Ambient has developed a Layer 1 blockchain that replaces conventional proof-of-work hashing with verified artificial intelligence inference, creating a network where miners carry out AI tasks in exchange for processing opportunities.

The blockchain relies on a mechanism called Proof of Logits (PoL) to cryptographically verify AI inference results. Instead of requiring miners to solve traditional hashing puzzles, Ambient routes them to process AI queries submitted through an onchain auction system. According to the project, its public API has already handled more than 134 billion tokens.

The approach targets a growing challenge in AI infrastructure: establishing whether an output was genuinely produced by a specified model using the claimed inputs. This concern has gained prominence as organizations increasingly rely on AI models hosted by third parties, where the end user typically has limited visibility into how results were generated. Ambient's PoL system is designed to confirm that inference was performed correctly and that the resulting output has not been fabricated or altered.

The core innovation substitutes verifiable AI inference for traditional proof-of-work computation, allowing blockchain miners to contribute useful computational work while creating an onchain record of the tasks performed. That contrasts with proof-of-work networks such as Bitcoin, where the energy expended on hashing puzzles produces no output beyond securing the chain.

Competitive marketplace for AI inference

Ambient's network currently supports large open-weight AI models, including GLM 5.2 and Kimi K2.7 Code, and targets models with more than 600 billion parameters.

AI queries are submitted through a competitive bidding process in which miners compete on factors such as price and processing speed. The auction-based structure is intended to create a marketplace where inference providers compete to fulfill requests while the blockchain records the associated activity.

By placing inference and settlement on-chain, the system maintains a verifiable record of which miner processed a request and whether the resulting computation passed the network's verification process.

This structure could give developers and businesses an alternative to conventional centralized AI services, particularly for applications requiring greater transparency around how inference is performed.

SVM compatibility targets high throughput

Ambient is compatible with the Solana Virtual Machine, allowing it to use an execution environment already familiar to developers building within the Solana ecosystem. The network targets approximately 65,000 transactions per second, reflecting the high throughput demands of an AI inference marketplace.

The architecture is designed to support a large volume of AI requests while maintaining on-chain settlement and verification. That combination could become increasingly relevant as AI applications demand greater access to computing resources and organizations look for alternatives to centralized inference providers.

Founded in 2025 by Travis Good, Ambient has progressed from its initial concept through testnet development to a live environment with an operational public API.

Decentralized AI infrastructure without a native token

Ambient's proposition centers on three areas where centralized AI services can face limitations: censorship resistance, privacy, and vendor dependence.

The decentralized model distributes inference workloads among independent miners rather than concentrating processing within a single provider. The use of open-weight models is also intended to reduce reliance on proprietary AI platforms and their associated access restrictions.

By distributing AI inference across independent miners and applying cryptographic verification, Ambient aims to offer developers an alternative infrastructure model that reduces dependence on a single centralized AI provider.

Another notable feature is the project's reported lack of a native cryptocurrency token. Unlike many blockchain projects that introduce tokens as part of their economic model, Ambient appears to be emphasizing infrastructure and service usage instead.

The model suggests the project could generate revenue through API consumption and inference services rather than relying primarily on token appreciation. That approach may also align the network more closely with enterprise infrastructure models, where customers pay for computing or API access.

Ambient has raised $7.2 million in seed funding from a16z's Crypto Startup Accelerator, Delphi Digital, and Amber Group. The funding provides capital as the company develops its blockchain infrastructure, expands supported AI models, and works toward broader adoption of decentralized inference.

The project ultimately seeks to turn AI computation into economically useful blockchain work, combining decentralized infrastructure, cryptographic verification, and a competitive inference marketplace in a single Layer 1 network.