Everyone in the neocloud space is fighting over the same thing: GPUs.
That’s the game right now. Buy up the latest NVIDIA chips. Deploy them as fast as possible. Sell access to AI labs scrambling for compute. Rinse and repeat.
But AI isn’t just about GPUs.
Not really.
Training massive models is one thing. Running them efficiently at scale is another. Because the real challenge isn’t just having the hardware. It’s integrating compute, storage, networking, and AI services into one seamless platform. And then securing data centre space for deployment.
The companies that solve that puzzle will win.
That’s exactly what this company is building.
It doesn’t just deploy compute. It owns data centres. Designs its own servers. Builds its own software. Runs its own AI research team. Abstracts away the complexity. It’s stacking every layer of AI infrastructure into a single, optimised system.
And it’s got $2.45 billion in cash.
Welcome to Nebius.
The GPU Audio Companion Issue #20
Want the GPU breakdown without the reading? The Audio Companion does it for you—but only if you’re subscribed. Fix that here.
Company Background
Nebius isn’t a typical startup.
It didn’t launch from a garage. It wasn’t built from scratch. It came from something much bigger.
For over a decade, Yandex N.V. was one of the most powerful tech companies in Russia.
A search giant, a cloud provider, and an AI powerhouse.
But when Russia’s geopolitical landscape changed, Yandex had to restructure. That meant selling off all its Russian assets and leaving behind only the international parts of the business. That process took two years.
When it was over, the cloud platform, data centres, and AI tools were rebranded under a new name.
The curious case of Nebius, the publicly traded AI infrastructure ‘startup’
— #TechCrunch (#@TechCrunch)
1:02 PM • Nov 24, 2024
Officially headquartered in Amsterdam, Nebius emerged from the restructuring with serious firepower:
A fully owned, high-performance data centre in Finland, housing 14,000 GPUs with room for expansion.
A full-stack AI cloud built for large-scale model training, fine-tuning, and inference.
In-house server and hardware design, giving it an edge on cost and efficiency.
A Nasdaq listing and a fresh $700 million capital raise from top-tier investors.
$2.45 billion in cash, no debt, and the flexibility to scale aggressively.
But Nebius isn’t just building an AI cloud.
It’s running a multi-pronged AI strategy.
Alongside its core infrastructure, it owns:
Toloka AI: A leader in AI data annotation, competing with Scale AI and Surge AI.
Avride: Its autonomous vehicle division, developing self-driving cars and delivery robots (now in a multi-year partnership with Uber).
TripleTen: A coding bootcamp helping people transition into tech, with a focus on the US and Latin America.
ClickHouse: Nebius retains a 28% stake in the open-source analytics database, valued at $2 billion.
From cloud computing to AI training to self-driving cars, Nebius is operating across multiple frontiers. The goal? Becoming the AI cloud that scales better, runs cheaper, and integrates deeper than the competition.
Now, it has to prove it can execute.
Executive Team
Nebius is led by experienced cloud operators, AI specialists, and business strategists:
Arkady Volozh (Co-founder & CEO): Founded Yandex, built its AI cloud division, and now leads Nebius’ global AI infrastructure push.
Roman Chernin (Co-Founder & Chief Business Officer): Drives Nebius’ commercial strategy and expansion.
Danila Shtan (Chief Technology Officer): Leads cloud infrastructure, platform engineering, and AI research.
Danila Pavlov, CFA (Chief Financial Officer): Manages financial strategy and capital allocation.
Daniel Bounds (Chief Marketing Officer): Oversees global market positioning and Nebius’ AI cloud brand.
The Edge
Nebius isn’t just another GPU cloud. Its competitive advantage comes from full-stack AI specialisation:
AI-Native Cloud Platform: Not just GPUs. Nebius runs a full AI infrastructure stack, covering model training, fine-tuning, inference, and data processing.
High-Performance Data Centres: Most neoclouds rely solely on colocation. Nebius operates its own high-efficiency data centre in Finland and works with partners like Patmos in the US, giving it greater control over costs and operations.
In-House Hardware & Software: Nebius designs its own servers, racks, and cloud architecture, integrating every layer for maximum efficiency.
AI Research & LLM Training: Nebius runs its own AI research division, using its infrastructure to train LLMs and optimise AI workflows.
Publicly Traded & Well-Funded: With a Nasdaq listing, $2.45 billion in cash, and no debt, Nebius can scale aggressively without waiting for VC funding.
Recent Moves
$2.45 Billion Cash on Hand: As of December 31, 2024, Nebius holds $2.45 billion in cash and cash equivalents, giving it one of the strongest balance sheets in the AI cloud sector.
$700 Million Capital Raise: In December 2024, Nebius raised $700 million from NVIDIA, Accel, and Orbis, securing long-term financial firepower.
Nasdaq Trading Resumed: After a two-year halt due to the Yandex restructuring, Nebius officially returned to public markets in October 2024.
Kansas City Expansion: Announced plans for a major US GPU cluster, starting in Kansas City, with additional customer hubs in San Francisco and Dallas.
Uber Partnership (Avride): Avride, Nebius’ self-driving division, signed a multi-year deal with Uber to deploy autonomous delivery robots and self-driving cars.
Avride launches sidewalk delivery bots on Uber Eats in Jersey City
— #TechCrunch (#@TechCrunch)
1:03 PM • Feb 13, 2025
Check out the Nebius Newsroom for the latest updates:
What’s Next?
With $2.45 billion in cash and a massive GPU deployment roadmap, 2025 is all about locking in long-term dominance.
Plans are in the works to triple the Finland data centre capacity. The Kansas City cluster comes online in Q1 2025, marking Nebius’ first major US deployment. Starting with thousands of NVIDIA Hopper GPUs, the facility has room to scale to 40MW. Meanwhile, the Paris cluster, already up and running with H200s, strengthens Nebius’ European presence. Then comes Blackwell.
Nebius is going all-in on NVIDIA’s next-generation architecture, with 22,000 Blackwell GPUs set to deploy across the US and Finland this year.
These chips will power the next generation of AI workloads, and Nebius is moving early to secure its place in the market.
Beyond training, AI Studio is positioning Nebius for the next wave of inference-driven AI adoption. Tracto.ai, a new serverless compute platform, extends this further by giving AI companies a full-stack environment to train, fine-tune, and deploy. Avride is scaling up with Uber and Grubhub partnerships, deploying delivery robots on US college campuses and city streets. Toloka AI continues landing top GenAI clients, while TripleTen’s edtech bootcamp is expanding across the US and Latin America.
Nebius is already seeing the impact.
March ARR is locked at $220 million, and with more GPUs, more data centres, and more enterprise deals, it’s aiming for $750 million to $1 billion ARR by December 2025.
The capital is there. The infrastructure is coming online.
Now, it’s about execution and securing the next wave of AI customers before the competition catches up.
Keep The GPU Sharp and Independent
Good analysis isn’t free. And bad analysis? That’s usually paid for.
I want The GPU to stay sharp, independent, and free from corporate fluff. That means digging deeper, asking harder questions, and breaking down the world of GPU compute without a filter.
If you’ve found value in The GPU, consider upgrading to a paid subscription or supporting it below:
☕ Buy Me A Coffee → https://buymeacoffee.com/bbaldieri
It helps keep this newsletter unfiltered and worth reading.
