Observed Signal · Sep 14, 2026 · Technical Release · Source: SemiAnalysis · Impact: 5/5 · Sentiment: Positive
Infrastructure Market: Rubin NVL72 Agentic Inference: 67x Better Performance per Dollar
SemiAnalysis reports first verified agentic inference results for NVIDIA's Rubin NVL72 platform using their AgentX benchmark. Even on early pre-release software, Rubin delivers up to 67x better performance per dollar of TCO compared to GB300 in certain configurations, and significantly higher throughput per MW. The analysis projects Rubin can generate over 2x more profit per gigawatt than Blackwell, with revenue and profit advantages of 39% and 42% respectively at a fixed power budget. Dynamic power shifting (DSX MaxLPS) allows more GPUs per datacenter footprint. The article highlights Rubin's superiority over H200 and MI355X, with recommendations for inference providers to adopt Rubin for cost-efficient token generation.
Technical release from major platform (NVIDIA) showing significant performance gains, which will impact AI infrastructure decisions industry-wide.
Key Takeaways & Evidence Grounding
- Rubin NVL72 achieves up to 67x the throughput per TCO of GB300 in specific scenarios.
- Rubin delivers up to 7x better token throughput per MW than Blackwell in real-world tests.
- Rubin can generate over 2x more profit per gigawatt than Blackwell.
- Rubin generates approximately 39% more revenue and 42% more profit per GW than GB300.
- 3-year rental cost for Rubin was over $8.5/hr/chip in July 2026.
Connected Companies & Entities
8 Entities mappedSemiAnalysis
AI infrastructure and semiconductor research, data models, tools and consulting.
“SemiAnalysis publishes the benchmark results and analysis....”
Meta
Consumer internet platforms monetised through advertising, apps, subscriptions and VR.
“Meta is mentioned as a buyer of compute and supporting the benchmark....”
NVIDIA
Accelerated computing company spanning AI software, cloud and gaming.
“NVIDIA's Rubin platform is benchmarked, with GTC 2026 presentation by Jensen....”
AMD
US semiconductor company designing processors and computing chips.
“AMD is collaborating on MI455X UALoE72, and MI355X is benchmarked....”
DeepSeek
LLM developer offering AI chat and API access.
“DeepSeek V4 Pro model is used as the workload for benchmarking....”
Microsoft Azure
Enterprise cloud infrastructure, data and AI platform.
“Microsoft Azure is mentioned as a buyer of compute and supporting the benchmark....”
Oracle
Enterprise cloud infrastructure and customer experience software provider.
“Oracle is mentioned as a buyer of compute and supporting the benchmark....”
Google Cloud
Enterprise cloud, data and AI services for businesses.
“Google Cloud is mentioned as a buyer of compute and supporting the benchmark....”
Ontology Mapping & Concepts
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.