Nvidia’s Vera Rubin Racks Hit Full Production as a $2 Billion Cloud Bet Reshapes the Narrative
Published on 07/22/2026 at 16:22 | Redaktion boerse-global.de
Nvidia is writing two very different stories this week. On one hand, the chipmaker’s next-generation Vera Rubin platform has entered full-scale production, with the first racks already shipping to major cloud customers. On the other, a regulatory filing has revealed a 9.3% stake in AI cloud provider Nebius — a position that underscores just how deeply Nvidia is weaving itself into the fabric of the AI infrastructure ecosystem.
The shares, meanwhile, are treading water. After closing at €181.88 on Tuesday with a 2.1% gain, the stock slipped 1.1% to €179.88 on Wednesday. That leaves the equity roughly 11% below its 52-week high of €202.50 set on May 14. The relative strength index sits at 51.2, signaling neither overbought nor oversold conditions.
Vera Rubin Goes Live With Eye-Popping Benchmarks
Nvidia officially began shipping its Vera Rubin platform on July 21, and by the following day multiple sources confirmed the system was already in volume production. The centerpiece is the NVL72 rack, which pairs 72 Rubin GPUs with 36 of Nvidia’s own Vera CPUs. Those CPUs pack 88 Olympus cores, 176 threads, and deliver 1.2 TB/s of memory bandwidth over LPDDR5X. The GPU connection runs over NVLink-C2C at up to 1.8 TB/s, while the system is tied together with sixth-generation Spectrum-X Ethernet fabric switches capable of 102.4 terabits.
Early performance numbers are striking. CoreWeave ran benchmarks using the DeepSeek-R1 model and recorded a tenfold improvement in token throughput per megawatt compared to the previous Grace-Blackwell generation — 800,000 tokens per second versus 80,000, at 150 megawatts of power draw. Nvidia also claims the Vera CPU runs Python workloads 1.8 times faster than AMD’s Turin processor.
Should investors sell immediately? Or is it worth buying Nvidia?
The launch partners include CoreWeave, Google Cloud, Microsoft Azure, and Oracle. OpenAI plans to deploy the racks at scale in the third quarter of 2026, while Anthropic and SpaceX are already testing the systems.
There’s a price tag to match the ambition: a Vera Rubin rack runs roughly $7 million to $8 million, well above the $5 million cost of a Blackwell rack. Nvidia is manufacturing across more than 350 facilities in 30 countries. SVP Andrew Bell put the theoretical maximum capacity at up to 1,000 racks per day — a figure that would translate to quarterly revenue north of $630 billion, though he was careful to note that remains aspirational rather than operational.
The Nebius Stake: More Than Meets the Eye
The Nebius filing, submitted to the SEC on July 20 via a Schedule 13G, reveals Nvidia holds 22.26 million shares — a 9.3% stake. The position traces back to a $2 billion investment Nvidia made to help Nebius expand its data center footprint.
But the headline number requires unpacking. Only 1.19 million of those shares come from earlier 13F filings. The remaining 21.07 million stem from a prepaid warrant that remains locked until September 11. Because the warrant can be exercised within 60 days of July 13, regulators require it to be counted toward beneficial ownership — which explains why the disclosed stake jumped from roughly 8.3% in March to 9.3% now.
The Nebius bet is part of a broader pattern. Nvidia has scattered similarly sized investments across the AI supply chain, including a separate $2 billion deal with Marvell Technology and stakes in Synopsys, CoreWeave, Coherent, and Lumentum. The largest single bet by far was the $30 billion stake in OpenAI closed in February 2026, embedded in a roughly $110 billion funding round. That was considerably less than the up-to-$100 billion framework Nvidia had floated the previous September.
CEO Jensen Huang hinted weeks after the OpenAI deal that the investment spree in the two leading AI labs might be winding down. The OpenAI investment could be “the last time,” he said, citing the startup’s planned IPO. A similar logic applies to the Anthropic stake, where Nvidia and Microsoft jointly committed up to $15 billion in November 2025 — Nvidia up to $10 billion, Microsoft up to $5 billion — in exchange for Anthropic agreeing to buy $30 billion in Azure compute capacity.
The Nebius relationship carries a more concrete business rationale. Meta signed a long-term agreement in March to invest up to $27 billion in Nebius’s AI infrastructure. Nebius specializes in GPU-intensive workloads, putting Nvidia in the unusual position of being both hardware supplier and co-owner.
Analyst Upgrade and Solid Fundamentals
Against this backdrop, Zacks Research upgraded the stock from Hold to Strong Buy on July 22. The call rests on recent quarterly results that beat expectations: earnings per share of $1.87 topped the consensus of $1.76, while revenue of $81.61 billion came in ahead of the $78.42 billion estimate — an 85.2% year-over-year jump. The Data Center segment grew 92% to $75.2 billion, with gross margin at 74.9%.
Nvidia at a turning point? This analysis reveals what investors need to know now.
The board also authorized an $80 billion share buyback and raised the quarterly dividend from $0.01 to $0.25, yielding roughly 0.5%. The payout ratio of 15.31% of earnings underscores that Nvidia remains a growth company first. Insider sales totaled $410.6 million over the past 90 days, while institutional ownership stands at 65.27%.
China Jitters and the Jevons Paradox
The broader semiconductor sector got a scare recently when Moonshot AI’s Kimi 3 model — boasting 2.8 trillion parameters and a 1-million-token context window — sent chip stocks tumbling more than 20% from their June highs, dragging the PHLX semiconductor index into bear market territory. Huang, speaking at the China International Supply Chain Expo, called Chinese AI models “world-class” and argued against banning Chinese open-source models, while continuing to support export restrictions on hardware like Blackwell and Rubin.
Analysts point out that Kimi 3, despite a 75% reduction in KV-cache requirements, still depends on powerful Nvidia racks due to its 896-expert architecture — a textbook case of Jevons paradox, where efficiency gains drive higher rather than lower chip demand. A meeting between Huang and the chiefs of Samsung, SK Group, and Naver is scheduled for next week near Silicon Valley, with OpenAI and Anthropic CEOs potentially joining.
Chart and Outlook
The stock has gained 13.48% year to date and 27.85% over the past twelve months. The RSI of 53.5 on Tuesday’s close confirms a neutral technical posture. With Vera Rubin now in production, a $2 billion cloud bet coming into clearer focus, and an analyst upgrade in hand, Nvidia is navigating a moment where operational momentum and strategic positioning are pulling in the same direction — even if the share price hasn’t fully caught up yet.
Ad
Nvidia Stock: New Analysis - 22 July
Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
Disclaimer regarding our articles: No investment advice, no buy or sell recommendation. Information on prices, companies, and markets is provided without guarantee; changes are possible at any time. Stock market transactions can lead to substantial losses. Our articles are created and reviewed in whole or in part automatically with the support of AI.
