GM. This is Milk Road AI, where we break down the businesses, bottlenecks, and capital allocation decisions shaping the future of technology.

​Here’s what we’ve got for you today:

Nexo is back in the U.S. - and new clients get 30 days of Wealth Club Premier perks! Higher yields, lower borrowing rates, and crypto cashback - start here.

THE CAPITAL ALLOCATION ARGUMENT NOBODY WANTS TO HEAR

In 1999, Berkshire Hathaway had its worst year in decades.

The S&P 500 was up 21%, while Berkshire was down 19%, and Warren Buffett was being openly mocked in the financial media.

Barron's ran a cover story asking if he'd lost his touch, and investors were fleeing to pile into Cisco, Yahoo, and whatever other dot-com was going up 400% that month.

Buffett didn't chase them, but he just kept asking the same question he always asked, is this a good company? But at this price, relative to everything else I can buy, does this make sense?

Two years later, the Nasdaq had lost 78% of its value, while Berkshire Hathaway remained stable.

That's one of the main questions that matter in capital allocation: Is this the best use of this specific dollar, right now, given everything else available?

That’s the question I keep coming back to every time I look at my portfolio lately, and it’s the exact question that led me to rotate out of Robinhood and into Oracle.

And before you close this email, I am genuinely, unambiguously bullish on Robinhood as a company, and that's actually the whole point.

Let's talk about what Robinhood actually is

Robinhood is a genuinely excellent business. 

27.4M funded customers at an all-time high, net margins above 40%. 

Gold subscribers at 4.34M and growing 36% year-over-year, compounding like a Costco membership, a sticky, recurring revenue layer that makes customers more profitable and less likely to leave regardless of what markets are doing. 

If that scales to 10M subscribers, the earnings math changes dramatically, and the stock deserves a completely different multiple than it carries today.

Prediction markets are also a genuine wildcard. 

Revenue went from $3M a year ago to $104M in a single quarter, and if new contract categories get regulatory approval and the product scales, that's a durable revenue stream that has nothing to do with whether Bitcoin is up or down on any given month.

And the long-term runway is real. 

Their user base skews young and as that demographic ages into retirement accounts, insurance, lending, and wealth management, Robinhood has a monetization opportunity that most financial platforms would kill for. 

The UK is live, EU crypto is coming, and the total addressable market keeps expanding.

None of that is in question. Here's what is.

The Q1 2026 earnings miss and what it actually revealed.

Robinhood reported Q1 earnings, revenue came in at $1.07B, and net income hit $346M. 

Operating cash flow exploded 217% year-over-year, but guess what? The stock dropped 13.24% in a single session.

The street had expected $1.20B to $1.24B, a 13% miss on the top line.

And the culprit was one line item, crypto transaction revenue cratered 47% year-over-year, falling from 43% of total transaction revenue to just 22% in a single quarter.

Bitcoin dropped roughly 30% from its October 2025 peak.

In a business with 40%+ net margins, the marginal dollar of crypto revenue hits the bottom line disproportionately hard, meaning a swing in one volatile asset class can make or break an entire quarter.

This is not a new risk. 

Over recent quarters, crypto has oscillated between 9% and 43% of total transaction revenue. 

That range tells you everything, Robinhood's earnings power is still substantially a reflection of how risk-hungry retail traders feel in any given month. 

Management is building to fix that through prediction markets, Gold subscriptions, and banking products, but those revenue lines still aren't large enough yet to fundamentally change the story.

And that’s really the core issue here.

Robinhood’s stock rose roughly 204% in 2025. 

When that happens, a stock stops pricing in a realistic range of outcomes and starts pricing in the optimistic scenario as the baseline. 

The Q1 consensus of $1.20B to $1.24B wasn’t aggressive, it was simply what perfection looked like on a spreadsheet.

Personally, I think earnings could decline over the next few quarters before the business reaccelerates again later on. 

Crypto activity still matters too much to Robinhood today, and with the current environment, I wouldn’t be surprised to see weaker numbers near term before things eventually recover.

That creates a difficult setup for the stock. When the next major catalyst still depends heavily on a crypto recovery, the market usually compresses the multiple first and asks questions later.

A great business can still have a bad risk/reward. 

A position in HOOD at $30 had enormous asymmetry and at today’s valuation, the cone of outcomes is much narrower. 

Upside is capped by valuation, while downside is exposed every time crypto has a bad month.

I'm not trimming because Robinhood is broken.

I'm trimming because the market already figured out it's great, and right next door, there's a company the market still thinks is boring.

THE CAPITAL ALLOCATION ARGUMENT NOBODY WANTS TO HEAR (P2)

Larry Ellison built Oracle into a relational database empire starting in the 1970s.

For decades, the company was synonymous with expensive, inflexible, on-premise software that Fortune 500 IT departments were contractually obligated to keep paying for. 

The reputation was so sticky that it's still the first thing most analysts picture when they hear the name.

But that cognitive lag is the entire opportunity.

Think about what it takes for a market narrative to fully update.

An analyst who spent a decade covering Oracle as a slow-growth database incumbent doesn't flip their mental model overnight because of one earnings print.

Institutions that haven't owned Oracle in years don't rebuild positions in a single quarter. 

The re-rating from legacy software company to critical AI infrastructure happens slowly and that gap between what Oracle actually is and what the market still thinks it is? 

That's where the money gets made.

The Q3 FY2026 print that changed the conversation

Oracle's Q3 FY2026 earnings dropped on March 10, 2026. Here's the scorecard:

But the number that actually matters isn't any of those, it's the Remaining Performance Obligations of $553B.

That's contractually committed future revenue, already signed and on the books. 

Oracle's entire market cap is nearly equal to that backlog.

OpenAI alone represents an estimated $300B of it, roughly 60% with that single relationship potentially accounting for nearly 30% of Oracle's total revenue in the coming years.

Management also raised FY2027 revenue guidance to $90B, well above the $86.6B, analyst consensus. 

And they've already secured more than 10 gigawatts of power capacity coming online in the next three years to service the demand pipeline.

What Oracle actually built

Here's what Oracle quietly became while everyone was watching Nvidia:

The fundamental difference between Oracle and Robinhood as investments is the nature of the underlying demand. 

Oracle sells capacity, committed, contracted infrastructure that AI companies need to exist before they can build anything. 

Robinhood earns activity revenue, fees generated when retail traders choose to speculate. 

Infrastructure revenue locks in at contract signing, and activity revenue has to be re-earned every quarter based on how risk-hungry people feel.

Combined hyperscaler CapEx for 2026 is expected to hit $630B to $770B, with roughly 75% of that directed specifically toward AI infrastructure.

Goldman Sachs projects $1.15T in combined AI CapEx from 2025 to 2027, more than double what was spent from 2022 to 2024. 

That spending does not turn off because Bitcoin is down 40% or retail sentiment sours.

The honest risk assessment

Oracle's risks are real and worth staring at directly.

The company is carrying roughly $134.6B in debt, and free cash flow is currently negative. 

The OpenAI concentration, 60% of that $553B backlog from a single counterparty, is a legitimate concern.

If that relationship changes for any reason, the math gets complicated fast. 

And Larry Ellison's role is evolving, which introduces management transition risk that doesn't exist on paper but absolutely exists in practice.

But here's the counterpoint on timing, the negative free cash flow is a function of the buildout phase.

I am projecting a cash flow waterfall around FY2029 as the heavy CapEx transitions into cash generation from that $553B backlog. 

Why I'm doing this now

The market's mental models update slowly. 

Oracle's Q3 was arguably one of the best earnings prints in enterprise tech history, 84% IaaS growth, $553B RPO, five consecutive quarters of acceleration, and the stock still trades at a discount to what those fundamentals should imply.

That lag between a fundamental inflection and market re-rating is exactly where asymmetric returns get made.

Robinhood is an excellent company, and the market knows it's an excellent company, and it repriced 204% in 2025 to reflect that knowledge.

Oracle is an increasingly excellent company. The market still partially thinks it's a legacy software vendor that got its cloud act together.

But here's the real reason I'm making this move: I don't think this is fintech's moment.

The next two to three years are going to be defined by where the capital is flowing, and that capital isn't flowing toward consumer trading apps. 

It's flowing toward the companies providing the plumbing that makes the entire AI buildout actually work.

Fintech is a consumer sentiment business, and it moves with retail confidence, crypto cycles, and risk appetite. 

Those tailwinds will come back, they always do. 

But in a market where the dominant macro story is an unprecedented infrastructure buildout, the companies sitting closest to that spending get re-rated first and fastest. 

Oracle isn't adjacent to that wave but rather sitting directly in the middle of it.

Sometimes investing isn’t about finding a better company but rather about finding a better setup.

The scenario where this trade is wrong is worth naming honestly, crypto recovers sharply, Robinhood's prediction markets scale faster than anyone expects. 

Gold subscribers hit 10M, and the international expansion drives a second re-rating cycle. In that world, HOOD outperforms.

I'm not ruling it out, I'm just looking at where the gravitational pull of capital is strongest right now.

And if I'm wrong about Oracle and right about Robinhood? I give up some alpha.

This isn't a prediction that HOOD falls but rather a bet that ORCL re-rates faster from here.

Buffett wasn't saying the dot-coms were bad companies, he was just asking a different question than everyone else in the room.

Alright, that's it for this edition of Milk Road AI. We want to hear from you.

BITE-SIZED COOKIES FOR THE ROAD 🍪

Anthropic sealed a deal with SpaceX, gaining 300 MW and 220K Nvidia GPUs instantly. Claude just got access to one of the world's largest GPU clusters, the compute race is heating up.

OpenAI launched GPT-Realtime-2, the first voice model with full GPT-5-class reasoning. It also translates 70+ languages live and transcribes speech in real-time as you speak.

Google I/O 2026 will unveil Gemini 4, AI glasses, and Gemini inside Boston Dynamics robots. Google is making its biggest AI push yet, across models, hardware, and robotics at once.

MILKY MEMES 🤣

ROADIE REVIEW OF THE DAY 🥛