GM, this is Milk Road AI, where we find the picks and shovels behind the AI gold rush.
Hereβs what weβve got for you today:
- βοΈ The AI picks-and-shovels play of the decade?
- ποΈ The Milk Road AI Show: The AI Model With An Unfair Advantage Over ChatGPT.
- πͺ Anthropic just became more valuable than OpenAI.
Kalshi is a regulated prediction market where you can trade on real-world events. Special offer for Milk Road Readers: Trade $10 on Kalshi β Claim $10 for free.
Prices as of 10:00 a.m. ET.
THE COMPANY RISING FROM THE ASHES
In 1997, a Soviet-trained mathematician named Arkady Volozh built a search engine in Moscow and nobody thought much of it at the time.
Google was still a PhD project in a Stanford dorm room, and the internet barely existed in Russia.
The idea that anyone would search for anything online was, in most parts of the world, still a novelty.
But Volozh had spent years studying how language worked, real language, with all its grammar quirks and Russian morphology and the ten different ways to conjugate a single verb.
He understood something that most Western tech companies were just beginning to figure out: that search wasn't about matching keywords, but rather about understanding what someone actually meant.
So he built Yandex.
Over the next 25 years, he turned it into Russia's answer to Google.
Yandex became the dominant search engine with 60%+ market share, and it even built ride-hailing, navigation, cloud services, e-commerce, and one of the most respected AI research labs in the world.
In 2011, it IPO'd on Nasdaq and raised $1.3B, one of the biggest tech IPOs of the year.
At its peak, the company was worth over $30B but then, on February 24, 2022, Russia invaded Ukraine, and everything fell apart overnight.
Within weeks, Yandex's Nasdaq shares were suspended from trading.
Volozh, who had moved to Israel years earlier, was placed on EU sanctions lists and forced to step down as CEO, and the company he had spent a quarter century building was suddenly a geopolitical grenade.
What followed was two years of one of the most complicated corporate restructurings in modern history.
In July 2024, Yandex sold every Russian asset to a consortium of local investors for $5.4B, a steep discount, but the price of walking away clean.
What was left was the Amsterdam parent company, a Finnish data center, a GPU cluster in Paris, over a thousand elite engineers, and a founder who had just watched his life's work get dismantled and was not done yet.
They renamed the company Nebius Group, changed the Nasdaq ticker from YNDX to NBIS, and relisted in October 2024.
And then Arkady Volozh pointed everything he had left at the AI infrastructure boom.
What Nebius actually is
Here's the simple version.
Every time an AI model does anything, answers a question, writes code, generates an image, or powers a drug discovery platform, it burns through enormous amounts of computing power.
That compute lives in giant warehouses filled with NVIDIA's most powerful chips, running around the clock and consuming electricity at rates that would power small cities.
Nebius owns those warehouses and rents them to the companies building AI.
They're what the industry calls a neocloud, a cloud provider built specifically for the AI era rather than retrofitted for it.
Amazon, Microsoft, and Google built their cloud businesses in the 2000s for websites, email, and file storage.
Nebius was built from day one for massive GPU clusters and AI workloads and that distinction matters more than it sounds.
According to Synergy Research, revenue across AI-focused cloud, SaaS, and data center providers is expected to grow at roughly 58% annually through 2031, creating a market worth hundreds of billions of dollars.
The problem isn't finding customers, it's building fast enough to serve them.
Which brings us to a seemingly strange question.
Microsoft owns Azure, Meta can spend tens of billions of dollars building data centers.
So why are two of the most powerful technology companies on earth paying a relatively young neocloud billions of dollars? Because they can't build fast enough.
AI demand is growing faster than new data center capacity can come online, building facilities takes years, securing power can take even longer, and companies training frontier AI models can't afford to wait.
They need dedicated GPU clusters now, not five years from now.
That's what Nebius sells, and the line of customers is growing faster than the company can build.
The global appetite for AI compute has become so intense that even the largest technology companies in history are running short on the physical infrastructure needed to train and run their models.
So they outsource the overflow to specialists.
In September, Microsoft signed a five-year deal with Nebius worth up to $19.4B.
Two months later, Meta followed with a $3B agreement.
Then, in March 2026, Meta came back for more, expanding the partnership to as much as $27B and making Nebius one of the first large-scale deployments of NVIDIA's next-generation Vera Rubin platform.
That expanded Meta agreement alone was worth more than Nebius's entire market value just months earlier.
Then NVIDIA joined the party, investing $2B directly into the company, Jensen Huang even called Nebius "the AI cloud built for the agentic era."
When the company selling the picks is buying equity in the shovel store, it's usually worth paying attention.
CRYPTOβS THIRD PMF MOMENT IS HERE
PMF = Product Market Fit
Until now, crypto has really had two major PMF moments: 24/7 trading and stablecoins.
Now a third one may be emerging with Kalshi: Prediction Markets.
Kalshi is a regulated prediction market exchange where people trade on the outcomes of real-world events (elections, inflation, sports, crypto prices and more).
And their growth has been massive:
- Recently raised $1B at a $22B valuation
- Institutional trading volume is up 800% in six months.
- Annualized volume is somewhere around $178B.
Kalshi is giving Milk Road readers $10 free when you trade $10.
Trade $10 β Claim $10 on Kalshi.
THE COMPANY RISING FROM THE ASHES (P2)
Alright. The backstory is wild, the contracts are enormous, and the customers are basically the Avengers of global tech.
So what does the actual investment case look like? Let's go to the numbers.
In Q1 2026, Nebius posted $399M in revenue, that's a 684% increase year-over-year.
This is a company that entered 2024 with roughly $120M annualized revenue, and by the end of 2026, management is guiding for $3 to $3.4B.
That's roughly 6x growth in a single calendar year at a company that is already generating nine figures in quarterly revenue.
And the growth doesn't stop there.
Current projections imply Nebius could generate roughly $35B in revenue between 2026 and 2028, driven by Microsoft's massive Azure AI agreements, Meta's expanding AI training commitments, and NVIDIA's strategic partnership and capital support.
Their balance sheet looks completely different from most AI infrastructure plays.
Nebius holds $9.3B in cash while CoreWeave, its most direct comparable, carries $7.5B in debt.
That difference matters enormously when you're trying to build gigawatt-scale data centers in a world where construction costs keep going up, and capital markets can get weird fast.
One person who appears to agree is former OpenAI researcher Leopold Aschenbrenner.
His fund recently disclosed a $2.6B position in Nebius, making it his largest holding.
For someone whose entire AI thesis is built around owning compute infrastructure, that's a pretty loud vote of confidence.
The honest risks
The biggest risk is customer concentration.
Microsoft and Meta collectively represent the vast majority of Nebius's contracted revenue.
If either one delays capacity deployment, renegotiates terms, or decides to build its own data centers fast enough to stop needing the overflow, the revenue trajectory breaks hard and fast.
The second risk is execution.
Nebius is simultaneously constructing multiple gigawatt-scale facilities across three continents.
Construction delays, permitting bottlenecks, transformer shortages, power interconnection timelines, any one of these can push 2026 revenue into 2027 and compress a multiple that has no room for disappointment built into it.
Still, it's hard not to appreciate what Volozh has built.
He turned Yandex from a small Russian search engine into a $30B technology giant, watched geopolitics tear it apart, sold the pieces he had to, and then redirected everything that remained toward the fastest-growing market on the planet.
Volozh is all in, Nebius represents roughly 90% of his net worth, controls tens of billions in signed contracts, and has billions of dollars available to fund its expansion.
The entire investment thesis comes down to one simple question: Will the world need more AI compute tomorrow than it does today?
That's the same bet former OpenAI researcher Leopold Aschenbrenner is making.
And it's also a bet I've been making.
While most investors were focused on the geopolitical risks and volatility across AI infrastructure names, I was making Nebius my largest position and dollar-cost averaging along the way.
Milk Road PRO members have been following the thesis from the beginning and watching the position evolve in real time as my conviction grew.
Today, that position is up more than 100%.
The crazy part? It's not even my biggest winner from the past three months.
That's what happens when you find great businesses before the rest of the market catches on.
If you want to see what I'm buying next, come join us inside Milk Road PRO.
Alright, that's it for this edition of Milk Road AI. We want to hear from you.
NBIS, are you in?
- Bullish: The picks and shovels always win.
- Skeptical: Too much execution risk at 12x revenue.
- Watching: Interesting story, not touching it until they prove the build.
AN UNFAIR ADVANTAGE OVER CHATGPT π€
In today's episode, we sat down with Yan Liberman, Managing Partner at Delphi Ventures, to talk about the intersection of crypto and AI and why some of these projects may be mispriced by the market.
Here's what you'll hear:
- What makes Venice different: an LLM app with anonymous, TEE, and end-to-end encrypted privacy tiers.
- The growth story behind Venice's run, from new Pro features and marketing to rising subscription revenue.
- Why Yan argues Venice is undervalued, with a tight token supply and falling emissions vs other projects.
- Grass, the opt-in network selling web data to AI labs to power model retraining and agentic flows.
Hit play and dig in π
YouTube | Spotify | Apple Podcasts
Real Finance Blockchain is an EVM-compatible L1 that is built specifically for RWA tokenization. Read more about Real Finance Blockchain here.
Arc is an open Layer-1 blockchain built for real-world financial applications. Public testnet is live, App Kits are here, and mainnet is getting closer. Start building on Arc.
Your carrier is selling your data. Use code MILKROAD for 33% off your first 6 months.
BITE-SIZED COOKIES FOR THE ROAD πͺ
Special offer for Milk Road readers. Kalshi is giving Milk Road readers $10 free when you trade $10 on their platform.*
Anthropic raised $65B at a $965B valuation, surpassing OpenAI as the most valuable. Claude's maker is now near $1T and an IPO could be right around the corner.
NVIDIA pledges $150B per year in Taiwan and breaks ground on a new campus. Jensen calls Taiwan "the epicenter of the AI revolution" and is putting money on it.
Huawei unveiled the Tau Scaling Law, a potential successor to Moore's Law. It shifts chip progress from shrinking transistors to cutting signal delay time.
*this is sponsored content.
MILKY MEMES π€£
ROADIE REVIEW OF THE DAY π₯
