AIxBlock

AIxBlock launched in 2024 as a decentralized end-to-end AI development platform built on Solana, but has since undergone a significant pivot: by mid-2025 the company had abandoned its blockchain roadmap in favor of becoming an enterprise-focused AI training data provider backed by the European Union.

Origins on Solana

The project was conceived in 2022 and accepted into NVIDIA's Inception Program early in its development. By June 2024 the team — co-founded by Swiss Dao (CEO), Tony Dinh, and Jackie Lu and headquartered in Sunnyvale, California — launched a public platform positioning AIxBlock as a comprehensive on-chain toolkit for AI developers.

At launch, the platform's core premise was that AI model development is prohibitively expensive and centralized. AIxBlock proposed a three-sided marketplace on Solana to address this:

• Compute marketplace: GPU and compute resources sourced peer-to-peer from individual providers, with the platform claiming savings of up to 90% versus centralized cloud providers.
• Dataset marketplace: Crowdsourced and curated training datasets available for purchase or contribution.
• Human validator network: A distributed workforce for data labeling, annotation, and quality assurance.

Developers could use the platform to train models from raw code, fine-tune foundation models, deploy existing models, annotate data, or contribute compute in exchange for rewards. The platform supported diverse AI modalities including computer vision, automatic speech recognition (ASR), and large language models. A self-hosted deployment option gave users the ability to keep sensitive data within their own infrastructure while still accessing the decentralized resource pools.

Blockchain Architecture and Token Plans

The platform ran on the Solana blockchain, with peer-to-peer transactions designed to be zero-fee — a key differentiator from cloud-based competitors. A Token Generation Event (TGE) was planned for Q2 2025, which would have introduced a native token to power the ecosystem's incentive and governance layer. The token was intended to reward compute providers, dataset contributors, and validators, and to enable platform governance.

In January 2025, AIxBlock announced a strategic partnership with Hivello, a platform simplifying node management for non-technical users. The partnership was designed to integrate Hivello's distributed GPU resources with AIxBlock's enterprise-level compute marketplace, broadening the pool of decentralized compute available to developers and lowering the technical bar for node participation.

Also in early 2025, AIxBlock reported a $1 million per year enterprise contract — evidence that even in its blockchain phase, the project was targeting large commercial clients rather than retail crypto users.

The Pivot: From DePIN to Enterprise Data

The planned Q2 2025 TGE did not materialize on any public exchange, and no AIxBlock token is found in major market data. By mid-2025, the company had repositioned entirely. The current website — retitled "Enterprise Training Data for Speech & LLMs" and subtitled "Sovereign AI Data Platform" — carries no mention of Solana, blockchain, decentralized compute, or a native token. Blog posts about decentralized AI development that previously appeared on the site have been removed.

The pivot coincided with the company receiving a EUR 1.74 million innovation grant from the European Union in August 2025, which appears to have reoriented the company's strategy and messaging toward regulated enterprise clients in Europe and the United States.

Current Business

As of mid-2026, AIxBlock operates as a specialized AI training data provider with roughly 73 employees and a client portfolio that includes Oracle, AWS, Apple, Uber, Uniphore, and Chainalysis. The company claims seven years of enterprise experience and more than 200,000 hours of audio delivered across 100-plus languages — figures that suggest the speech data capabilities predate the blockchain phase.

The current product suite includes:

• Speech Data Services: Large-scale collection, transcription, and annotation of voice recordings across 100-plus languages and accents, used to train voice AI, ASR systems, and speech analytics platforms.
• Sound and Environment Audio: Real-world noise, environmental audio, and acoustic scene recordings for classification, detection, and recognition models.
• Text and Dialogue Data Services: Conversation annotation, named entity recognition, RLHF preference data, and fine-tuning datasets for LLMs.
• OTS Audio Library: Hundreds of thousands of hours of pre-licensed call center audio from US, India, and Philippines sources, available for bulk licensing.
• Self-Hosted Platform: A deployable AI development environment including a data pipeline, training tooling, and GPU marketplace that clients run on their own infrastructure — the one remaining feature from the original decentralized platform.

Quality assurance is a central differentiator in the current positioning. The platform uses a consensus mechanism to establish quality baselines, continuous blind testing to detect drift, a three-tier QC structure (QA, QC, QC2), biometric KYC onboarding for contributors, and behavioral anomaly detection to flag automation and proxy work. Multi-layer contributor verification is presented as an institutional-grade safeguard for clients in regulated sectors such as banking, healthcare, and government.

The self-hosted option allows enterprise clients to connect their own storage from day one, so data collected on their behalf flows directly to their infrastructure without AIxBlock retaining any copy. This data sovereignty angle is positioned as a direct response to enterprise security concerns about proprietary data being retained or resold.

Ecosystem Fit and Solana Relevance

In its original form, AIxBlock was an ambitious DePIN-adjacent project: applying Solana's low-cost, high-throughput infrastructure to the problem of decentralized AI compute and data. The model was similar in concept to other Solana AI infrastructure projects, proposing to disaggregate the expensive, centralized compute stack that underlies modern AI development.

That Solana-native strategy did not survive into 2026. No token launched, the decentralized compute marketplace is not publicly described anywhere on the current site, and the Hivello partnership announcement appears to have been the project's last public blockchain-related activity. The company is now an enterprise B2B data firm with EU institutional backing and has dropped its crypto-native positioning.

AIxBlock's trajectory illustrates a pattern visible across several early-cycle AI-blockchain projects: a strong product insight — AI training data and compute are expensive and centralized — paired with blockchain token economics that did not ultimately reach market, followed by a pivot to a conventional enterprise software and services model using the underlying capabilities.