GM. This is Milk Road AI, where we track who’s funding, building, and owning the AI stack.
Here’s what we’ve got for you today:
- ✍️ How Microsoft is quietly reshaping the AI race.
- 🎙️ The Milk Road AI Show: 6 Winning AI Stocks Beyond Nvidia and Big Tech w/ Ronnie.
- 🍪 Nvidia deepens AI push with CoreWeave.
Yield Seeker gives you a personalized AI Agent that goes out and earns yield on your stablecoins. Hurry & deposit on Yieldseeker now using the “milkroad” referral code!
Prices as of 10:00 a.m. ET.
INSIDE MICROSOFT’S SECRET PLAN TO DOMINATE AI
Do you remember the biggest scandal in music history?
In 1989, Milli Vanilli won the Grammy for Best New Artist.
They were cool. They had the dreadlocks. They had the dance moves.
But they had one tiny problem:
They weren’t actually singing.
They were just looking pretty while someone else hit the high notes backstage.
For the last three years, Microsoft has been the Milli Vanilli of tech.
They have been dancing on stage, selling billions in Copilot subscriptions, and accepting the applause.
But the voice coming out of the speakers? That was OpenAI.
Microsoft was just holding the microphone for Sam Altman.
But before we get to the drama (and oh boy, is there drama), we have to answer a simple question:
What the heck does Microsoft even do? Aside from selling Windows to the three people left on Earth who don't own a MacBook?
Well, they basically have three massive buckets that print money:
- Server & cloud services now make up 35% of revenue. This is the server farm that runs the internet, if it breaks, we revert to the Stone Age.
- Commercial Office 365 is the second biggest slice at 31%. If you want a corporate job in 2026, you basically have to pay Satya Nadella rent.
- Gaming ($23.5B) now makes more than Windows ($17.3B).
That is a lot of cash, but if you want to know where the stock price is going, you can ignore all of it.
The entire company’s future is currently hinging on a single division:
Azure (the cloud infrastructure that basically runs the internet).
But in 2025, the landlord realized they didn't hold the keys to the penthouse, a very erratic tenant did.
The AI power struggle goes public
If you were watching Microsoft in 2025, you were watching a celebrity marriage implode in slow motion.
For a while, the "Cheat Code" strategy worked:
Microsoft invested $1B in 2019, got exclusive rights to OpenAI, and looked like a genius.
But by mid-2025, the relationship started to rot.
Sam Altman and OpenAI realized they were the ones with the talent, and Microsoft was just the bank.
OpenAI realized it was the "bottleneck" and started negotiating harder for control over pricing and IP.
Then came the infidelity:
- In March 2025, OpenAI announced a $22.4B deal with CoreWeave , signaling they wanted to cheat on Microsoft’s cloud.
- By June, they were reportedly considering accusing Microsoft of "anticompetitive behavior".
The market realized a terrifying truth:
If OpenAI walked away, Microsoft’s entire Copilot strategy, the thing they bet the farm on, would evaporate overnight.
It looked like one of the most valuable companies in the world was about to be dumped by its cooler, younger boyfriend.
But wait, Satya Nadella, CEO of Microsoft, didn't let them walk.
In October 2025, he forced a "renewal of vows" that made history.
He signed a $250B multi-year deal to lock OpenAI down.
It was the ultimate "Prenup" and the deal structure is pure genius:
- Microsoft traded "cloud exclusivity" for a guaranteed $250B in Azure spending.
- They secured a real 27% equity stake worth ~$135B.
- Microsoft now guarantees access to all models, including future AGI, through 2032.
Wait, why does Microsoft even want this deal? Because it buys them exactly 7 years to build their own brain.
They aren't trying to ride OpenAI forever. They are buying time to learn how to sing.
But in Silicon Valley, seven years is a lifetime.
The landlord isn't waiting for the lease to expire. Microsoft is already remodeling the building to kick the tenant out.
Just wait until you hear about what they bought a year ago.
GROW YOUR STABLECOINS ON AUTOPILOT WITH AI
Our boss said he’d double our salary if we could automate 73% of our work.
Naturally, I thought: “Can I automate earning yield on crypto?”
That’s where Yieldseeker comes in. It gives you a personal AI agent that hunts the best stablecoin yields 24/7.
How it works:
- Sign up & get your personal AI Agent
- Deposit as little as $10 USDC on Base
- Your AI agent handles the rest
Due to popular demand, we’re running the previous campaign again:
The first 100 Milk Road users to deposit $100+ on Yieldseeker get 10,000 bonus points.
(oh also, their platform just got a full rebrand)
Hurry & deposit on Yieldseeker now using the “milkroad” referral code!
INSIDE MICROSOFT’S SECRET PLAN TO DOMINATE AI (P2)
Back in 2024, Microsoft realized they had a problem:
OpenAI had all the "IQ" (raw intelligence), but Microsoft had zero "EQ" (emotional intelligence).
Let's face it, Microsoft’s historical idea of 'empathy' was Clippy popping up to ask: 'It looks like you're having a mental breakdown.
Would you like help formatting that?'
So, they targeted Inflection AI, a startup run by DeepMind co-founder Mustafa Suleyman.
Inflection AI built something called "Pi", an AI designed to be supportive, empathetic, and conversational.
They were building a machine that felt human.
But instead of buying the company (which would trigger an immediate FTC war), Microsoft executed a "reverse acqui-hire" playbook:
- Microsoft paid exactly $650M, split into $620M for "non-exclusive licensing" and a $30M payment to waive legal claims.
- Because they didn't technically "buy" the company, they argued they didn't need to file standard merger notifications. Inflection "technically" still exists as an independent business, just without its staff.
- They took Mustafa Suleyman, co-founder Karén Simonyan, and most of the 70-person engineering team to launch the new "Microsoft AI" division.
Microsoft gutted a competitor, absorbed their top engineers directly into HQ, and finally got what they really paid for:
A world-class product team that answers to Satya Nadella, not Sam Altman.
Microsoft’s plan to lower the AI tax
With the brain (Inflection team) already secured, the landlord looked at his biggest expense report, grabbed a red pen, and declared war on his hardware suppliers.
For years, Microsoft was addicted to NVIDIA GPUs.
Every time you used Copilot, Microsoft had to pay a "tax" to Jensen Huang.
But this week, they officially unveiled the Maia 200:
- It packs 140B transistors on a bleeding-edge 3nm process.
- It delivers 3x the throughput of Amazon’s latest Trainium chip.
- Crucially, it beats their previous hardware setup in cost-efficiency by 30%.
But here is the detail most people missed: This chip is optimized for inference (running the model), not training.
Why? Because OpenAI burned roughly $5B on inference in just six months.
By moving those workloads to Maia, Microsoft stops bleeding margin to NVIDIA immediately.
Does this mean Jensen Huang should trade his leather jacket for a cardigan? Is NVIDIA finished? Absolutely not.
While Microsoft is using Maia to drive the car (inference), they still need NVIDIA to build the car (training).
Training a model like GPT-5 requires raw, unadulterated power and flexibility.
NVIDIA's H100s and Blackwell chips are still the undisputed kings of this hill.
Microsoft is still buying thousands of them to train the next generation of intelligence.
Microsoft is just one customer (albeit a huge one).
Meta, xAI, Google, and every sovereign nation on Earth are still queuing up to buy as much NVIDIA compute as they can get their hands on.
But here is the irony: by making AI cheaper to run (via Maia), Microsoft will likely increase the total usage of AI.
More usage means more demand for better models, which means more demand for NVIDIA training clusters.
Maia fixes Microsoft's margin problem. It doesn't fix the world's hunger for intelligence.
The final verdict
Microsoft reports earnings today after the bell.
And while the headlines may obsess over the CapEx, the real story is the moat.
The thesis is simple: Microsoft is deliberately depressing its own margins today to build an insurmountable infrastructure lead for tomorrow.
Here are the only three numbers that matter for this earnings print:
- The "Intelligent Cloud" segment is projected to hit $32.5B in revenue this quarter alone.
- Commercial bookings surged 111% year-over-year. That is a doubling of future contracts in a single year.
- Management is guiding for roughly $80B in revenue for the quarter. That is 15% growth for a company worth $3.5T.
So, is the stock a buy? I need to see the acceleration.
If they spend $35B on CapEx but Azure growth stays flat, that is a red flag.
It means the "landlord" is building empty apartments with no tenants or cash flow.
However, what if that massive spend converts into Azure growth accelerating past 39%?
Then the thesis is confirmed.
It means the demand is real, the new Maia chips are helping the margins, and the "Milli Vanilli" era is officially over.
Just look at the scoreboard.
While the media talks about "competition", the gap is actually widening.
Amazon (#1 in market share) is sitting at a 17% growth. Google Cloud is hustling at 32%.
But Microsoft (#2)? They are leaving them in the dust at 39%.
And say what you want about Microsoft and their ability to execute in AI, but betting against Satya Nadella is historically a bad idea.
He is arguably one of the best CEOs in the world.
Don't believe me? Just look at the chart.
Since taking over in 2014, he hasn't just grown the stock; he resurrected it.
I would not count out Microsoft in 2026.
Satya Nadella has pulled off the greatest save in corporate history.
He rented OpenAI's talent just long enough to build his own infrastructure.
They might not have the best voice in the band anymore, but they own the stadium.
And if growth accelerates today? You buy the stadium.
Alright, that’s it for this edition. Catch you on the next one.
But before you go, I want your honest opinion on this newsletter overall so far. 👇️
What'd you think of today's edition?
I’ll be replying to every single comment, yes, even the ones that hurt my feelings a little.
I’ll just be over here stress-eating cookies for emotional support.
6 AI STOCKS THAT FLY 🚀
In today’s episode, we sat down with Ronnie V to talk about AI driven stock opportunities beyond Nvidia and Big Tech.
Here’s what you’ll hear:
- Why AI demand is pushing power, grid, and infrastructure plays like Fluence Energy.
- How advertising is shifting from SEO to GEO with Zeta Global and what that means for marketers.
- Ronnie’s framework for spotting durable AI businesses using margins and free cash flow.
- His favorite sub $10B names, including DLocal as a 12 month pick and Mobileye as a long term bet.
Don’t sleep on this one 👇
YouTube | Spotify | Apple Podcasts
Bridge helps businesses send, store, accept, and launch stablecoins instantly. Learn how Bridge is powering stablecoin-backed money movement.
Mercuryo is the simple way to buy, sell, and spend crypto with low fees (yes, even with Apple Pay). Discover an easier way to use crypto.
Bitfinex is the first major exchange to remove trading fees entirely. Trade on Bitfinex for free now.
BITE-SIZED COOKIES FOR THE ROAD 🍪
Nvidia is investing $2B in CoreWeave to expand AI data center capacity. The deal will power new AI factories with Nvidia’s latest chips.
Moonshot AI released Kimi K2.5, an open-source multimodal model strong at coding. It also launched Kimi Code to rival tools like Claude Code.
Nvidia launched AI weather models that predict storms faster and more accurately. Its Earth-2 system outperforms Google’s GenCast on most metrics.
MILKY MEMES 🤣
ROADIE REVIEW OF THE DAY 🥛
