A100 Rental Price Index — SDA100RT
Silicon Data A100 Rental Price Index
Daily benchmark for NVIDIA A100 hourly rental pricing, with separate neo-cloud and hyperscaler readings. The Neo-Cloud index is published daily as ticker SDA100RT.
ABOUT THIS INDEX
What the A100 Silicon Index tracks
The A100 Rental Price Index (Neo-Cloud ticker SDA100RT) is the Silicon Data benchmark for NVIDIA A100 GPU rental pricing. It is published daily as a standardized, like-for-like hourly rate expressed in USD per GPU-hour, drawn from observations across neo-cloud providers, hyperscalers, colocation markets, and private rental platforms. The A100 remains the workhorse GPU for inference pipelines, fine-tuning, and cost-sensitive training workloads.
The index separates neo-cloud and hyperscaler pricing into distinct readings so buyers can compare segments directly. Every observation entering the calculation is standardized for rental term length, cluster scale, and interconnect — the index reflects the underlying compute cost, not the headline rate on any single provider's website.
Operators sustaining A100 fleets for inference workloads use SiteIQ — the Silicon Data data center capacity-planning platform — to model deployable density, climate-adjusted PUE, and break-even utilization for ongoing capacity planning.
A100 is the most widely distributed data-center GPU in the rental market. Older fleets remain in active service across inference-heavy providers.
PRICE DRIVERS
What drives A100 rental prices
A100 rental prices are pulled between three forces. Training workloads that can afford higher per-GPU throughput increasingly migrate up to H100 and B200, softening A100 demand at the high end, while established inference and fine-tuning pipelines stay on A100 for its unit economics — a durable price floor even as frontier training moves up the stack. At the same time, as large operators retire A100 clusters in favor of newer accelerators, that released capacity re-enters the on-demand market and compresses hourly rates. Because the index reports a standardized hourly rate, these forces show up as movements in a like-for-like benchmark rather than in any single provider's list price.
Methodology
How the A100 index is calculated
The A100 Silicon Index is calculated daily from observations across cloud providers, colocation markets, brokered cluster sales, and private rental platforms. Observations are standardized for machine specs, rental terms, platform performance, and geolocation, then filtered for outliers and independently validated before publication.
RELATED INDICES
Other Silicon Indices
Frequently Asked Questions (FAQ)
The current A100 rental price is $1.29 per GPU-hour, based on the Silicon Data A100 Rental Price Index (Neo-Cloud ticker SDA100RT). The reading is a standardized, like-for-like rate across neo-cloud and hyperscaler providers, and the two segment readings are published separately on this page.
Hyperscaler on-demand capacity typically prices above neo-cloud and marketplace supply. The index publishes both segment readings so you can see the spread and how it moves.
New cluster availability and provider capacity additions, substitution pressure as other accelerators come online, and contract length (on-demand vs reserved). Every observation is standardized before it enters the index.
The realized, standardized hourly rental price of NVIDIA A100 capacity in USD per GPU-hour, drawn from neo-cloud providers, hyperscalers, colocation markets, and private rental platforms.
A100 rents well below H100 and B200 on a per-GPU-hour basis, reflecting its position as the prior-generation workhorse. The spread to H100 widens or narrows with how quickly training demand migrates to newer accelerators, while A100's floor is set by inference and fine-tuning workloads that stay on it for unit economics. All three indices are published on the same standardized, like-for-like basis, so the differences reflect real market pricing rather than methodology.
SDA100RT is published daily as a standardized, like-for-like hourly rate across tracked A100 providers; observations are standardized for machine specs, rental terms, platform performance, and geography, statistical outliers removed, and each day's value independently validated.
Get access to Silicon Data
Full access to every Silicon Index, the complete portal, and historical pricing — distributed via web and API.
Silicon Data Portal Access
7-day trial →Daily GPU and LLM market intelligence, built for institutional use.
- SiliconIndex™ Dashboard — all GPU, Memory & accelerator indices
- Market direction tracking
- PriceIQ predictive modeling
- Daily index values & 90-day historical pricing
- NeoCloud vs Hyperscaler comparison
- SiliconCarbon™