At COMPUTEX 2025, NVIDIA showcased a trio of AI-native computing platforms—DGX Spark, DGX Station, and the newly announced RTX Pro Servers—aimed at accelerating the adoption of AI infrastructure across enterprises of all sizes. Together, these systems reflect NVIDIA’s strategic commitment to making high-performance AI compute scalable, modular, and accessible—not only for hyperscalers, but for startups, SMEs, and IT-led enterprises looking to deploy AI locally.

DGX Spark: AI Development, Now Within Reach

First unveiled earlier this year, DGX Spark returned to COMPUTEX 2025 with expanded availability and strong OEM support. With 1 petaflop of performance in a compact workstation form factor, it enables developers and small teams to build, fine-tune, and test AI models on-premise—without the operational overhead of cloud provisioning.

Now available from vendors such as Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI, DGX Spark systems were actively showcased across partner booths at COMPUTEX. Shipping begins in July, making it a timely and practical choice for SMEs looking to adopt generative AI, vision systems, or RAG pipelines.

DGX Station: Datacenter-Class AI, Desk-Side Format

While announced earlier at GTC, DGX Station made its COMPUTEX debut with OEM-branded variants now available through major NVIDIA hardware partners. The system delivers enough performance to fine-tune trillion-parameter models—all within a workstation-sized footprint that runs off standard power.

DGX Station is now offered by enterprise vendors such as ASUS, Dell, HP, Lenovo, and MSI, providing SMEs and mid-sized organizations a convenient path to advanced AI development infrastructure—especially where data privacy, latency, or local control is a priority.

RTX Pro Servers: Enterprise AI at Production Scale

The headline launch at COMPUTEX 2025 was the RTX Pro Server, a new category of enterprise-grade AI infrastructure designed to run multimodal AI agents, real-time inference, and simulation workloads.

Unlike the DGX systems, which are optimized for AI training at massive scale (often in large research labs or datacenters), and EGX systems, which focus on edge deployments and specialized verticals (e.g. smart cities, industrial automation), RTX Pro Servers are designed as general-purpose AI production systems that integrate directly into standard enterprise IT environments.

What makes RTX Pro Servers different—and SME friendly:

• Standard x86 compatibility: They run in typical server rooms and support existing infrastructure like VMware and Kubernetes.
• Scalable GPU configurations: Up to 8 GPUs per node, but flexible for smaller setups.
• Cost-effective compared to DGX-class clusters, while still supporting demanding inference workloads (e.g. LLMs, copilots, creative agents).
• OEM support from Dell, Lenovo, ASUS, GIGABYTE, HP, and others, enabling localized procurement, service, and financing options.

This makes RTX Pro Servers a compelling choice for mid-sized enterprises and advanced SMEs looking to deploy AI agents across functions—without needing to overhaul their IT environment or invest in specialized datacenter infrastructure.

Conclusion: The AI Infrastructure Stack is Maturing—and Democratizing

At COMPUTEX 2025, NVIDIA sent a clear message: AI infrastructure is no longer just for tech giants.

With DGX Spark for prototyping, DGX Station for advanced development, and RTX Pro Servers for production deployment, NVIDIA is delivering a full-stack approach that meets businesses where they are—whether that’s a single AI engineer or an enterprise IT department scaling digital agents across operations.

For SMEs, this is a rare moment of convergence: enterprise-grade AI compute, widely available, OEM-supported, and designed to fit both technical and business realities.