Nvidia Wields Software to Control AI Factories as Vera Rubin Production Kicks Off

Nvidia shifts from chipmaker to data center orchestrator with DSX software platform and Vera Rubin hardware, promising 40% more GPU density and 5x inference throughput.

Nvidia is rewriting its playbook. The chipmaker that dominated the AI accelerator market is now positioning itself as the central nervous system of the modern data center, bundling hardware with a software stack that manages everything from power distribution to digital simulation. The shift was laid bare at Computex and GTC Taipei, where the company unveiled its DSX software platform alongside confirmation that its next-generation Vera Rubin architecture is already rolling off production lines.

The dual announcement sent a clear message: Nvidia intends to lock in its lead not just through brute silicon performance, but by dictating how entire AI clusters are designed, deployed and operated. At the same time, CEO Jensen Huang used his Taipei keynote to address the market’s biggest anxiety head-on — supply chain reliability.

A Platform Play, Not Just a Chip Refresh

The DSX platform is Nvidia’s most ambitious software push to date. Its core module, DSX MaxLPS, manages real-time power allocation with enough precision to let data center operators run 40 percent more GPUs within the same electricity budget. That is not an incremental efficiency gain; in an era where energy costs can make or break a cluster’s economics, the savings translate directly into competitive advantage.

DSX is modular. DSX Sim builds digital twins inside Nvidia’s Omniverse environment for pre-deployment validation. DSX Exchange acts as a developer API hub, DSX Flex handles network connectivity, and DSX OS provides a modular operating system, part of which is being released as open source. Nvidia is not building the ecosystem alone. Vertiv announced on June 3 that it will integrate a digital twin of its SmartRun infrastructure system into Omniverse DSX, allowing cooling and power validation before a single server rack is installed. ASUS is using DSX to optimise network and cooling planning for its liquid-cooled systems built on the Vera Rubin architecture, while network specialist Netris is extending its NAAM platform across Asia-Pacific as “Nvidia DSX Air” to speed up GPU cloud deployments.

Vera Rubin Enters the Fray

While the software story dominated the presentations, the hardware pipeline remains the headline growth driver. Vera Rubin, which combines a Vera CPU with a Rubin GPU, has entered full production. First systems are already reaching customers including OpenAI, Anthropic, xAI, Dell, Oracle and CoreWeave. Performance claims are striking: the platform is said to deliver 3.5 times the training performance of Blackwell and five times the inference throughput. Full system deliveries are slated for autumn.

The Vera CPU itself is now moving into volume production. Nvidia says it achieves 1.8 times the performance of conventional x86 processors on agentic workloads, a signal that the company is serious about challenging Intel and AMD in the CPU arena. The broader Vera Rubin rollout gives the market a concrete timeline for the next product wave, which analysts will watch closely as a bellwether for Nvidia’s ability to sustain its blistering revenue growth.

Financial Rearrangement Reflects Changing Customer Mix

Underlying the product news is a significant reorganisation of Nvidia’s financial segments. The former catch-all “Data Center” line has been split into two categories. The Hyperscale segment — sales to the biggest cloud operators — generated approximately $37.9 billion in the first quarter of fiscal 2027, a 115 percent increase year-over-year. The new ACIE segment (AI Clouds, Industrial, Enterprise) grew 74 percent, underscoring that industrial and corporate customers are becoming a discrete growth pillar in their own right.

That shift is already showing up in the stock price. Nvidia shares traded at €191.72 on Tuesday, down 0.61 percent on the day after a 6 percent pop on Monday following the RTX Spark announcement. The stock remains 19 percent higher since the start of the year and sits just 4.64 percent below its 52-week high of €201.05. Over a 12-month horizon the gain stands at more than 54 percent, with shares trading roughly 19 percent above their 200-day moving average.

Management has guided for second-quarter revenue of about $91 billion. Shareholders are being rewarded, too: at least half of free cash flow is being returned through buybacks and dividends. The buyback programme has been expanded by $80 billion, and the quarterly dividend has been raised to $0.25 per share, payable on June 26, 2026. With a market capitalisation of roughly $5.4 trillion, the absolute payout is large, but relative to the valuation it remains more a symbolic gesture — the real value driver is the growth trajectory of the core business.

PC Battlefront Opens with RTX Spark

Nvidia is not content with the data center. At Computex, it unveiled the RTX Spark chip, developed in collaboration with Microsoft, to bring local AI inference to personal computers. The product is scheduled for an autumn launch, putting Nvidia in more direct competition with AMD, Intel and Apple in the PC space. The announcement sent shares up more than 6 percent on Monday, and the slight pullback on Tuesday looks more like a pause than a reversal.

Huang also touched on a longer-term infrastructure shift: the transition from copper cabling to silicon photonics for data transmission in large systems. Nvidia plans to use optical interconnects where they are needed and copper where it remains efficient, a pragmatic approach that avoids betting the farm on a single technology.

The autumn window will be the first real test. Between Vera Rubin system deliveries, the RTX Spark launch and the broader adoption of the DSX platform, Nvidia is laying out a dense timeline of execution milestones. Investors have priced in high expectations; the stock’s proximity to its 52-week high leaves little room for shipment or margin disappointments. But with production lines running, a software stack that ties the data center together, and a growing customer base that now spans hyperscalers, enterprises and PC users, the company is making a credible case that its growth story is only entering its next chapter.