Nvidia shows off its latest data center GPU: The Blackwell Ultra

Nvidia shows off its latest data center GPU: The Blackwell Ultra

  • 18.03.2025 00:00
  • datacenterdynamics.com
  • Keywords: AI

Nvidia introduced its Blackwell Ultra GPU for data centers, part of an annual release strategy driven by AI growth. The chip offers high performance in new DGX systems and is supported by major cloud providers and server manufacturers.

Nvidia ServicesNVDAsentiment_satisfied
open_in_newOriginal article

Estimated market influence

Nvidia

Nvidia

Positivesentiment_satisfied
Analyst rating: Strong buy

Announced the Blackwell Ultra GPU and new DGX systems, which are expected to boost AI performance and expand their market share in data center GPUs.

Context

Business Insights and Market Implications

Key Product Details

  • GPU Architecture: Built on Nvidia's Blackwell architecture, introduced in 2024.
  • Performance:
    • 15 petaflops of FP4 performance.
    • 288GB HBM3e memory.
    • Combines two reticle-sized GPUs.

System Configurations

  • Nvidia GB300 NVL72:
    • Connects 72 Blackwell Ultra GPUs and 36 Arm Neoverse-based Nvidia Grace CPUs in a single liquid-cooled rack.
    • Offers 1.5x more AI performance than the previous generation (GB200 NVL72).
  • Nvidia HGX B300 NVL16: Another system configuration for hyperscale deployments.

Release and Availability

  • Delivery Timeline: Expected in the second half of 2025.
  • Launch Delays: Blackwell GPUs missed their initial launch, though exact impact is unclear.

Ecosystem and Partnerships

  • Server Manufacturers:
    • Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, Supermicro.
    • Additional partners: Aivres, ASRock Rack, ASUS, Eviden, Foxconn, Gigabyte, Inventec, Pegatron, Quanta Cloud Technology (QCT), Wistron, Wiwynn.
  • Hyperscalers and Neoclouds:
    • Amazon Web Services, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure.
    • Additional cloud providers: CoreWeave, Crusoe, Lambda, Nebius, Nscale, Yotta, YTL.

Market Impact

  • AI Performance Boost: The 1.5x performance improvement positions Blackwell Ultra as a significant upgrade for AI workloads.
  • Density and Scalability: Liquid-cooled rack design enables high-density deployments, critical for hyperscale data centers.
  • Competition in Cloud Infrastructure: Hyperscalers offering Blackwell Ultra-powered instances will intensify competition in the cloud services market.

Strategic Considerations

  • Nvidia's Leadership: Reinforces Nvidia's dominance in AI and GPU technology.
  • Customer Demand: High-performance GPUs are essential for advancing AI applications, ensuring sustained demand.
  • Ecosystem Expansion: Extensive partnerships ensure broad adoption across industries.

Long-Term Effects

  • AI Growth: The shift to annual GPU releases aligns with the rapid evolution of AI technologies.
  • Regulatory Impact: No immediate regulatory hurdles mentioned, but ongoing scrutiny of tech monopolies could affect market dynamics.