Nvidia GTC: Transforming Data Centres into AI Factories

Nvidia has unveiled a comprehensive suite of data centre technologies at GTC 2025, which are a testament to company CEO Jensen Huang's vision for transforming traditional computing facilities.

Supporting what Nvidia refers to as ‘AI factories’, the GTC announcements range across hardware, software and services. Overwhelmingly, they are designed to power the next wave of AI applications - particularly those that are focused on AI reasoning and agentic AI.

“AI has made a giant leap - reasoning and agentic AI demand orders of magnitude more computing performance,” says Jensen Huang, founder and CEO of Nvidia, during his keynote address at the event.

Blackwell Ultra to power next-generation AI

A focal point of Nvidia’s announcements was the evolution of its Blackwell architecture with the Blackwell Ultra platform. 

Building upon the original Blackwell architecture introduced last year, Blackwell Ultra represents a significant leap forward with the GB300 NVL72 rack-scale solution, which connects an impressive 72 Blackwell Ultra GPUs and 36 Arm Neoverse-based Grace CPUs in a single rack design.

Jensen Huang shares: “We designed Blackwell Ultra for this moment — it's a single versatile platform that can easily and efficiently do pretraining, post-training and reasoning AI inference.”

According to Nvidia, the GB300 NVL72 delivers 1.5x more AI performance than its predecessor and increases Blackwell's revenue opportunity by 50x for AI factories compared to those built with Nvidia Hopper. 

The platform is particularly important for organisations looking to implement “test-time scaling” - the technique of applying more compute during inference to improve accuracy, which is critical for reasoning and agentic AI.

DGX SuperPOD: Redefining AI Supercomputing

Nvidia has also announced what it refers to as “the world’s most advanced enterprise AI infrastructure” - the Nvidia DGX SuperPOD.

Built with Blackwell Ultra GPUs, these systems come in two main configurations and are designed to be deployed as turnkey AI factory supercomputers.

Notably, the DGX GB300 systems feature Nvidia Grace Blackwell Ultra Superchips in a rack-scale, liquid-cooled architecture designed for real-time agent responses on advanced reasoning models. 

These systems deliver up to 70x more AI performance than AI factories built with Hopper systems and include 38TB of fast memory to offer what Nvidia calls “unmatched performance at scale for multistep reasoning”.

For data centres that prefer standard air cooling, Nvidia has introduced the DGX B300 systems, which leverage the B300 NVL16 architecture. These systems deliver 11x faster AI performance for inference and a 4x speedup for training compared to the previous Hopper generation, with each system providing 2.3TB of HBM3e memory.

The networking infrastructure for these systems consists of the DGX GB300 connecting its 72 GPUs via fifth-generation NVLink technology to create one massive shared memory space. 

The systems seamlessly integrate with both Nvidia’s Quantum-X800 InfiniBand and Spectrum-X Ethernet platforms, ensuring data can flow without bottlenecks.

RTX PRO Blackwell: Supporting the data centre

Complementing the high-end DGX systems, Nvidia also announces the RTX PRO Blackwell series, including the RTX PRO 6000 Blackwell Server Edition specifically designed for data centre deployments. 

The GPU features a passively cooled thermal design that can be configured with up to eight GPUs per server, making it suitable for a variety of enterprise workloads.

RTX PRO 6000 also supports Nvidia vGPU software, allowing organisations to power AI workloads across virtualised environments and deliver high-performance virtual workstation instances to remote users - an increasingly important capability in today's distributed work environments.

Supporting an industry under high demand

For data centre management, Nvidia has introduced Mission Control - an AI data centre operation and orchestration software specifically designed for Blackwell-based DGX systems. 

It aims to automate the management and operations of AI infrastructure and address one of the key challenges organisations face when scaling up AI deployments.

The company also highlighted Nvidia AI Enterprise, its software platform for building and deploying enterprise-grade AI agents. This includes Nvidia NIM microservices and the newly announced Nvidia Llama Nemotron open reasoning model family.

Also recognising that not all organisations are able to build their own AI infrastructure, Nvidia has unveiled its ‘Nvidia Instant AI Factory’ - a managed service that features the Blackwell Ultra-powered Nvidia DGX SuperPOD with Nvidia Mission Control software.

Equinix will be the first to offer this service, with preconfigured liquid or air-cooled AI-ready data centres across 45 markets around the world. 

The service also promises to provide businesses with fully provisioned, intelligence-generating AI factories optimised for state-of-the-art model training and real-time reasoning workloads.

Nvidia's ecosystem continues to expand, with major organisations like Cisco, Dell Technologies, HPE, Lenovo and Supermicro expected to deliver servers based on both Blackwell Ultra and RTX PRO Blackwell products. 

Making such comprehensive announcements at GTC highlight how Nvidia is committed to transforming the traditional data centre with the power of AI to support rising workloads. As AI applications continue to develop, the company is positioning itself at the forefront of this transition.

Explore the latest edition of Data Centre Magazine and be part of the conversation at our global conference series, Tech & AI LIVE and Data Centre LIVE. 

Discover all our upcoming events and secure your tickets today.

Data Centre Magazine is a BizClik brand