WEKA has announced a significant expansion of its AI-native data platform, integrating with the NVIDIA AI Data Platform reference design to optimise AI infrastructure for agentic AI and reasoning models. The move, revealed at GTC 2025, includes multiple NVIDIA storage certifications aimed at enhancing enterprise AI efficiency and scalability.
Advancing AI with High-Performance Storage
The NVIDIA AI Data Platform is designed to redefine AI infrastructure by integrating NVIDIA Blackwell GPUs, BlueField DPUs, Spectrum-X networking, and AI Enterprise software with high-performance storage. WEKA’s data platform now aligns with this architecture, providing a scalable foundation for AI-driven enterprises. The partnership ensures that AI models can connect more effectively to business knowledge, achieving higher accuracy in complex reasoning tasks.
As part of this development, WEKA has achieved new certifications, including the NVIDIA Cloud Partner (NCP) Reference Architecture with GB200 NVL72 and the NVIDIA-Certified Systems™ Storage designation for enterprise AI deployments. These certifications validate WEKA’s compatibility with NVIDIA best practices, ensuring optimised storage performance and cost efficiency.
Breaking the AI Memory Barrier
A key innovation unveiled by WEKA is its Augmented Memory Grid™, which integrates WEKA Data Platform software with NVIDIA’s accelerated computing and networking solutions. The technology aims to significantly improve AI inference speed by increasing token efficiency and the number of tokens processed per second.
AI models require larger context windows, expanded parameters, and greater memory capacity for effective reasoning and decision-making. WEKA’s Augmented Memory Grid addresses this challenge by extending memory capacity by petabytes, significantly beyond today’s standard single-terabyte increments. Near-memory speed performance at microsecond latencies ensures faster token processing, leading to better reasoning outcomes.
Key benefits of the technology include:
- 41x Faster Time to First Token: WEKA’s solution reduces the time needed to process 105,000 tokens compared to conventional prefill recalculations.
- Improved Token Processing Efficiency: AI inference clusters can achieve higher token throughput, cutting overall token processing costs by up to 24%.
AI Innovation and New Storage Certifications
WEKA’s WEKApod™ Nitro Data Platform Appliances have also been certified as one of the first high-performance storage solutions for NVIDIA Cloud Partner (NCP) deployments. These appliances support NVIDIA HGX H200, B200, and GB200 NVL72, providing enhanced infrastructure for AI developers.
Additionally, the WEKApod Nitro appliances have received NVIDIA-Certified Systems Storage designation, confirming their compliance with NVIDIA’s best practices for enterprise AI factories and high-performance computing (HPC) workloads. A single 8U entry-level WEKApod configuration can support up to 1,152 GPUs, offering high-density, energy-efficient performance.
Nilesh Patel, Chief Product Officer at WEKA, said:
“WEKA Augmented Memory Grid is breaking the AI memory barrier, expanding GPU memory and optimising token efficiency. This breakthrough will enable faster AI innovation at lower costs without sacrificing performance.”
Rob Davis, Vice President of Storage Networking Technology at NVIDIA, added:
“The combination of NVIDIA and WEKA technologies will allow AI agents to process data with unprecedented speed and accuracy.”
Availability
The WEKA NCP reference architecture for NVIDIA Blackwell systems will be available later this month, while WEKA Augmented Memory Grid is set to launch in Spring 2025 for WEKA Data Platform customers.