NVIDIA GTC 2025: Recap and Reflections

nvidia gtc 2025 blog post

Zadara’s Role in the AI Revolution

At Zadara, we have spent years developing scalable, automated multi-tenancy solutions for on-premises and sovereign edge clouds. Our close collaboration with NVIDIA and partners ranging from enterprises to manage service providers, telecom providers, and NCPs has enabled us to optimize our platform for AI training, AI reasoning, and digital AI sovereignty. We have been adapting our software to fit into the NVIDIA Reference Architecture for NCPs. Moreover, we also developed an entire framework for the deployment of a complete sovereign Retrieval Augmented Generation (RAG) system on an AI factory or NCP. We successfully deployed this system within organizations to provide advanced RAG capabilities for tenants leveraging multiple LLMs supporting digital sovereignty, laying the groundwork for the next wave of AI innovation.

NVIDIA GTC 2025

NVIDIA GTC 2025 once again showcased groundbreaking advancements in AI, highlighting key trends and developments from NVIDIA and its extensive ecosystem. The conference provided a unique opportunity to engage with industry leaders at the forefront of AI innovation. Several of us from Zadara participated, gaining invaluable insights and fostering meaningful discussions with NVIDIA, partners, and customers. While much has already been said about CEO Jensen Huang’s visionary keynote and the numerous announcements, this post reflects on the major trends observed, particularly in comparison to GTC 2024.

Key AI Market Trends

From a market perspective, AI factories, AI reasoning, and digital AI sovereignty emerged as dominant themes, shaping discussions with NVIDIA and customers alike. NVIDIA’s announcements reinforced these trends, but further advancements are required to fully meet the surging demand in these domains.

AI Factories: Expanding AI Infrastructure

AI factories – cloud or on-premises infrastructures optimized for large-scale AI workloads are extending AI capabilities beyond hyperscalers to telecom providers and enterprises. This shift is driven by the growing need for independent AI processing, reducing reliance on centralized cloud providers. Enterprises and telecom companies are increasingly adopting AI factories to complement hyperscalers and NVIDIA Cloud Partners (NCPs).

The Rise of AI Reasoning

AI reasoning, an evolution of AI inferencing, requires significantly higher computing power than AI inferencing. Unlike AI training, AI reasoning benefits from proximity to users, making AI factories and NCPs ideal for hosting these workloads. The transition from training-centric AI to reasoning-centric AI is fueling demand for decentralized AI infrastructure. Yet, decentralized AI requires powerful infrastructure to fulfill AI reasoning demands.

Digital AI Sovereignty: A Growing Priority

Digital AI sovereignty is not new but its momentum accelerated recently due to geopolitical shifts, stringent regulatory compliance, and increasing efforts to build local technology stacks. Telecom providers leveraging AI factories and NCPs, are well-positioned to meet these demands, ensuring greater control over AI data, infrastructure, and decision-making processes.

NVIDIA’s Announcements: Powering the Future of AI

Several key announcements at GTC 2025 set the stage for accelerated development to support the growing demand for AI factories and AI reasoning. The unveiling of NVIDIA Blackwell Ultra, combined with NVIDIA Spectrum-X™ Ethernet and NVIDIA Quantum-X800 InfiniBand, addresses AI reasoning bottlenecks. Additionally, NVIDIA BlueField®-3 DPUs enhance multi-tenant networking, GPU elasticity, accelerated data access, and real-time cybersecurity threat detection—critical components for AI factories.

Software: The Key to AI Factory and NCPs Optimization

To maximize efficiency, AI factories and NCPs require advanced software solutions that provide scalable, secure multi-tenancy with full isolation and automation, optimizing resource utilization without burdening AI workloads. Deploying such software on control clusters or preferably on NVIDIA BlueField®-3 DPUs transforms DPUs into the control plane for AI factory operations while preserving data plane performance.

Zadara’s Role in the AI Revolution

At Zadara, we have spent years developing scalable, automated multi-tenancy solutions for on-premises and sovereign edge clouds. Our close collaboration with NVIDIA and partners ranging from enterprises to manage service providers, telecom providers, and NCPs has enabled us to optimize our platform for AI training, AI reasoning, and digital AI sovereignty. We have been adapting our software to fit into the NVIDIA Reference Architecture for NCPs. Moreover, we also developed an entire framework for the deployment of a complete sovereign Retrieval Augmented Generation (RAG) system on an AI factory or NCP. We successfully deployed this system within organizations to provide advanced RAG capabilities for tenants leveraging multiple LLMs supporting digital sovereignty, laying the groundwork for the next wave of AI innovation.

Stay tuned for upcoming announcements as we continue working alongside NVIDIA and our partners to drive advancements in AI infrastructure and software solutions for AI workloads.

Sovereign AI by Zadara Edge Cloud

sovereign ai solution brief cover

Share This Post

More To Explore