Nvidia’s Vera Rubin Platform Enters Production as AI Factory Strategy Takes Shape

Nvidia announces Vera Rubin production, DSX software for AI factories, and strong Q1 earnings; stock near highs as growth outlook solid.

Nvidia is no longer content being merely the world’s dominant chip supplier. At Computex in Taipei, chief executive Jensen Huang laid out a vision in which the company orchestrates entire AI data centres — and backed it up with tangible production milestones and a fortified supply chain. The message to investors: growth will not be constrained by delivery bottlenecks.

Shares in the tech heavyweight edged lower on Tuesday, trading at €191.72 in Frankfurt, a 0.6% dip from the prior session. That trims the year-to-date advance to 19%, but the stock remains just 4.6% below its 52-week high of €201.05 — territory that leaves little room for missteps on execution.

Vera Rubin hits the factory floor

Production of the next-generation Vera Rubin platform began on 1 June, with first systems already reaching customers such as OpenAI, Anthropic, xAI, Dell, Oracle and CoreWeave. The platform pairs the Vera central processing unit with a Rubin graphics processor, and Nvidia claims it delivers 3.5 times the training performance of its Blackwell predecessor and five times the inference speed.

Vera itself is no ordinary processor. Designed as an orchestration accelerator for “agentic” AI — autonomous systems that execute multi-step tasks and draw logical conclusions — it handles agent workloads 1.8 times faster than traditional x86 architectures, particularly in Python environments and code compilation. The broader Rubin ecosystem integrates BlueField-4 memory controllers and Spectrum-6 Ethernet switches, creating what Nvidia calls a unified fabric for high-performance AI clusters.

Full system deliveries are slated for the autumn, giving the market a concrete timetable for the next product wave.

A software layer for AI factories

Alongside the hardware reveal, Nvidia launched the DSX platform, a software control plane for managing large-scale AI factories. The system introduces “tokens per megawatt” as a new efficiency metric, reflecting the company’s push to measure output relative to energy consumption rather than raw compute.

DSX has already been adopted by a who’s who of server builders — ASUS, Dell, HPE, Lenovo and Supermicro — while early enterprise users include Oracle Cloud Infrastructure, CoreWeave and several prominent AI laboratories. The platform provides rack-scale infrastructure with security-by-design based on zero-trust storage, aiming to lower the operational complexity of running thousands of accelerators in parallel.

Record numbers underpin the strategy

The technology blitz follows stellar financial results. In May, Nvidia reported record quarterly revenue of $81.6 billion, a year-on-year jump of 85%, with the data-centre segment alone contributing $75.2 billion. The company boosted its share buyback programme by an additional $80 billion and raised its quarterly dividend from $0.01 to $0.25 per share — a move that signals management’s confidence that the AI boom is no flash in the pan.

For the second quarter of fiscal 2027, Nvidia guided for revenue around $91 billion, explicitly excluding contributions from regions subject to tightened export restrictions. With a market capitalisation of roughly $5.1 trillion, it remains the world’s most valuable technology stock.

Supply chain assurance and the PC offensive

Huang took pains to address investor concerns about production capacity. He highlighted investments by Taiwanese partners TSMC, Amkor, SPIL, Wistron and Foxconn in US manufacturing ecosystems, and stressed that Nvidia had secured sufficient supply across accelerators, memory, packaging and server assembly to support the ramp-up. The strategy aims to create redundancy and cushion political as well as logistical risks.

Nor is Nvidia limiting its ambitions to the data centre. A new RTX Spark chip, developed with Microsoft, will bring AI capabilities to personal computers, launching in the autumn. The announcement propelled the stock more than 6% on Monday, and the modest pullback since looks more like consolidation after a strong run — the shares still trade 19% above their 200-day moving average.

Huang also touched on the evolving role of interconnects in large systems, noting that Nvidia is deploying silicon photonics for optical connections where bandwidth demands are highest, while retaining copper where it remains efficient. The approach reinforces the company’s shift from a component vendor to a full-stack infrastructure provider, with the autumn delivery cycle shaping up as the real-world test of whether the strategy can sustain the blistering pace of growth.