hey Fintech Nerds đź‘‹

What a week.

Robinhood’s new Gold card credit card is the biggest product launch of the year so far. The 3% cashback looks like a loss leader designed to attract an affluent customer base. The news also broke Chime is likely to IPO in 2025, and the idea that it could “compete” with banks seems to have everyone upset 🤷‍♂️

That all happened in the same week merchants “won” a $30bn settlement with Visa and Mastercard to cap fees and enable surcharging.

This means your Robinhood Gold, or Chase Sapphire Card could all wind up with additional fees at checkout. And that’s before we even mention preferring wallets (Apple, Google, Paze??!)

Skip to đź‘€ Things to Know for a roundup of the important bits.

The đź“Ł Weekly Rant was a FUN one. We’re at Peak AI. Jensun and Zuck are wearing bad jackets and doing deals for H100s. Weirdly, in Fintech, it’s contrarian to think Generative AI might actually be useful.

PS. I was on the Moody’s KYC podcast this week and had a BLAST. Check it out here or search for KYC decoded in your podcast app.

PPS. This one is too long for Gmail users, so head to the substack for more.

Here's this week's Brainfood in summary

đź“Ł Rant: Gen AI and Word salads

đź’¸ 4 Fintech Companies:

  1. Invest Alert AI - The GenAI investment co-pilot

  2. Simple closure - The business closing service 

  3. Lettuce - Automated tax & accounts for solopreneurs

  4. Nuke - Instantly kill financial apps if your phone is stolen.

đź‘€ Things to Know:

Weekly Rant đź“Ł

Gen AI and Word salads

Behold.

A Mckinsey word salad about Generative AI and finance

TL;DR creates committees and a command structure to figure out the low-hanging fruit opportunities.

Plus, whatever the hell this cube thing is.

Why does this matter?

It matters because people often tell me Generative AI is a sustaining innovation. It benefits incumbents in financial services and not disruptors. 

They have a strong case on current evidence.

Accenture has done more revenue from Generative AI projects than all foundation model companies combined.

I disagree.

Generative AI is disruptive, but not for the reasons you think. It's not about AGI (Artificial General Intelligence); it's about disrupting the industry's cost structure, which is exactly what Cloud and mobile did in the past decade.

How?

  • Incumbents can't get out of their own way

  • Fintech disrupted the cost structure of distributing finance

  • Generative AI disrupts the cost structure all over again

  • There are lots of tasks that couldn't be automated with APIs or UIs that now are

  • Most disruptors have already adopted Generative AI in their workflow and product

  • The SaaS era of Generative AI is here for at least a little while

  • What's next is these small wedge products displace outsourcing and consulting

The Generative AI disruption has already started. Incumbents won't benefit from it because they'll be so busy forming committees and doing PowerPoint slides. They can't execute.

Fintech can.

Let's make that tangible.

1. Most incumbents cannot get out of their own way

Every board met to discuss what to do about AI and ended up giving the CEO the three talking points so they have an "AI story." How did that story come about? They hired strategy consultants.

Those strategy consultants used slides from previous work to plot this innovation against a matrix and help many people internally reach some consensus about the jargon.

Meanwhile, the security team blocked ChatGPT access.

The level of bottoms-up innovation is 0 when you restrict staff access entirely. 

Microsoft (and, to some painful extent, Salesforce) has enabled "Corporate GenAI" co-pilots inside their product suites. So now you can compose a corporate PowerPoint or email with AI instead of doing it yourself. 

The revenue generated by Accenture generates the sheer demand for Generative AI projects to understand how GenAI might be used. I wouldn’t be surprised if most Fortune 500 companies had more than one $1-2m GenAI “discovery” project running. Add those up, and you quickly get into the billions.

I’ve written about how Generative AI is this decades Robotic Process Automation. It is a sensible starting point as a sort of “Magic duct tape” to take cost out of an organization. But most of this is not novel or industry-changing. Incumbents gonna incumbent.

There are exceptions. Some novel things are happening, like the ability to directly pull in financial market data and make sense of it. I’ve even seen banks run incredibly sophisticated “hallucination checker” experiments. These companies have incredibly smart people doing things that would surprise you.

But that doesn't shift the market; it's a glorified iOS update.

It doesn’t change the paradigm or their fundamental cost structure.

What shifts the market is a step function change in the unit economics or the product being offered.

2. Fintech has disrupted cost structures

a) APIs, Mobile, and Cloud computing massively reduced the cost of distribution. From 2010 onwards, companies like Stripe, Square, Chime, Plaid, and Monzo began building their entire businesses using modern software engineering principles as the default. 

Unlike incumbents, they rarely had their own data centers, mainframes, or or physical stores. These entirely digital businesses benefitted from the massive investment Amazon, Google, and Apple made in creating data centers and mobile device app stores for manufacture and distribution.

b) Everything is an API. Next the entire financial services value chain became API-first. What used to be a bank department or an outsourced provider was now 10x more available. Canonically, Stripe allowed anyone to accept payments online with 9 lines of code without speaking to sales. 

Today, this is table stakes, but for the financial services industry, you might as well have made gravity work backward. Everything from customer onboarding to late payment collections and fraud detection now has a suite of API-first providers who massively reduce time to market for new entrants.

This might seem obvious to you, person who works at a Fintech company. But to the bankers, it's still not obvious

But here's the rub.

c) CAC and CTS are 10x lower. Nubank's cost to onboard a customer is less than $10; its cost to serve is in that ballpark for the entire year. Compared to its peers, it's 10 times cheaper. This means they can do two things.

  1. Serve lower income segments profitably.

  2. Make more profit from higher income segments

This is true for most Fintech companies that serve end customers and reach scale. It's also changing the shape of companies that use Fintech products. I wrote several weeks ago about how, on average, finance teams are 10x smaller at digital companies because modern B2B Fintech technology companies have solved so many problems.

It's common for a Series A company to have one person in their finance department and as few as 3 in Series B. APIs dramatically reduced the number of people needed to run any part of an organization. 

d) The next phase of financial services is the cost battleground. Incumbents generate a lot of revenue by being diversified and having a wide product catalog. Disruptors come from beneath with lower costs and gradually diversify their offerings to generate more revenue. But the reality is that this cost battleground plays out at all levels. 

(Below: The market map I’ve been using for a while, showing where it’s a unit economics battle vs where it’s a features and new distribution battle).

  1. By product line: Direct deposit or brokerage accounts compete for the same customers.

  2. By demographic: Robinhood and Public.com are chasing similar demographics, while E*Trade and Schwaab do. Incumbent vs. incumbent and disruptor vs. disruptor are just as, if not more, competitive.

  3. By vintage: Companies born in 1990 have tech and process debt, but so do companies born in 2010+. The companies heading for IPO are vulnerable to disruption from new 3rd Gen Fintech companies that might be able to outcompete on costs or features. 

e) The best adopters of any new technology are the companies that were born at (or after) its invention. The banks I consider quite good at technology, like Capital One and Chase, are still on their path to using cloud-native technologies. They have higher fixed costs, and they're winning vs their direct competitors, but they're not capturing the opportunity that a 10x reduction in costs brought to Fintech companies.

GenAI has entered the chat and is naturally suited to attacking costs. 

Klarna's AI "doing the work of 700 people is the start." In case you missed it, in a fantastic bit of PR, Klarna announced their GenAI-powered chatbot:

  1. Took user queiries from 11 mins to 2 mins to resolution

  2. Increased customer satisfaction

  3. Speaks over 45 languages

  4. Did the equivalent work of 700 staff

This is why people say GenAI is a sustaining innovation. Here, we have a company that's at scale and using it to run its business. My take is slightly different. 

Generative AI is impacting Fintech companies because they can actually use it. Big banks are still being McPowerpointed and comitteed to death.

A generation of companies will be born without those cost structures in the first place. Just as fintech companies disrupted the industry's cost structure through APIs, cloud computing, and mobile technology, Gen AI will do that all over again

3. Generative AI Disrupts cost structures all over again

To see where Generative AI changes cost structures, we have to look at where cost is created, even in the most mobile, cloud, and API-first businesses in the universe. 

a) APIs, machine learning, and UIs don't solve everything. User experience and APIs have trended towards power through simplicity. The fewer choices you give a user, the more friction you can remove, and generally, the better. Machine learning also needs a well-defined problem set, well-labeled data, and solid data scientists to ensure it works correctly.

In a perfect world, we'd design everything this way.

But the world isn't perfect. 

Some variables are fixed outside the control of any one entity or company. Regulators love a PDF, every country has a different format, every state does company incorporation differently, and legacy banks love a PDF, and don't always have great (or any) API. 

With enough time and money, you can engineer the perfect waterfall to manage these edge cases.

The problem is without the time and money, you can't.

Small companies don't have time or money. They'll use whatever can solve the variable, complex tasks, and edge cases.

b) Generative AI is good at conditional tasks. The list of things GenAI can do quite well is staggering, from diagnosing symptoms and passing law exams to summarizing long documents. It can generate realistic images, audio, and video in the broader set. 

So much is circumstantial in financial services. 

For example, the right financial advice for you might differ from someone with the same financial profile. Why? Maybe you value sustainability and health while they're a risk taker who loves the outdoors. 

Another example is two identical companies applying for lending. Still, they operate in different industries with different risks of bribery, corruption, or sanctions (e.g., oil and gas vs big boutique retail). 

Circumstances matter. Reducing the data set to make them fit in a UI or API doesn't always work. 

The tyranny of great UIs and APIs is one-size-fits-nobody designs that aren't contextually relevant to you or your circumstances. At scale, these user experiences are designed to boost engagement KPIs and attract and retain new users. 

The gap is for the bespoke, personalized offering or one that can understand its conditions and circumstances. Building a process or beautiful software is hard if the possible inputs and outputs are highly varied.

Gen AI does this well.

c) GenAI is good at complex tasks. Complexity is another dimension to consider. Knowing the detail of a 2,000-word document and how that relates to a 5,000-word compliance manual is complex. It requires a lot of background experience and reading comprehension. Knowing which compliance rule applies, at which point, for which customer can take a lifetime. 

Every organization has its own "risk appetite" for applying controls. In a perfect world, there would be an unlimited budget and headcount for a specialist in everything. But again, the world isn't perfect. So, compliance officers become generalists expected to do incredibly complex things at high volumes.

Gen AI does this well. 

d) Tasks I've seen automated. There's a GenAI startup for every compliance activity; I've also seen it for complex customer onboarding (like business onboarding). There are two low-hanging fruit areas.

  1. Wherever a PDF or process can't be automated or turned into software because it is too circumstantial, GenAI is appearing.

  2. Wherever there's a complex problem that UIs and APIs haven't solved (like wealth advisory or personal finance management)

Surprisingly, the vanilla GPT, Claude, and Gemini models are pretty bad at this type of task without extremely complex prompts. 

In fact, even with custom GPTs (the new feature designed to turn ChatGPT into an app store), they're still quite poor unless incredibly well prompted

That's the problem; it's also the short-term opportunity. 

4. The problem with GenAI today

A narrative says the 10,000+ GenAI companies will die because they're all paying OpenAI, which is paying Nvidia, so the only real winners are chip makers. In the short term, I buy that. In the long term, I don't.

The market is more nuanced.

  1. Mainstream LLMs (like ChatGPT) are as good (or bad) as their users' prompt ability. LLMs (Large Language Models) mirrors how we use spreadsheets. Everyone has their own. Some are better than others. Most people are bad at prompting and haven't found the edges of where models help them.

  2. No single LLM model rules them all. Outside of ChatGPT, there are lots of smaller, open-source models and specialists. A well-balanced smaller model, tuned to a use case, will outperform a generic model that is poorly prompted by a user. The specialist and open-source models deliver much more bang for the buck.  

  3. GenAI and LLM performance isn't static. The prompt that worked yesterday might not work today. I'm also seeing people getting fatigued with marketing copy that is clearly generated. (e.g., "in a complex and evolving landscape" is a dead give away)

  4. LLMs raise the floor, not the ceiling. LLMs have become an aggregation of above-average intelligence but not world-class intelligence. That could change given the development speed over the past 18 months, but consider this final point. 

  5. Even with incredible progress, demand outstrips supply. Sam Altman is chasing $trillions in Abu Dhabi to build a new chip manufacturing project. I'd speculate that's because we've reached the ceiling of what the current chip fab capacity can deliver. Nvidia just smashed Moore’s law and is building incredible chips with the B100, but the first one costs $10bn, and it will take a while to scale. We can make chips 5x faster and 5x more efficient, but that’s useless if we can’t build them fast enough to make progress.

Even if you assume the existing high costs of inference and API calls continue to fall, we still don’t know where financial services regulators will land on the topic.

My broad frame for LLMs and GenAI is that it's definitely not artificial general intelligence. It's incredible, but on the scale of software-as-a-service and cloud. That's awesome, but not AGI (Artificial General Intelligence). 

At least not until we can scale up chip production hyper-exponentially.

It's the alternative when a UI or API doesn't cut it. 

We're in the SaaS era of GenAI.

At least until we can build 20 more TSMC fabs in Abu Dhabi.

5. GenAI's SaaS era

a) GenAI makes existing Fintech companies more cost-competitive and pushes them into new parts of the market. Most Fintech companies have a GenerativeAI offering or are using it internally. The next step is the massive disruption of cost structures.

Generative AI impacts Tech Fintech companies because they can use it, unlike incumbents. It makes them even more cost-competitive vs incumbents. That cost advantage could manifest as a business model shift.

Investment banking analysts used to be too expensive for most companies or consumers. What if instead of a team of 3 analysts at $600,000 per year, you can have similar quality for $1,000 per month. How might that disrupt parts of financial services that are traditionally human-led and white-glove only?

b) Vertical GenAI companies solve specific problems, which could be a wedge into more. The blitzscaling of new vertical GenAI companies is useful because they solve many of GenAI's problems today. Companies are good at promoting, knowing when to use which model and when to fine-tune. 

My "4 Fintech companies" have been littered with examples of this type of company. e.g.

  • Casca* pre-qualifies and manages the loan application process for SMBs and banks

  • Tennis Finance helps companies screen marketing copy for UDAAP compliance (supported with a fractional compliance officer)

  • Coris checks a merchant you're onboarding matches the industry they claim and watches their public-facing site

c) There's a gap for things that didn't work as a UI or API to try GenAI. Over time, this will look like SaaS, but for jobs that used to be outsourced, even at tech companies like Meta. Areas like content moderation, complex business customer queries or onboarding, and stuff that never quite worked as a mobile UI.

The disruption is on the cost side of the equation.

But in turn, this creates new markets.

The net impact of SaaS wasn't sustaining incumbents. It was a generation of companies born with 10x less fixed costs for going to market. 

The next decade will see a generation of companies born with the skill of knowing which GenAI models to bake in and how to make them effective and fine-tuned. 

6. What's next

a) Fintech companies will start to spin out specialist GenAI teams. A team from Ramp just left to found cognition labs, the world's "software engineer" AI company.  

Devin can plan and execute complex engineering tasks requiring thousands of decisions. Devin can recall relevant context at every step, learn over time, and fix mistakes.

b) Consultants are having a bad year. They're about to have a bad decade. Consulting's resourcing model always took frameworks and a few smart senior leaders with relationships backed by lots of college grads. Cynically, they'd then take PowerPoint formats that worked elsewhere, interview internal staff (pump), and play that back in the PowerPoint (dump). GenAI is very good at that entry-level work. Consulting budgets at big organizations are 

c) We'll see new creative financial planning solutions. PFM and financial planning were always the last miles the UI and API couldn't solve, but for the 10% of the population who love that stuff. GenAI can understand user context and complexity.

d) Open source becomes a much bigger part of the mainstream conversation. To date, we think of closed models and a chat interface like ChatGPT, Claude, or Gemini when we think of LLMs or Generative AI. However, consider just how many A100 chips Meta bought and their positioning with Llama 2 as the open-source standard bearer. Llama 3 and what comes after could be very interesting.

e) Supply will eventually catch demand. The investment in global chip manufacturing is astonishing. Models are becoming smaller and more power-efficient. We’re seeing astonishing progress; watching Nvidia keynotes feels like living in a sci-fi movie. These supply chain issues will take another 5 years to solve. But new paradigms always benefit the new, smaller companies before it does the incumbents.

So here's your summary:

  • Gen AI is a disruptive force in fintech, unlike the sustaining innovation narrative often presented by incumbents and consultants.

  • Fintech companies are better positioned to leverage Gen AI due to their nimble and innovative nature, while bureaucracy and committees often bog down incumbents.

  • Gen AI has the potential to significantly reduce cost structures across various aspects of fintech, from customer service to compliance and beyond.

  • While the current "SaaS era" of Gen AI is limited by chip production capacity, the future holds the promise of companies born with the ability to seamlessly integrate and fine-tune Gen AI models into their workflows.

  • Human-only tasks like investment banking analysis could get democratized in the next cycle setting off a whole new wave of disruption.

We're past peak hype for GenAI. 

Every company had it in their pitch. 

But right at the edge, a new platform is being born. 

It's fashionable to roll your eyes at GenAI in Fintech and wonder where the innovation is. You could do the same with anything bigger than the industry like cloud or mobile.

But if Generative AI is magical duct tape, the future is builders who are amazing at using this magical duct tape to problem solve in a way incumbents never could.

They’ll build amazing things for ultra-low costs because they can.

Cost innovation is innovation in opportunity. 

We wouldn't have Nubank or CashApp unless it was 10x cheaper to serve SMBs and consumers with new technology. 

What does it look like if it's 10x cheaper to have a financial advisor or CFO in your pocket?

ST.

4 Fintech Companies đź’¸

1. Invest Alert AI - The GenAI investment co-pilot

Invest Alert will risk assess an investment portfolio, provide recommended actions, and offer "AI-powered" allocation suggestions. It's aimed at consumers to explain past performance clearly and the rationale for recommendations. The service connects to brokerage accounts like Robinhood, Schwaab, eToro, and robo advisor Nutmeg. 

🧠 Is this ChatGPT in disguise? I always try to get hands-on with any service that's got a chatbot and AI in the name, but couldn't in this case. An "alerts" service as a chatbot is an interesting frame. They've obviously built enough to monitor a portfolio and provide alerts when the market changes. At worst, this is some decent prompting through the GPT-4 API; at best, this is a low-key financial advisor, which triggers my second thought. 

🧠 There's a fine line between financial advice and general best practice. If GenAI i giving you best practice advice based on your portfolio  ut only ever serves recommendations, how's that different to advice? 

2. Simple closure - The business closing service 

Simple closure aims to cut the legal fees, stress, and 100s of hours of admin for founders shutting down a company. It pulls together accounts and investor asset distribution and manages dissolution, winds up, and shuts down with a step-by-step process.

🧠 Not every company should succeed. In the post-ZIRP era, it's not surprising that so many companies are closing down, and accordingly, a whole host of companies are popping up to help them close down (a smidge of irony there). Nine out of 10 companies fail, but there's no end-of-life support for companies due to closing, and founders want to move on to the next thing. 

3. Lettuce - Automated tax & accounts for solopreneurs

Lettuce helps companies file as an S Corp, automatically calculates taxes owed, and allows companies to make quarterly payments. The service will make recommendations, produce cash flow statements, help file payroll taxes, and aims to save users $10,000 per year. Fintech nerds will also note they offer a debit card (issued by Trans Pecos Bank)

🧠 This is the accounting platform that also does business banking. I love this product idea and company name; I hope it succeeds. They've started at the admin and work backwards to the banking. I like that, and it makes the monetization route obvious. S-Corps are designed for freelancers to save money, but they are often not taken advantage of because of the admin. Things like mileage and use of a home office are deductible. 

PS. The reason I love the name is twofold. 1) It's a vegetable that doesn't get enough love for its comedy value. 2) A Lettuce lasted longer than UK rime Minister Liz Truss in a famous UK live stream. Which is the most British thing ever. 

4. Nuke From Orbit - Instantly kill financial apps if your phone is stolen.

Nuk is a consumer-facing app and service that allows you to instantly wipe your mobile device if stolen. It will auto-lock banking, brokerage, and Fintech apps and cancel any associated cards. It will also block new SIMs to prevent SIM swap attacks and secure any social media or email accounts. Users add their bank cards and create networks of other devices (or people) they trust to secure their accounts in the event of device theft.

🧠 This should be a service every bank and Fintech company offers by default. We live our entire lives on mobile phones, and while some of us have PIN codes or 2-factor authentication, even those can be stolen. The service can't order new bank or SIM cards, which is why I think it's the natural partner to banks. Banks deal in trust, which is more trustworthy than "we'll fix things if they go wrong." 

Things to know đź‘€

The big three stories this week are all on a theme. How much money can you make from a card payment? First up. Robinhood.

Robinhood’s Gold card will cost $5 per month or $50 annually and be available to Robinhood’s existing 1.24m Gold members. It offers 3% back on everything, 5% back on travel booked through the Robinhood portal, no foreign transaction fees, and up to 5 authorized users.

🧠 On the surface, this looks like a loss leader. 3% cashback is more than the revenue.

🧠 They need the cross sell to work into the wider Robinhood ecosystem for the economics to make sense. The Visa Signature preferred card will do a blended 2.2% interchange revenue, way less than the cashback. (h/t Matt Jones for the image below)

🧠 The reality is more complex. The two target segments are high net worth (transactors who pay the balance in full) and revolvers building their credit. Robinhood earns interest on the revolvers. That competes squarely with Apple Card.

🧠 The high net worth value proposition is about cross-selling. Once someone joins Gold, they tend to take more products and increase ARPU. Robinhood expects high net worth customers will be transactors (pay full balance) but take lots of other products like high yield savings. That competes with JP Morgan.

🧠 I’d love to see the data on Gold customers. Is the ideal client of Gold a YOLO memestock trader, trading options who’d need to build their credit, or are they more financially savvy? I imagine that Robinhood wants more high-net-worth customers, and its wider high-yield savings and pension product is designed to attract them. But the reality of the YOLO traders is they’re more likely sub prime. So they’d make money on the interest.

🧠 It uses a separate app from the main Robinhood, which could break the cross-sell opportunity. The ideal client profile for Robinhood is someone who has and will grow wealth. This card is designed to attract that customer, but it’s in a whole separate app. How does cross-selling work in that scenario? It’s an interesting trade-off because transactional and card are very different experiences and sets of jobs to be done in brokerage.

🧠 The fact they’re going after affluent segments makes them much more of a threat to big banks than Chime. Travel, points, and cashback are the sweet spot of the big banks. The trick is to determine if they can make those customers profitable over that lifetime. Or is the 3% just an introductory offer to hook a bunch of customers?

🧠 The terms and conditions on that 3% back? It can be used for more card spend or within the Robinhood ecosystem. I guess this is designed to solve the two apps problem and, at a minimum, make the cost worth the expense. (net-net, it’s 0.8% cashback if they’re collecting 2.2% on the next spend)

PS. How might the Visa & Mastercard settlement with merchants impact all of this?

ICYMI: Visa and Mastercard have agreed to cut US interchange fees in a settlement merchants say will “save $30bn in fees per year.” It will also allow merchants to charge different prices to users based on their credit or debit card. 🧠 Visa and Mastercard will be fine. This won’t impact their revenue and likely won’t save merchants all that much. The much bigger fee (the merchant discount rate or MDR) is set by issuers. The Credit Card Competition Act (AKA, Durbin 2.0) aims to cap the fees on credit card interchange, which would be much more impactful.

🧠 The end of "honor all cards." Merchants must accept Visa and Mastercard if they pay $0.21c or 3%. Today merchants can surcharge with AMEX but not Visa or Mastercard. Now they can surcharge whatever is lower: 3% of the transaction or 3% of the acceptance cost.

🧠 This does make the experience worse for premium rewards-based cards. Consumers might not use their Chase Sapphire or similar premium card if they have to pay a surcharge everywhere.

🧠 This will likely impact Google and Apple Pay. Merchants can choose which wallets get accepted and preferred. You have to wonder where Paze and Visa+ would fit in this picture. (huge h/t to Scott Wessman)🧠 This is only for debit cards, not credit. Fintech will be fine. Fintech is propped up on Durbin-exempt cards for Fintech companies. Smaller (sub $10bn asset) banks can charge more for their swipe fees than the big banks. They share the 2% transaction fee with large Fintech companies who use it as a key revenue source. 🧠 The merchant win may be bad for consumers. The surcharges for using another card could also be framed as a "junk fee." You'll now have to think hard about which card to use and when.

Chime planning a 2025 IPO, according to Bloomberg. With 38m customers, Chime may have more “primary” users than Bank of America. Ron Shevlin says about half “consider” Chime to be their primary checking account, which would put them between PayPal and Wells Fargo in market share.

🧠 Chime has reshaped the market. An IPO in 2025 would be a crowning moment for them. The “get paid early” wedge has turned into 10s of millions of customers and active usage.

🧠 A low cost to acquire and run accounts is crucial. Their demographic today skews lower income, which is harder to make a profit from, despite their low-cost operating model. The path to profitability is likely via going up market and/or cross-selling.

🧠 Chime will cross-sell successfully, but I’m not sure its brand can go upmarket. Nobody has quite nailed that premium millennial Fintech consumer experience high ground Monzo and Starling took in the UK. The UK digital banks’ first cohorts are aging into being high deposit balance customers. Chime’s wedge in get paid early and credit building just doesn’t feel aspirational. That’s not a knock; the market needs what it does desperately. It’s just a brand question.

🧠 “Primary” checking account is such a BS term. Chase defines it as anyone with over $500 monthly deposits plus 5x transactions. But that could also mean it’s a paycheck motel. Money comes in and goes out. Primary used to mean the top of the wallet and first in line to cross-sell. Today it just means collecting a good chunk of consumer deposit balances 🤷‍♂️

🧠 Does Chime ever want to be a bank? When they IPO they’ll have that sweet tech company multiple. Banks get valued very differently. But I think the trick is to be a growth stock. You can do that with a balance sheet. It’s working for Nubank, it’s sort of working for SoFi (which, arguably, it shouldn’t), and it will work for Monzo. If I were advising them, I’d stay away from the charter for ~3 to 5 years until they figure out lending, and then buy a charter to improve economics further down the line.

For more on the unit economics of digital banks vs incumbents check out the longer form piece I wrote a couple of weeks ago, “Fintech has already disrupted banking.”

Good Reads đź“š

This, by my good friend Luca, is so good it made me guffaw and clap several times. It's worth it for the first two paragraphs alone. This is a dense read, but one I promise you, filled with absolute bangers like 

"Yeah, you are reading this correctly, [it is a] a derivative of a derivative of a re-hypothecated digital asset"

That's all, folks. đź‘‹

Remember, if you' e enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)

(1) All content and views expressed here are the authors' personal opinions and do not reflect the views of any of their employers or employees. 

(2) All companies or assets mentioned by the author in which the author has a personal and/or financial interest are denoted with a *. None of the above constitutes investment advice, and you should seek independent advice before making any investment decisions.

(3) Any companies mentioned are top of mind and used for illustrative purposes only. 

(4) A team of researchers has not rigorously fact-checked this. Please don't take it as gospel—strong opinions weakly held 

(5) Citations may be missing, and I've done my b st to cite, but I will always aim to update and correct the live version where possible. If I cited you and got the referencing wrong, please reach out