Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx
$NBIS JUST IN: Famed Billionaire Ron Baron’s Fund Baron Capital Group initiated a position in Nebius in Q1
From Baron Capital:
“Nebius is a Big Idea with a remarkable origin story – the company was born from Yandex, popularly known as the “Google of Russia,” which founder Arkady Volozh built into a $30 billion business over 25 years with leading positions in online search, e-commerce, ride-hailing, music streaming, maps, and cloud services. After Russia’s invasion of Ukraine, Volozh divested all Russian assets in the largest corporate exit from Russia in history ($5.4 billion), reconstituting the company as Amsterdam based, Nebius. The company boasts a world-class 1,300-strong team of engineers with decades of experience building large-scale computing systems and a vision to build a leading AI cloud business from first principles – a purpose-built vertically-integrated hardware and software stack optimized for AI workloads.
Nebius’ long-term vision requires significant resources to build the physical infrastructure and acquire customers. In the interim, Nebius is strategically and very selectively signing bare-metal GPU deals (renting the data center with the GPUs installed but no software on top of the GPUs) with Microsoft (up to $19 billion) and Meta (up to $27 billion). While there is a range of outcomes on the long-term value of GPUs in a bare-metal model (with the main concern revolving around the rapid depreciation of old GPUs as new more efficient ones are introduced), the useful life of AI accelerators appears to be longer than previously anticipated, as the economic output of GPUs (as measured by token throughput) has been increasing over time as models have improved, meaning project returns get better with age. Dylan Patel of SemiAnalysis explained the dynamic on the Dwarkesh Podcast (March 2026)14: GPT-5.4 from OpenAI generates more tokens on H100s than GPT-4 did, despite being a far more capable model, because newer architectures (such as sparse mixture-of-experts) are more computationally efficient per token. The implication is that an H100 is worth more today than when it was purchased three years ago. For Nebius, this means the residual value of its GPU fleet after long-term contracts expire could be substantially higher than depreciation schedules assume. The rapid growth in AI-demand has also driven H100 one-year rental prices higher by approximately 40% from $1.70 per hour in October 2025 to $2.35 per hour by March 2026 (SemiAnalysis, March 2026),15 even as NVIDIA’s newer Blackwell GPUs entered the market. In our view, as long as incremental AI token demand exceeds the token supply enabled by the new chips produced in a given year, older GPUs retain – and can even gain – value (as long as token throughput increases over time as models improve – despite an increase in intelligence). More importantly, these deals function as a quasi-financing mechanism for the AI cloud buildout. Meta's deal in-particular provides access to investment grade borrowing costs with no equity dilution, while acting as a backstop customer if enterprise demand for the AI cloud doesn't materialize on schedule.”
At today's price, we are essentially buying a contracted bare metal business and a portfolio of valuable stakes at a fair price, and getting what could become a very big idea, as one of the world's next great AI cloud platforms with a world-class Founder/CEO and a team of engineers who have succeeded before at an attractive entry point.
We have conviction in Nebius' ability to build a large AI cloud business. They are leading the neocloud space in building a full suite of software offerings on their platform, much in line with what the big three hyperscalers have built for cloud workloads, i.e., multi-tenant compute, unified storage, inference-as-a-service, and security certifications, the kind of platform depth that took incumbents years to assemble.”
https://t.co/Hk6tSP47bM
Nebius choisit la France.
L’acteur du Cloud IA dévoile un projet de plus de 8 milliards d’euros d’investissement pour déployer des infrastructures et des services de cloud, avec une capacité cible de 240 MW, positionnant le site parmi les plus puissants du continent.
Bedankt!
@AntonLaVay@Frenchie_ 🙏 thanks. Kind of wondering how their yandex engineers settled in Netherlands? Hope they are happily settled there for their working environment?
@compassbow Thanks. NOK is supposed to report earnings on July 23. Management probably can’t buy stock 1 month before the earnings (regulatory). Guess their coming quarterly or this year’s earnings should be very positive
🚨 Jim Cramer on $NBIS and Situational Awareness led by Leopold Aschenbrenner:
“When this man touches something it turns to gold. Jensen Huang introduced me to Nebius two years ago. He said ‘I want you to keep an eye on them.’ The stock was at $16… always worth listening.”
@mikealfred@mattragland They don’t care but there is an unspoken assumption that their is money coming into their home so that they can eat, live well and do all sorts of school hobbies related activities