Hello fellow investors.
How's the market treating you so far in 2025?
Well, the S&P500 is still negative for the year, but it looks like Trump maybe easing up on tariffs, so that may bring us into positive territory soon. We'll see.
But if you've been following along here and investing like I have, you are likely doing far better than 95% of investors out there, even in a rough year like this. My personal YTD return is over 20%, I am VERY proud of this specifically in a year where everybody has been struggling and losing money.
The best hedge funds in Manhattan tend to unofficially promise returns between 10-20% in a year, so for me to consistently beat them puts a huge grin on my face.
In the last year, I am up over 100% in my portfolio. I spend a lot of time reading market news, researching companies, and devote a lot of time to this. But I genuinely love doing this and it doesn't even feel like work to me, so I am very happy to be at least rewarded financially for all of this.
Now, that being said a lot of this is due to my high conviction in PLTR which I have been promoting since 2021 here, but
my very recent pick, HIMS has also taken off like I expected.
So naturally, some of you are already asking me what's next. I've been working on this one for a while, but here is my latest investment:
It’s called Nebius, ticker symbol $NBIS, and it just might be the best kept secret in AI.
So What Is Nebius?
$NBIS is a next-gen cloud provider. Not just your typical storage and compute shop either. Nebius builds highly specialized AI infrastructure, offers developer tools, and runs one of the most impressive hardware-software stacks I’ve seen outside of hyperscalers. Think AWS and Azure, but streamlined for AI workloads and far more cost efficient.
The company is founder-led. It carries no debt, holds 2.4 billion dollars in cash, and is quietly building a monster. They own three high-potential subsidiaries and a meaningful stake in ClickHouse. And somehow, they’re still trading at just a 6.6 billion dollar valuation.
The core business alone could 4x or more from here. And that’s before even accounting for the other units.
The Big Picture
Nebius is gunning for dominance in the AI infrastructure race. The total addressable market for AI cloud is expected to grow at a 35 percent CAGR through 2030. Everyone wants compute power, and Nebius is offering some of the most cost-effective, high-performance solutions in the game.
They operate two key platforms:
-
Nebius AI Cloud — optimized for AI training and inference
-
Nebius AI Studio — a platform for researchers and developers to fine-tune, test, and deploy models
And now they’ve introduced TractoAI, a serverless platform that makes deploying AI workloads even easier, priced by usage instead of cluster time.
This isn't just a "cool tech" company. This is an extremely well-funded, engineering-heavy operator that designs its own servers, racks, and even motherboards. They run world-class data centers in Finland, the US, Paris, and Iceland. Their Finland site ranks among the top 20 supercomputers globally.
Let that sink in.
Why It Stands Out
A few things make Nebius unique:
-
They’re a preferred Nvidia partner, meaning they get early access to top GPUs like Blackwell. That’s a massive edge in the AI compute wars.
-
Their infrastructure is 20 to 25 percent cheaper to operate compared to others.
-
They’ve got over 15 years of experience in server optimization.
-
They’re expanding hard. By the end of 2025, Nebius expects to have 100 megawatts of capacity in the US and over 400 megawatts globally.
Let’s talk about scale for a second. At full capacity, their Finland data center alone could generate 1 billion dollars in annual recurring revenue. Add Missouri, Iceland, and Paris, and you get a path to 2.5 billion dollars ARR by mid-2026.
Apply even a modest 9 times ARR valuation, and that core business alone could be worth 22.5 billion dollars.
But that’s just part one.
Now Enter the Subsidiaries
Avride — The Sleeper in Autonomy
I was shocked this company wasn’t getting more attention.
Avride is Nebius' autonomous vehicle arm. Think robotaxis and delivery robots. They were the first to launch robotaxis in Europe and have already driven 22 million kilometers. That’s more than Cruise, Zoox, or Motional in some cases. Zero accidents.
They’re already operating in Moscow and testing in the US and South Korea. Starting this year, they’ll be running paid rides in Dallas through Uber. They’ve partnered with Hyundai to move toward full autonomy.
This is serious. Cruise raised money at a 30 billion dollar valuation. Zoox sits around 6 billion. Motional, which hasn’t even tested highway driving, was last valued at over 4 billion. Avride is at least on par, or maybe even further ahead.
They also have autonomous delivery bots already deployed in the US, UAE, and Korea. Over 200,000 deliveries completed. Uber Eats integration. They expect to scale to 1,000 units by the end of the year.
Serve Robotics, which is doing something similar, is valued at 500 million with half the deployment. So conservatively, the delivery division alone should be worth a few hundred million. The robotaxi unit? Could be 6 to 10 billion by itself once it scales.
Toloka — Data Is the New Oil
Toloka is one of the world’s top AI data labeling platforms. Clients include Microsoft, AMD, Amazon, and more.
They’ve evolved from basic crowdsourcing to offering high-quality, targeted data for training large language models. This is one of the biggest bottlenecks in AI development right now, and Toloka is well positioned to dominate.
2024 revenue is projected at 60 million dollars, growing at over 40 percent annually. That puts 2030 revenue at around 322 million. At just 5 times sales, you’re looking at a 1.6 billion dollar valuation.
Again, that’s conservative.
TripleTen — Educating the AI Workforce
TripleTen is a fast-growing online bootcamp focused on AI and software training. Over 1,000 students enroll monthly, and they’ve recently expanded into B2B training for companies looking to reskill employees in AI.
They expect 50 million dollars in revenue by 2025 and are growing rapidly. The educational tech market is projected to grow at over 26 percent annually. If TripleTen hits 160 million dollars in revenue by 2030 and trades at 2 times sales (like Coursera), that’s another 320 million dollars in value.
ClickHouse Stake
Nebius owns 28 percent of ClickHouse, an open-source columnar database built for fast analytical queries. It’s used by giants like Microsoft, Spotify, Lyft, Block, Meta, and more.
ClickHouse was valued at 2 billion dollars in 2021. Since then, it’s only expanded. That makes Nebius’ stake worth around 560 million dollars, minimum.
Valuation Summary
So what do we have here?
-
Core cloud business could be worth 22 to 27 billion by 2026 to 2028
-
Avride could reach 6 to 10 billion in value
-
Toloka could be worth 1.6 billion by 2030
-
TripleTen around 320 million
-
ClickHouse stake valued around 560 million
And all of that... is bundled into a company that today is valued at just 6 billion dollars.
Yes, There Will Be Dilution
Let’s be real. The cloud business is CapEx intensive. Avride will need funding. Nebius raised 700 million in December and has 235 million shares outstanding, plus 126 million shares held in treasury for incentive plans and future financing.
Even if total dilution reaches 53 percent (I think that’s the ceiling), the upside is still massive.
Where This Could Go
I believe $NBIS could be worth 10 billion dollars by the end of this year. They’re expected to hit 750 million dollars in ARR and are on track to break even on adjusted EBITDA.
By 2026? The company could conservatively be worth 35 to 40 billion.
By 2030? I wouldn’t be surprised to see a 50 billion dollar valuation, maybe more, depending on how AI compute demand evolves and how Avride scales.
Even if you assume full dilution, you’re still looking at a 5 to 10x return over five to seven years.
At sub $30, I find this to be a bargain. I think this can reach $50 within this calendar year.
Final Thoughts
$NBIS is my highest conviction pick right now after PLTR & HIMS.
It’s got strong fundamentals, massive growth, an unbeatable TAM, and one of the most talented engineering teams in AI. They’re building everything from the ground up, solving multiple bottlenecks in AI: compute, data, and talent all under one umbrella.
This is the kind of company that becomes obvious only in hindsight. But I’m not waiting for hindsight.
We'll see.