Lambda needed to raise token throughput and improve AI-cloud economics, but fixed rack and facility power budgets were leaving GPU capacity stranded and limiting how many nodes could run safely at once.
Reported impact
Lambda increased cluster-wide token throughput by 24 percent, improved performance per watt by 23 percent, raised hybrid inference throughput by 20 percent, and lifted hybrid training throughput by 17 percent without expanding the facility power budget.
Lambda validated NVIDIA's DSX AI Factory Platform as an operational layer for power-constrained AI clusters rather than just a benchmark feature. On a five-rack HGX B200 environment, DSX MaxLPS and its Dynamic Power Software monitored GPU and rack-level draw in real time, then redistributed available power so more nodes could run under the same facility ceiling. Operators set policy targets such as 80 or 85 percent of maximum node draw without modifying active workloads, which allowed Lambda to pack additional inference and training nodes into the same footprint while keeping visibility through power-observability dashboards. Because mixed inference and training jobs have different power curves, the system could reclaim capacity that static provisioning leaves idle. That architecture turns power shaping into a control plane for AI-factory utilization, not just a hardware setting, and gives Lambda a path to lower cost per token as clusters scale.
Training and inference workloads are scheduled onto HGX B200 nodes, where Dynamic Power Software tracks live consumption and DSX MaxLPS applies per-policy power targets. The adjusted node budgets let Lambda run additional servers, and throughput and observability metrics feed back into cluster operations for ongoing tuning.
Dynamic Power Software (DPS) monitors GPU and rack-level consumption in real time, reallocating power to where it is needed to maximize token throughput within fixed budgets.
NVIDIA DSX MaxLPS enabled Lambda to add two inference nodes within the same power budget, increasing throughput by 20% in concurrent inference and training configurations.
NVIDIA DSX MaxLPS increased cluster-wide throughput per watt by 23% while operating within the same fixed power budget.
Using NVIDIA DSX MaxLPS, Lambda ran 19 nodes instead of 16 within the same power budget, increasing cluster-wide token throughput by 24% and GPU capacity by 19%.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
The Organization JSON-LD lists addressLocality "San Francisco", addressCountry "US", and numberOfEmployees value 106.
Related
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph NVIDIA Sample Sep 1, 2026 Onetera Onetera needed to help more than 90 municipalities turn fragmented regulations, zoning maps, permit rules, and legacy documents into fast resident-facing services without imposing heavy IT overhead or expensive manual… 1.7x increase in data throughput GPT OSS 20B · Government Intelligence Agent · LangGraph NVIDIA Sample Aug 12, 2026 CYRAN AI Solutions CYRAN needed to deliver interactive geospatial intelligence in air-gapped and sovereign environments, but JPEG 2000 decode latency on large satellite scenes was the bottleneck that kept edge analytics from being… 26 GB satellite scene NVIDIA DGX Spark · NVIDIA nvJPEG2000 · PixelFlux 