Nvidia’s Two-Front Blitz: Mass Production of Vera Rubin and a Desktop AI Ambush

Nvidia is no longer content selling chips by the box. With Vera Rubin entering serial production and the RTX Spark poised to invade the Windows PC market, the company is executing a wholesale shift from component supplier to integrated AI infrastructure provider. Jensen Huang unveiled the full scope of that transformation at GTC Taipei on 1 June 2026, and the Frankfurt-listed shares responded with a 5.18% gain to €190.80, bringing the year-to-date advance to 18.44%.

The Vera Rubin NVL72 system sits at the heart of the data-centre push. It marries 72 Rubin GPUs with 36 Vera CPUs to deliver 3.6 exaFLOPS of compute, all packed into liquid-cooled racks that Nvidia now sells as a turnkey solution. The company claims inference throughput jumps tenfold against the Blackwell architecture, while the cost per token falls to one-tenth of the prior generation — a critical metric for the always-on compute patterns of AI agents. Early adopters include OpenAI, Anthropic and SpaceX, and Dell Technologies and CoreWeave already have operational installations. To guarantee supply, Nvidia has locked in HBM4 memory from SK hynix, Samsung Electronics and Micron; the Vera CPU itself houses 88 “Olympus” cores and is said to be 1.8 times faster than x86 processors for agentic workloads.

On the desktop side, the RTX Spark represents Nvidia’s boldest assault yet on turf long held by Intel, AMD, Qualcomm and Apple. Developed in partnership with MediaTek and manufactured at TSMC, the chip pairs 20 Arm-based Grace CPU cores with a Blackwell GPU packing 6,144 CUDA cores and fifth-generation tensor cores, linked via NVLink-C2C. The result is 1 petaflop of local AI performance and up to 128 GB of unified memory — enough to run large language models entirely off the cloud. Microsoft, Dell, HP, ASUS, Lenovo and MSI will ship Spark-powered laptops from autumn 2026, with Microsoft’s Surface Laptop Ultra among the first flagships. A companion DGX Station for Windows, driven by the GB300 Grace Blackwell Ultra desktop superchip, offers 748 GB of coherent memory and 20 petaflops of FP4 performance for billion-parameter models; it lands in the fourth quarter of 2026 with partners including ASUS, GIGABYTE and Supermicro.

Should investors sell immediately? Or is it worth buying Nvidia?

The product offensive comes atop a blistering financial run. In the first quarter of fiscal 2027 Nvidia generated $81.6 billion in revenue, up 85.2% year on year, with the data-centre segment alone contributing $75.25 billion — a 92% surge. For the full fiscal 2026, revenue reached $215.9 billion, 65% higher than the prior year, at an operating margin of roughly 60%. The board recently authorised an additional $80 billion share buyback and raised the quarterly dividend from $0.01 to $0.25 per share, payable on 26 June 2026. The market capitalisation now exceeds $5.1 trillion, and the stock sits about 5% below its 52-week high of €201.05 in Frankfurt.

Nvidia is also deepening its footprint in Taiwan, where annual spending has ballooned from $10-15 billion to roughly $100 billion. A new campus dubbed “Constellation” is slated to open around 2030 as the company’s overseas headquarters, housing some 4,000 employees. On the software front, Huang unveiled Cosmos 3, an open world model that unifies image understanding, world generation and action prediction. Meanwhile, the DRIVE Hyperion Level 4 robotaxi platform added Foxconn (Taiwan rollout from 2028, then Asia), VinFast (Southeast Asia), Uber (Munich programme) and Humain (Saudi Arabia) as new partners.

Yet the regulatory environment is tightening. On 1 June 2026 the US sharpened export controls on AI chips, with particular scrutiny on Chinese subsidiaries of Western firms. Nvidia’s CFO Colette Kress noted that hyperscalers could push spending to $1 trillion by 2027 and that total AI infrastructure investment may reach $3-4 trillion by 2030. To lock down the supply chain, the company is partnering with Corning on three new factories and holds an option to invest up to $2.7 billion in the glass-maker. Huang remains defiant, arguing that “useful AI” has arrived and that billions of AI agents will actually increase demand for software developers. The product roadmap stays relentless: Vera Rubin will be followed by the Feynman architecture in 2029 and Rosa Feynman Spark in 2030. For now, the market is watching whether the Taipei keynote can turn that pipeline into a sustained upward move.

Ad

Nvidia Stock: New Analysis - 1 June

Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.

Read our updated Nvidia analysis...

en | US67066G1040 | NVIDIA’S | boerse | 69465784 |