Stoa Markets: A Marketplace for GPUs and AI Servers with Verified Quotes

Stoa (stoaexchange.com) is a marketplace for buying and selling new and used GPUs and AI servers. The company, founded by Eren, Berat, and Kaan (YC S26), started brokering GPU deals manually and found the process fragmented—trades happen over phone calls, spreadsheets, and email threads, with no shared price discovery.
Key features from the source
- Structured RFQs: Buyers submit a detailed request (configuration, quantity, condition, warranty, location, delivery terms, inspection criteria) that is sent to KYB-verified dealers. Dealers return firm quotes without seeing each other's bids.
- Verified counterparties: Stoa performs know-your-business (KYB) checks on dealers, verifying company ownership and who is authorized to trade.
- Settlement tracking: Payment, shipping, delivery, and inspection are tracked through settlement, with evidence required at each step. Stoa does not take possession of hardware.
- Tiered fees: Stoa charges a tiered fee on completed trades, with lower fees at higher volumes.
- AI-assisted RFQ drafting: Users can describe hardware in plain language or attach a PDF, and AI drafts the RFQ while the user retains control.
Market traction
In its first month, Stoa received over $300M in requests for quotes (RFQs). The platform lists indicative prices for popular SKUs: A100 80GB at $9,490, H100 SXM5 at $25,860, H200 SXM at $33,850, B200 at $43,860, and GB200 NVL at $67,400. These are illustrative levels, not live market data.
Who it's for
Stoa targets GPU brokers, data center operators, AI labs and enterprises, cloud/neocloud providers, OEMs, VARs, SIs, resellers, liquidators, lenders, and funds involved in GPU trading or financing.
📖 Read the full source: HN AI Agents
👀 See Also

Flavian: A WordPress Development Framework with 24 Specialized Claude Code Agents
Flavian is an open-source WordPress development framework built around Claude Code, featuring 24 specialized agents for tasks like frontend development, security audits, and Figma-to-WordPress conversion. The creator found domain-specific agents significantly outperform general-purpose ones for WordPress development.

Memctl: Open Source MCP Server for Persistent Memory in AI Coding Agents
Memctl is an open source MCP server that provides AI coding agents with persistent memory across sessions, machines, and IDEs. Built primarily with Claude Code in two weeks, it stores project context and serves it back in subsequent sessions.

Benchmark: Gemma4 12B vs Qwen3 8B quantized on 24GB Mac Mini
A developer tested Gemma4 12B against Qwen3:8b-q4_K_M on a 24GB Mac Mini using two prompts. Qwen3 processed prompts 4-5x faster, while Gemma4 generated output slightly faster.

PRECC Tool Cuts Claude Code API Costs with Pre-Tool-Call Compression
A developer built PRECC, an open source tool that intercepts Claude Code tool calls and compresses payloads using RTK (Redundancy-aware Token Kompression), reducing input tokens by 40-66% with no perceptible latency impact.