NVIDIA outlines how AI factories boost return on investment with productivity, durability and flexibility
NVIDIA’s blog post explains how its AI factories maximize return on investment through three key factors: earning capacity, useful life, and demand. Each megawatt-scale factory costs about $60 million, and operators prioritize clear ROI before committing capital. NVIDIA’s AI factories are designed to be productive—delivering the highest throughput per megawatt and lowest cost per…
Key points
- NVIDIA AI factories cost about **$60 million per megawatt** and focus on productivity, durability, and flexibility to maximize ROI
- **Vera Rubin NVL72** systems offer **30x higher throughput per megawatt** and **45x lower token costs** than **GB300 NVL72** for **DeepSeek V4 Pro**
- **CUDA-X libraries** enable running AI and non-AI workloads across generations of NVIDIA GPUs, extending hardware’s useful life
The company highlights NVIDIA Vera Rubin NVL72 systems, which deliver over 30x higher throughput per megawatt than NVIDIA GB300 NVL72 and up to 45x lower cost per million tokens for the DeepSeek V4 Pro model. NVIDIA’s CUDA-X libraries enable running any accelerated workload, while continuous software optimization keeps installed hardware productive for years. The blog also notes that A100 GPUs, introduced in 2020, remain in commercial use six years later, with operators extending depreciation schedules. A September 2026 Sprout analysis estimates useful life for an eight-GPU H100 system at five to six years and for GB300 NVL72 at nine to ten years, based on resale value.
The story so far
3 episodes →- NVIDIA outlines how AI factories boost return on investment with productivity, durability and flexibilitythis story
Productive, Durable, Fungible: How NVIDIA AI Factories Maximize Return on Investment
NVIDIA Blog · 1 October 2026
Loading the full article…
This text was published by NVIDIA Blog and written by Shruti Koparkar. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- Alphabet to launch Google AI chips into orbit on SpaceX Falcon 9 · 2 src
- Nvidia-backed Nebius buys Inferize to cut AI inference GPU costs · 1 src
- Deepseek releases open-source TileLang for Huawei Ascend chips · 3 src
- Legato launches AI hearing glasses for $999 · 1 src
- Satlyt raises $8M to run AI on satellites with Google’s Gemma model · 1 src
Comments
via GitHub Discussions