DDN, the worldwide chief in AI and knowledge intelligence options, introduced continued progress in its collaboration with Nebul, a European chief in offering sovereign-hybrid cloud options, to optimize large-scale AI inference efficiency by superior KV Cache acceleration and high-performance knowledge infrastructure.
Additionally Learn: AiThority Interview with Matej Bukovinski, Chief Know-how Officer at Nutrient
The collaboration brings collectively Nebul’s AI inference platform, DDN’s Infinia knowledge intelligence structure, and NVIDIA accelerated computing applied sciences to deal with some of the vital challenges dealing with manufacturing AI environments: maximizing the financial return on AI infrastructure investments by greater GPU utilization, sooner token era, and decrease cost-per-token.
As AI adoption strikes from experimentation to manufacturing, organizations are confronting a brand new actuality: the problem is not whether or not AI works, however whether or not the economics work at scale.
Throughout the business, enterprises, cloud suppliers, and sovereign AI initiatives are more and more measuring success by enterprise outcomes akin to GPU utilization, cost-per-token, tokens-per-watt, and time-to-production. On this atmosphere, knowledge infrastructure has emerged as a important determinant of AI profitability.
“The AI dialog has basically modified,” mentioned Alex Bouzari, CEO and Co-Founder at DDN. “For years, the business targeted on buying GPUs. Immediately, the query is how effectively these GPUs generate worth. Inference has turn out to be the financial engine of AI, and decreasing the price of each token produced is now some of the vital challenges dealing with the business.”
As a part of an ongoing proof-of-concept engagement, DDN and Nebul are validating next-generation KV Cache acceleration capabilities designed to assist NVIDIA DSX-based AI manufacturing unit deployments by bettering inference effectivity, decreasing latency, and growing utilization of AI infrastructure.
“For years, the business targeted on constructing bigger fashions. Immediately, the problem is making these fashions economically viable in manufacturing,” mentioned Arnold Juffer, CEO at Nebul. “Each group is in search of methods to generate extra worth from its AI infrastructure investments. By our collaboration with DDN and NVIDIA, we’re demonstrating how KV Cache optimization and high-performance knowledge architectures can enhance inference effectivity, cut back latency, and assist unlock the following section of AI adoption.”
“AI infrastructure is more and more outlined by effectivity at scale,” mentioned Rod Evans, Vice President of Cloud Infrastructure at NVIDIA. “As organizations deploy bigger fashions and agentic AI workloads into manufacturing, applied sciences that enhance GPU utilization, cut back latency, and speed up token era turn out to be important. DDN continues to be an vital collaborator in advancing the information and infrastructure capabilities wanted to assist the following era of AI factories.”
Latest benchmarking efforts have demonstrated promising early outcomes, together with measurable enhancements in Time-to-First-Token (TTFT) efficiency with KV Cache enabled. The collaboration has efficiently accomplished RoCE-based infrastructure validation and continues to increase benchmarking actions throughout bigger inference sequence lengths whereas figuring out further optimization alternatives inside DDN’s Infinia platform.
The mission has additionally expanded to incorporate collaboration with NVIDIA round benchmarking methodologies, scalability validation, and future technical publications.
The Economics of AI Have Shifted
Trade consideration has traditionally centered on mannequin efficiency and GPU availability. Nevertheless, as organizations deploy AI into manufacturing, a brand new bottleneck has emerged: knowledge motion and inference effectivity.
Agentic AI workloads, retrieval-augmented era (RAG), and large-scale inference environments place unprecedented calls for on infrastructure, together with networking and storage. Each millisecond of latency and each share level of GPU idle time immediately impacts profitability.
“Coaching builds the asset. Inference is the place it earns,” mentioned Bouzari. “The following era of AI functions shall be powered by brokers that carry out significant enterprise transactions and selections. The economics of these methods rely upon infrastructure that may ship knowledge on the velocity AI operates.”
DDN’s Infinia platform was purpose-built to deal with these challenges by distributed KV Cache companies, GPU-native knowledge motion, clever knowledge orchestration, and high-performance storage architectures that maximize accelerator effectivity.
Accelerating the AI Manufacturing unit Period
The Nebul collaboration reinforces DDN’s broader imaginative and prescient that AI infrastructure should evolve past conventional storage architectures and turn out to be an energetic participant in AI execution.
The collaboration helps the broader NVIDIA DSX platform strategy to AI factories, the place compute, networking, storage, software program, and operations are designed collectively to enhance tokens-per-watt, cost-per-token, and time-to-production.
As organizations search to operationalize AI at scale, DDN believes the defining metrics of success will more and more turn out to be:
- GPU utilization
- Value-per-token
- Tokens-per-watt
- Time-to-first-token
- Time-to-production
Organizations that optimize these metrics will obtain sustainable AI economics. Those that don’t danger deploying infrastructure that continues to be underutilized regardless of vital funding.
“We don’t promote storage,” Bouzari added. “We assist organizations maximize the financial return on each GPU, each token, and each watt. That’s the inspiration of the AI financial system.”
Additionally Learn: AI methods – Interoperable AI methods: Connecting fashions throughout platforms
[To share your insights with us, please write to psen@itechseries.com]
