Ponsbloom is a decentralized AI inference network that connects AI workloads with idle Apple Silicon computers. Its service offers private, hardware-verified compute through an OpenAI-compatible API, allowing developers to use existing OpenAI SDKs for tasks such as text generation, function calling, streaming, and speech-to-text. The network supports models including Gemma 4, Qwen 3.5, Nemotron 3.5, and GPT-OSS.

Ponsbloom emphasizes end-to-end encryption, Apple Secure Enclave-based hardware attestation, and signed responses that can be independently verified. The company says operators cannot access request data and that its hardened runtime restricts debugging and memory inspection. Its per-token pricing has no subscription requirement, with listed rates positioned below centralized API alternatives.

For Mac owners, Ponsbloom provides an agent that enrolls and attests a machine, serves models, and schedules compute around user preferences, such as idle time or power availability. Operators can track usage and earnings in a console, with the service stating that they keep 95% of revenue and pay no platform fee. Ponsbloom is presented as a research preview.