375ai

TLDR: 375ai is a DePIN project on Solana that deploys edge sensing hardware and a crowdsourced mobile app to collect structured, real-world multimodal data for AI training, paying contributors in its native $EAT token. Launched on Solana mainnet in October 2025, the network had already logged 2.7 billion cumulative events and 300,000-plus app downloads before going live.

Artificial intelligence development has a structural data problem: the overwhelming majority of frontier AI training data is synthetic, generated by AI systems rather than captured from direct observation of the physical world. 375ai was founded in 2023 to address this gap, building a decentralized data layer that continuously captures labeled, multimodal real-world data and makes it accessible to AI researchers and enterprise buyers.

The project launched its mainnet on Solana in October 2025, joining a growing category of DePIN (Decentralized Physical Infrastructure Networks) that use on-chain token incentives to coordinate real-world hardware contributions at scale.

How the Network Works

375ai operates three distinct data collection channels, each targeting a different segment of the physical environment.

375edge is the network's flagship hardware node — a fixed sensing unit built for installation on premium urban infrastructure. Through a partnership with Outfront Media, which manages more than 40,000 billboard locations reaching roughly 70 percent of the American population, 375ai has access to a national grid of high-traffic vantage points for sensor deployment. Each 375edge device carries six 4K cameras, ambient microphones, environmental sensors, and an NVIDIA-powered onboard GPU. Processing happens entirely at the edge: rather than transmitting raw video streams to the cloud, each device distills approximately one terabyte of raw daily input into roughly 35 megabytes of structured insight data, a compression ratio of around 26,000x. The output includes frame-level object tracking across more than 100 categories, motion vectors, transition dynamics, and spatial geometry with known vantage coordinates. Privacy protections are built into the hardware pipeline: faces and license plates are anonymized on-device before any data leaves the unit, with no facial recognition performed. The architecture is designed to comply with GDPR and CCPA.

375go is a mobile app that enables crowdsourced data collection for environments fixed infrastructure cannot reach — retail shelves, building interiors, points of interest, and outdoor spaces in markets beyond the US. Contributors use smartphones to capture labeled data on demand and verify ground truth for model predictions. By the time of mainnet launch, the app had attracted more than 300,000 downloads across more than 170 countries, with 200,000 nodes actively participating during the Discovery Testnet phase.

375x, launched in April 2025 in partnership with AYDO, extends the network by ingesting data from third-party IoT devices. AYDO's infrastructure connects to more than 7,000 device types, allowing owners of existing cameras, environmental sensors, and audio equipment to route their data streams into the 375ai network in exchange for token rewards. An incentivized pilot launched with 40 select devices during the testnet phase.

Together the three channels give 375ai multi-format, multi-vantage coverage of the physical world: purpose-built urban infrastructure, mobile human-driven collection, and the broader third-party IoT ecosystem.

Data Products and Use Cases

Enterprise buyers receive the network's structured outputs via S3, Parquet files, or streaming API. The data products include continuous temporal sequences with labeled geometry — a format suited for training autonomous systems, logistics models, and world models that need to understand how real-world environments change over time. Target verticals include autonomous vehicles, CPG retail intelligence, logistics and freight, government analytics, advertising and media, and physical AI research.

At mainnet launch in October 2025, the network reported 2 million vehicles observed per day, 2.7 billion cumulative events logged, and more than 13,000 hours per week of labeled video produced.

Token: $EAT

The network's native token is the Edge AI Token ($EAT), distributed as a reward to participants contributing data through 375go, 375edge deployments, or 375x IoT connections. Mainnet launched on Solana in October 2025, following the conclusion of the Discovery Testnet. Testnet participants received an airdrop of $EAT based on a snapshot taken prior to the mainnet transition. Detailed tokenomics — total supply, vesting schedule, and allocation breakdown — have not been publicly disclosed in any sources reviewed for this profile.

Partnerships

Beyond the Outfront Media infrastructure agreement, 375ai has announced several protocol-layer collaborations:

• Akave — partnership to build verifiable edge intelligence, enabling cryptographic attestation of data provenance at the point of collection
• Irys Network — integration to make collected real-world data verifiable on-chain, anchoring the network's data integrity pipeline
• Wingbits and Beacon Protocol — collaboration to connect physical-world data streams to downstream AI consumers
• AYDO — the IoT integration layer underpinning the 375x product line

Team and Funding

375ai raised $10 million in October 2025 in a round led by Delphi Ventures, Strobe Capital, and HackVC. The executive team includes Harry Dewhirst (CEO), Rob Atherton (COO), Trevor Branon (CBO), and Chad Partridge (Chief AI Officer). No publicly disclosed security audits of smart contracts or edge device firmware were found in the sources reviewed.

Solana Fit

375ai sits squarely in the physical DePIN category that has become one of Solana's most active growth sectors. The network's reward structure — distributing $EAT tokens to contributors of data across hardware nodes, mobile applications, and third-party IoT devices — follows the DePIN coordination model that Solana's throughput and low transaction costs are well suited to support. The project's combination of purpose-built edge hardware, a large consumer-facing mobile app, and third-party IoT integration gives it a broader data acquisition surface than single-modality DePIN networks, positioning it as infrastructure-layer tooling for the physical AI training pipeline rather than an application targeting end users directly.