DI

DI (Didotxyz) is a decentralized AI infrastructure platform that aggregates global GPU resources into a unified computing pool. Its core business is providing OpenAI-compatible API interfaces, allowing developers to access models like Llama and DeepSeek at costs 60-85% lower than centralized clouds such as AWS or Azure. The platform also operates AI Arena for model evaluation and a PoQ (Proof of Quality) mechanism for on-chain verification of inference outputs.

The market pain point DI addresses is the structural shortage of GPU supply: major cloud providers control over 60% of computing resources, with delivery wait times stretching to months. DI unlocks idle consumer and gaming GPUs, making small open-weight models deployable at roughly half the cost of traditional clouds.

In the past six months (2026-03-04 to 2026-09-04), DI's API has processed over 1 billion inference requests, serving more than 15,000 paying users and generating $23 million in economic revenue. Its AI Arena has attracted over 500,000 participants for model benchmarking. However, like the broader DeAI sector, DI faces governance challenges—token-driven development incentives can misalign with long-term research goals, and the industry has seen disputes (e.g., Bittensor subnet exits) that highlight unresolved coordination issues.