Dell and NVIDIA Unveil Next-Gen AI Factory: Parabolic Demand Drives Vera Rubin, Cost Breakthroughs
At Dell Technologies World 2025, Michael Dell and Jensen Huang painted a vivid picture of an AI landscape shifting from pilot projects to full-scale production. With worldwide AI infrastructure spending predicted to hit $3–4 trillion by 2030 and token consumption soaring 3,400%, the duo introduced a suite of new hardware—including the Dell PowerEdge XE9812 built on NVIDIA Vera Rubin NVL72—that cuts inference costs by 10x and accelerates agentic AI. Here’s what you need to know.
What did Michael Dell reveal about AI spending and token growth?
Michael Dell kicked off the keynote by sizing the opportunity: global AI infrastructure spending could reach $3–4 trillion by 2030, while token consumption is projected to grow an astonishing 3,400% in the same period. He described the current pace as “parabolic” and emphasized that the productivity boom is already underway in companies like his own. “The rate of change has gone parabolic, and it’s not slowing down,” Dell said. This sets the stage for the massive investments in accelerated computing and networking that followed.
How did Jensen Huang define “useful AI” and parabolic demand?
Jensen Huang took the stage after Dell and declared that “useful AI” has arrived, which is why demand is “going parabolic, utterly parabolic.” He explained that tasks that once took months now take weeks, weeks become days, and days shrink to hours. This productivity leap, however, demands a gigantic leap in computation. Huang emphasized that enterprise AI has moved beyond pilots into agentic AI and inference at scale—secure, autonomous agents running inside the corporate perimeter, powered by the Dell AI Factory with NVIDIA.
What is the Dell PowerEdge XE9812 and how does it lower cost per token?
The centerpiece of the hardware refresh is the Dell PowerEdge XE9812, built on the NVIDIA Vera Rubin NVL72 platform. It delivers up to 10x lower cost-per-token compared to the previous NVIDIA Blackwell architecture, making it ideal for massive-scale agentic AI inferencing. By drastically reducing the expense of processing tokens, enterprises can deploy more sophisticated models and autonomous agents without breaking their budgets. This server is part of a broader lineup designed to handle the explosive growth in AI workloads.
What other Dell PowerEdge servers were announced alongside the XE9812?
Dell also introduced the PowerEdge XE9880L, XE9885L, and XE9882L—the first Dell systems built on NVIDIA HGX Rubin NVL8. These servers support up to 144 GPUs per rack with 100% direct liquid cooling for the compute nodes, delivering up to 10x the performance of the HGX B200. Designed for the most demanding AI training and inference tasks, they enable enterprises to scale their AI factories efficiently while managing power and thermal constraints.
What networking updates did Dell and NVIDIA announce?
To keep data flowing at AI-scale speeds, Dell unveiled the new Dell PowerSwitch portfolio featuring NVIDIA Quantum-X800 InfiniBand—with liquid-cooled, co-packaged optics—and NVIDIA Spectrum-6 Ethernet. These switches are engineered to reduce latency and increase bandwidth for distributed AI workloads. The combination of high-performance networking and the new servers ensures that the Dell AI Factory can handle the massive data movement required by agentic AI and large language models.
What is the Dell PowerRack and why is it important?
Dell introduced Dell PowerRack, a fully integrated system that combines compute, networking, and storage into a single engineered chassis. It features unified thermal design, power management, and software optimization—all built to work together from the ground up. The result is accelerated AI and HPC workloads at enterprise scale without the overhead of assembling individual components. PowerRack simplifies deployment and maintenance, making it easier for organizations to transition from pilots to production.
What Vera CPU-based servers did Dell announce for agentic AI?
On the CPU side, Dell launched the PowerEdge M9822 and R9822 servers, powered by NVIDIA Vera CPUs. These systems are purpose-built for agentic AI, where each task—data pipelines, analytics, sandboxed tools, and code workloads—depends on the previous step. The Vera CPU runs these sequential jobs 50% faster than traditional CPUs, and enterprise data queries are up to 3x faster. This makes them ideal for running autonomous agents that require low-latency, step-by-step processing.
How many enterprises are already using Dell AI Factories with NVIDIA?
Michael Dell confirmed that 5,000 enterprises—including industry giants like Lilly, Samsung, and Honeywell—are currently running AI workloads on Dell AI Factories with NVIDIA. These companies are moving from experimentation to production at scale, leveraging the full stack of accelerated computing, networking, and software. The sheer number underscores the urgency of the parabolic demand Huang described and validates Dell’s strategy of offering integrated, end-to-end AI infrastructure.