Case study: RentServers.com
Autonomous GPU and bare-metal clearinghouse
A GPU and bare-metal infrastructure marketplace engineered as an autonomous deal desk: Rust/Axum server-rendered quote ingestion, a 36-month TCO normalization engine, and automated multi-vendor inventory clearing.
The problem
GPU and bare-metal pricing is rarely public and changes week to week. Buyers cannot compare a reserved cloud instance against a dedicated server and a colocation quote, because every vendor prices setup, IP addresses, bandwidth and term commitments differently. The platform had to turn incompatible quotes into one comparable number, automatically.
Architecture
- 01
Ingestion
Rust/Axum services ingest buyer requirements and provider offers into a normalized catalog of GPU, dedicated, VPS and colocation inventory, rendered server-side.
- 02
Normalization
Each quote becomes one 36-month total: the monthly rate held for 36 months, plus setup fees, IP addresses and bandwidth overage. Offers with different terms and fee structures land on the same basis.
- 03
Clearing
Inventory from multiple vendors is matched against a single normalized specification, so a buyer sees comparable offers instead of incompatible price sheets.
- 04
Agent surface
A remote MCP server and llms.txt expose price benchmarks and the 36-month TCO calculation, so AI assistants answer infrastructure questions with real numbers.
- 05
Execution
Designed, built and shipped to production in under 14 days, with CI/CD, monitoring and server-rendered pages delivered in under 40 ms.
Components
- Rust/Axum quote ingestion
- Buyer requirements and provider offers ingested and rendered server-side in Rust: fast, indexable and inexpensive to run under crawler and agent traffic.
- 36-month TCO normalization
- Every quote converted to one 36-month cost that includes setup fees, IP addresses and bandwidth overage, so cloud, reserved, bare metal and colocation compare directly.
- Multi-vendor inventory clearing
- Offers from multiple providers matched against one normalized specification, so buyers compare like for like and suppliers quote without another portal.
- Native MCP agent discovery
- A remote MCP server and llms.txt let AI assistants query live price benchmarks and run the 36-month TCO calculation directly.