How InstaDeep improved its R&D and smoothed processes with Dell AI Factory with NVIDIA

TL;DR

• InstaDeep increased computational performance and design speed by 10x each using Dell AI Factory with NVIDIA
• InstaDeep reduced costs 30% by using Dell AI Factory with NVIDIA

InstaDeep, a leader in advanced manufacturing and Industry 4.0, partnered with Dell Technologies, NVIDIA, and SCC France to scale its compute infrastructure. By deploying the Dell AI Factory with NVIDIA, InstaDeep powered its proprietary DeepPCB software—an agentic AI tool that fully automates the design, placement, and routing of printed circuit boards (PCBs). The upgrade eliminated infrastructure bottlenecks, allowing them to scale large-scale AI models and accelerate industrial R&D.

Key outcomes & business impact

• 10x Design Speed: Cut PCB placement and routing design time tenfold using reinforcement learning .
• 10x Compute Power: Multiplied computational performance by 10x to drive rapid AI innovation.
• 40% Faster Training: Boosted AI model tuning and training efficiency by up to 40% .
• 30% Cost Reduction: Reduced infrastructure and operational costs through optimized resource utilization.

The tech stack behind the innovation

• The Infrastructure: The Dell AI Factory with NVIDIA, built on Dell PowerEdge XE Series Servers and NVIDIA Spectrum Networking.
• The Supercomputing Cluster: Named Kyber, this cluster is powered by NVIDIA H100 GPUs, achieving ~0.5 exaFLOPs in FP16 performance and ranking among the top 100 global AI clusters.
• Strategic Deployment: SCC France acted as the end-to-end integration partner, ensuring seamless deployment of the high-performance computing (HPC) systems so InstaDeep's engineers could focus purely on innovation .

What’s next?

"This will be the year of smart agents – systems that operate in complex environments and make decisions," says Karim Beguir, co-founder and CEO of InstaDeep

Moving forward, InstaDeep is expanding its "smart agent" reinforcement learning models beyond electronics into advanced fields like humanoid robotics and complex autonomous industrial tasks. Read the full case study here.