Corespan Blog
Explore insights, innovations, and perspectives from the Corespan team.
Disaggregated NVMe Scratch Pad: Breaking the GPU Memory Barrier
Corespan’s disaggregated NVMe scratch pad creates a shared, high-performance storage tier that extends GPU memory, enabling scalable AI workloads with better utilization and predictable performance.
Read more about Disaggregated NVMe Scratch Pad: Breaking the GPU Memory Barrier
Disaggregated GPU Memory Pools
Past 8 GPUs, network hops stall syncs and strand vRAM. Corespan disaggregates GPU memory into one photonic PCIe pool, so hosts draw needed capacity on demand—higher utilization, lower cost, at scale.
Read more about Disaggregated GPU Memory Pools
AI’s Second Wave: From Training Hype to Inference Reality
AI’s second wave shifts from model training to inference efficiency—optimizing cost-per-token and energy use. Dynamic, composable GPU fabrics unlock stranded capacity and maximize utilization.
Read more about AI’s Second Wave: From Training Hype to Inference Reality 
Dynamic AI Infrastructure for Energy
AI in energy often fails in the field due to legacy infrastructure, siloed data, and I/O limits. Corespan pools GPUs and NVMe into a composable PCIe fabric for low-latency, reliable AI.
Read more about Dynamic AI Infrastructure for Energy
Drut Becomes Corespan Systems
Drut Technologies is becoming Corespan Systems—a name that reflects our focus on intelligent compute cores and the high-bandwidth spans that connect them into unified, high-performance systems.
Read more about Drut Becomes Corespan Systems
Redefining ROI in AI Hardware: Not Every GPU Has to Be Expensive
AI ROI isn’t about buying the most expensive GPUs. It’s about utilization, right-sizing, and dynamic resource allocation to match real workload demands.
Read more about Redefining ROI in AI Hardware: Not Every GPU Has to Be Expensive
Beyond Static Servers: Dynamic Infrastructure for AI Efficiency
Static servers can’t keep up with AI. Dynamic, composable infrastructure improves utilization, reduces cost, and adapts in real time to changing workloads.
Read more about Beyond Static Servers: Dynamic Infrastructure for AI Efficiency