How AI makes Uber and Nubank stronger, not weaker

GM. This is Milk Road PRO, hunting the consumer platforms the market is mispricing one active user at a time.

Today's edition features our latest PRO report on a group of scaled apps that get tougher to disrupt as machines get smarter, and the single number Milk Road uses to prove it. The bet: each loyal, repeat customer is the one asset automation can't easily rebuild, and Wall Street is still underpricing it. You'll get the opening below, with the rest on the site.

Here's what we've got for you today:

📊 The metric behind the whole thesis: profit per active user, and why cheaper automation raises what each loyal customer is worth.
🚗 Uber as the top pick at 6.1%: 202M monthly active users, $28 operating profit each, with AVs and takeout optionality as the swing factors.
🏦 Nubank at 4.3%: $13.30 monthly revenue per active customer against under $1 to serve, a U.S. bank charter around 2027, and stablecoin rails across Latin America.

Bitget Stock 2.0 is a complete rethink of how tokenized U.S. equities should work. Check out Bitget Stock 2.0 and start trading tokenized equities today.

Prices as of 2:00 p.m. ET. Powered by CoinGecko.

AI'S 'LOSERS' ARE ACTUALLY WINNING

I've been bullish on what we've called super apps for years.

But the only way I owned that thesis was through crypto super apps: Coinbase and Robinhood. Both live and die with trading cycles.

When volume dries up, so does the profit. The idea was right, but the expression was too aggressive.

So in the last few months I added Uber, Nubank, and Klarna.

They might be the better representation of the same thesis: their businesses are more durable, far less cyclical, and the repricing upside is wide open.

Today I want to look at all three through the distribution-platforms lens I've been building.

All three are scaled consumer platforms with real, recurring users, where AI and blockchain cut costs and deepen the relationship instead of threatening it.

The more AI fear grips the market, the more a real human who opens the app and transacts every week is worth. An owned, recurring relationship is the one asset agents and automation can't cheaply rebuild, and that scarcity is exactly what the market is discounting in my view.

The number that separates a durable platform from a big-but-stale app is profit per active user (operating profit divided by active users).

That's the core of my “Distribution Platforms” thesis. I still hold Coinbase and Robinhood, but not as the only bet in this thesis.

Go PRO to see my full portfolio and trade alerts - here!

THE FRAMEWORK

The easiest mistake in consumer investing is confusing users with economics. A big app is not automatically a good business.

But before I even get to economics, I ask a harder question: is the core business durable?

This is the lesson from Coinbase and Robinhood.

Both convert users into profit better than anything else in the book. The problem is what that profit rests on. Trading revenue rises and falls with the cycle, so profit per active user swings with the market instead of compounding through it.

A great conversion rate on a cyclical base is still a cyclical business.

So the core has to clear two hurdles:

First, is the revenue durable and ideally growing across cycles, not just in good years? People move, eat, bank, and pay in every market.

Second, does that durable usage convert into operating profit? Uber, Nubank, and Klarna pass the first gate in a way the crypto names can't. That's why I added them to my portfolio.

For each company I track five things:

Source: Milk Road PRO

Read together, they tell one story. Active users are the asset. Revenue and profit per user show whether that base is getting more valuable, not just bigger.

If S&M (Sales & Marketing) costs per user fall while profit per user climbs, growth is paying for itself. EV (Enterprise Value) per user is the price tag you are paying on the market.

The setup I want is profit per active user rising while EV per user stays low, because that gap makes up the repricing opportunity.

The active-user metric differs by company (Uber's MAPCs - Monthly Active Platform Customers, Nubank's active customers, Klarna's active consumers, Robinhood's funded customers, Coinbase's MTUs), but the underwriting question is the same for all: is the active base getting more profitable over time?

The full thesis rests on four pillars:

Source: Milk Road PRO

That fourth pillar is where the money is: the re-rating, not just better fundamentals. This shift occurs as the marketplace recognizes that disruption, when wielded correctly, transforms into significant leverage.

THE FY2025 SCOREBOARD

Using FY2025 as the base period, here's the scoreboard (profit and EV per active user):

Source: Milk Road PRO

Uber and Nubank already turn users into profit. Klarna is the cheap, pre-inflection name. Robinhood and Coinbase screen highest on profit per user, but the market already pays for it.

TOKENIZED STOCKS JUST GOT AN UPGRADE

Onchain stocks have always had the same three problems:

Terrible liquidity
No dividend tracking
Dead capital

Bitget Stock 2.0 fixes all of that. It's a complete rethink of how tokenized U.S. equities should work:

0.04% trading fees
Near-zero slippage
Real-time dividend distributions
Stocks that can actually be used across DeFi
Sourced directly from Nasdaq and NYSE

The gap between TradFi and DeFi is getting smaller.

Check out Bitget Stock 2.0 and start trading tokenized equities today.

THE FY2025 SCOREBOARD

1. Uber: the action layer

Uber is the cleanest starting point. The behavior is obvious: people move, eat, and want local convenience delivered.

FY2025 numbers:

202M monthly active platform consumers.
13.6B trips.
$193.5B gross bookings.
$52.0B revenue.
$5.6B operating income.

That works out to roughly:

$258 revenue per MAPC.
$28 operating profit per MAPC.
$24 sales and marketing per MAPC.
$781 enterprise value per MAPC.

In multiple terms, the market is valuing each active Uber customer at roughly 3x revenue, or 28x operating profit.

The bull case: Uber's platform now spans rides, delivery, grocery, advertising, and Uber One. Once a user is on it, Uber can lift profit per user through trip frequency, ad load, better matching, routing efficiency, and AV (Autonomous Vehicles) partnerships.

The peer comparison backs the argument:

Source: Milk Road PRO

Uber generates roughly twice DoorDash's operating profit per active user at a lower EV per user. The comparison asks the right question: is Uber valued too narrowly?

The technology layer is where I think the market is most asleep.

AI lifts monetization (sharper matching, dynamic pricing, an ad engine trained on real purchase intent) and lets Uber ship features faster on a flatter cost base.

Both push operating profit per MAPC higher.

AVs are the bigger prize, and I think Uber wins as they scale. It has committed more than $10B across roughly 30 AV partners, runs autonomous rides in eight cities, and targets 15-plus by the end of 2026.

Management's case: AVs change who supplies the car while Uber still owns demand, which Khosrowshahi calls a multi-trillion-dollar opportunity. Robotaxi fleets still need someone to fill seats at high utilization, and Uber has 202M consumers to point them at.

Blockchain is a minor lever here (stablecoin rails for cheaper driver payouts), not a reason to own it.

The bear case is real. Regulation, insurance, driver classification, take-rate ceilings, and delivery competition all bite.

The sharpest risk is AV disintermediation, already visible as Waymo expands through its own app across roughly 20 cities, partners with Lyft in Nashville, and exits its Uber tie-up in Phoenix. If the best AV operators aggregate their own demand, Uber's edge thins. That is what I’m watching the closest here.

One piece the market rarely prices: Uber as a takeout.

A giant that’s building AVs but is short on consumer demand (Tesla, Amazon, or Alphabet) may find it faster to buy distribution than build it, and Uber offers an instant global rider base, operations, payments, and a marketplace that fills cars.

There's a real chance one of them makes a run once robotaxi economics are proven. I don't underwrite the name on it, but it's live optionality that cushions the downside.

My take on Uber: this is my highest-conviction name, a 6.1% position, because the thesis already shows in cash flow, buybacks, and margin expansion, not in a number I'm projecting.

Grow MAPCs while lifting profit per MAPC, and it earns a local-commerce multiple instead of a rideshare-and-delivery one. AVs are the swing factor, and I think they break Uber's way.

The takeout optionality is a bonus, not the reason I own it.

2. Nubank: low-cost banking at scale

Nubank is the cleanest financial example of the theme and the mid-sized compounder in the book.

FY2025:

131M customers.
83.4% activity rate.
~109M estimated active customers.
$13.30 monthly ARPAC (average revenue per active customer).
$0.80 monthly cost to serve per active customer.
$10.6B revenue.
$3.9B pre-tax income proxy.

Roughly:

$97 revenue per active customer.
$35 profit per active customer.
$2 sales and marketing per active customer.
$465 enterprise value per active customer.

In multiple terms, the market is valuing each active Nubank customer at roughly 4.8x revenue, or 13.2x operating profit.

The engine is the spread between what a customer generates and what they cost to serve: $13.30 a month against under $1.

Add products (deposits, cards, loans, payroll, insurance, investments) and profit per active customer climbs.

The SoFi comparison helps:

Source: Milk Road PRO

SoFi earns more revenue per member and similar profit, but spends far more on marketing to get there. Nubank's edge is scale, low cost to serve, and cheap distribution.

The technology levers look a lot like Uber's, but pointed at a bank.

AI sharpens underwriting, fraud detection, and support automation, which is how a customer stays profitable at under $1 a month to serve, and it lets Nubank build and localize products faster as it enters new markets. Same operating-leverage story, applied to credit.

Blockchain matters more here than for any other name I own.

In Latin America, stablecoins are dollar access, and that is the product: roughly 90% of Brazil's crypto activity already runs through them, with heavier demand in higher-inflation markets.

Nubank has built for it, with 15-plus assets in Nubank Cripto, a yield on USDC balances, in-app swaps, and its own custody, plus a former Brazil central bank governor now on its board pushing stablecoin payments on Nubank cards.

As dollar-denominated money spreads across the region, Nubank can own the rail.

The U.S. ties both together. In Jan. 2026 Nubank won conditional OCC approval for a national bank, Nubank, N.A., covering deposits, cards, lending, and digital asset custody, with a launch targeted around 2027.

That opens the world's largest banking market to a model proven at 109M customers, and lets Nubank carry its AI cost base and stablecoin rails in from day one.

One caveat: Nubank is a bank. Profit per active customer has to be risk-adjusted.

ARPAC growth without credit discipline isn't value creation. Grow lending too fast, let losses season badly, and your whole business is at existential risk.

My take on Nubank: it can be a long-duration financial compounder, a 4.3% position, if ARPAC rises, cost to serve stays low, and credit quality holds. People will keep banking.

The risk I weigh most is credit losses outrunning ARPAC, which would expose the AI-credit edge as cyclical or overextended rather than structural.

The U.S. charter and stablecoin rails are real upside the ~$465 EV per customer isn't paying for yet.

3. Klarna: the BNPL stigma discount

Klarna is the newest name here, and the only one I own as a smaller position rather than a core position.

It IPO'd on the NYSE in September 2025 and is building from payments, into a digital bank.

FY2025:

118M active consumers (+28% YoY).
$127.9B GMV (+22% YoY).
$3.5B revenue (+25% YoY).
$65M adjusted operating profit (1.9% margin).
A net loss for the year, with EPS of -$0.79.

Roughly:

$30 revenue per active consumer.
~$1 operating profit per active consumer.
~$65 enterprise value per active consumer.

In multiple terms, the market is valuing each active Klarna customer at roughly 2x revenue.

The profit multiple matters less here: Klarna is still early, only recently reached operating breakeven, and is clearly choosing growth over margin.

The whole case is that profit per active user climbs from here.

The repricing gap is a different shape than the others.

The market prices Klarna like a fragile pay-in-four lender: credit losses, discretionary-spend exposure, consumer weakness, and a commoditized checkout button.

That stigma keeps the multiple low, and the ~$65 EV per active consumer against Nubank's ~$465 shows how low.

My case is that the market has the category wrong. Klarna is turning into a checkout distribution network with bank funding, widening product density, and AI-led operating leverage.

In Q1 2026, non-transaction operating expenses rose only 3% while revenue grew 44%.

Since Q4 2022, revenue is up 104% while operating expenses fell 8% and headcount dropped 49%.

Post-IPO neglect is part of the setup. Underrated U.S. growth is part of the upside. The BNPL stigma discount is the main source of the mispricing.

The pure-play comparison helps to sharpen the point. Affirm is the closest listed BNPL peer:

Source: Milk Road PRO

Affirm earns roughly 5x Klarna's revenue per user because it is mostly an interest-bearing lender (about 72% of volume), monetizing each user harder and carrying more credit risk.

Klarna's base is larger and lighter, full of low-monetization Pay Now users it can graduate into financing, card, and banking.

The number that matters is the multiple: Klarna trades near 2x revenue, Affirm around 7x.

The market is paying up for the pure-play lender and discounting the deposit-funded bank with five times the users - and that gap makes up the trade.

The risk that breaks it: Klarna never escapes BNPL economics.

If take-rate compresses as PSP distribution scales while credit losses season badly on growing Fair Financing volume, then Klarna is just another pay-in-four lender adding risk to buy growth.

That single combination, falling take-rate and rising losses against rising volume, is the one I keep the closest eye on.

My take on Klarna: I own it as a smaller, probationary starter, a 2.2% position, not as the next Nubank.

The thesis is attractive, but one good quarter isn't proof.

I want two more clean quarters of widening product density and contained credit before I'd write it up as a high-conviction compounder rather than an asymmetric starter.

4. Coinbase and Robinhood: crypto exposure

I own both, but as crypto exposure rather than core repricing bets. Both score high on profit per active user. Both are also the two names where the market has already closed most of the gap.

Robinhood, FY2025:

27.0M funded customers.
$4.5B revenue, $2.1B operating income.
~$166 revenue and ~$78 operating profit per funded customer.
~$3,046 enterprise value per funded customer.

In multiple terms, the market is valuing each active Robinhood customer at roughly 18.3x revenue, or 39.0x operating profit.

Source: Milk Road PRO

Profit per funded customer is strong. The price is the problem. Schwab earns far more per account because its accounts are deeper and more mature, yet Robinhood's EV per funded customer already sits close to Schwab's.

The market is paying for the platform transition before it's proven. Robinhood also leans on cyclical trading, so the metric only holds through a cycle if durable products (assets, subscriptions, cash, retirement) carry it.

Coinbase, FY2025:

9.2M MTUs.
$7.2B revenue, $1.5B operating income.
~$781 revenue and ~$158 operating profit per MTU.
~$4,030 enterprise value per MTU.

In multiple terms, the market is valuing each active Coinbase customer at roughly 5.2x revenue, or 25.5x operating profit.

Profit per MTU is the highest in the group, but the metric flatters a cyclical business.

The stronger story is the mix shift: $2.8B subscription and services revenue, nearly 1M Coinbase One subscribers, $17.8B average USDC balances, and 12 products above $100M annualized revenue.

Coinbase gets more durable as profit leans on institutional, custody, stablecoin, and subscription revenue instead of retail trading. But at ~$4,030 EV per MTU, the market already prices a lot of that in.

So both stay in the book as crypto exposure. Neither is the cleanest expression of the repricing thesis, because for these two the repricing has largely happened.

THE DISRUPTION DISCOUNT

Two things happen at once as intelligence gets cheap. Cost to serve falls (which drops straight to operating profit) and the customer base gets harder to copy.

An agent can spin up a checkout flow or a lending screen in seconds. It can't conjure 202M people who already trust Uber with their card and open the app out of habit.

Software is the part that commoditizes. Distribution is the part that holds.

So the question I ask of any AI or blockchain claim is narrow: does it improve the unit economics of a user relationship that already exists at scale?

A slide about AI proves nothing, and neither does a blockchain product nobody uses.

The only evidence that counts is profit per active user moving up, through lower cost to serve, higher conversion, or a new revenue line that sticks.

The concrete cases sit in the sections above.

AI shows up in Uber's matching and ad engine, Nubank's underwriting and sub-$1 cost to serve, and Klarna's operating expenses falling while revenue climbs.

Blockchain shows up in Coinbase's stablecoin and custody revenue and Nubank's dollar rails across Latin America. Each one is a technology that reaches the income statement.

That's where the re-rating comes from.

The market is still charging these platforms a disruption discount while the data already shows the leverage building.

Close that gap and the multiple moves.

WHAT PROVES THE THESIS RIGHT

The thesis is working if I see:

Active users grow or stay highly engaged.
Revenue per active user rises.
Operating profit per active user rises.
Sales and marketing per active user stays flat or falls.
Product attach increases.
AI or automation shows up by creating lower cost to serve, better risk outcomes, faster product velocity, or higher conversion.
Blockchain or new financial rails show up in stablecoin, custody, tokenized asset, settlement, payment, or onchain monetization where relevant.
Margins expand without damaging retention or trust.
Valuation migrates from narrow category multiples toward platform-quality multiples.

Company-specific proof points:

Source: Milk Road PRO

WHAT PROVES THE THESIS WRONG

The theme breaks if:

Active users grow but operating profit per active user doesn't.
Sales and marketing rise just to hold growth.
Product expansion fails to attach.
AI stays as just a narrative and never lowers cost to serve, improves risk, or raises product velocity.
Blockchain products fail to build durable revenue or weaken trust and regulatory standing.
Monetization damages trust or retention.
Regulation compresses economics.

The clean falsifier: if active users grow but operating profit per active user doesn't improve, I'm wrong.

MY POSITION

Go PRO to see my full position - here !

BITE-SIZED COOKIES FOR THE ROAD 🍪

Pre-internet money rails meet the agentic economy? Yikes. Arc is Circle’s open Layer-1 built for the internet-native economy. Humans and AI welcome.*

Saylor: If BTC appreciates 0% a year, Strategy has 30-40 years of dividend coverage. "If Bitcoin appreciates about 3% a year, we can pay the dividends forever."

Kyle Reidhead: Eli Lilly is heading to $5,000 and it has almost nothing to do with an AI miracle drug.

Tom Lee: AI agents are about to become wealthier than humans. "We're going to start to question whether we work for the AI agent or they actually work for us."

Bitcoin lending is making a comeback. Several major US banks (like Ledn) have begun offering Bitcoin-backed credit lines.**

*this is sponsored content. **this is partner content.