The Artificial Super Intelligence (ASI) Alliance said ASI:Cloud has begun processing live workloads, marking a significant step in its push to provide decentralized, enterprise-grade AI infrastructure to developers and researchers.
The platform offers permissionless access to GPU compute and AI inference services, positioning itself as an alternative to centralized cloud providers at a time of rising demand for production AI.
ASI:Cloud is designed to simplify access to high-performance AI infrastructure by unifying compute, billing, and developer experience within a single platform.
According to the alliance, this approach addresses growing frustration among AI builders navigating fragmented vendor ecosystems and escalating costs in traditional cloud markets.
A unified, permissionless AI infrastructure
The launch comes as enterprises increasingly seek to regain control of their AI workloads.
Many are looking to repatriate compute from centralized providers due to cost volatility, capacity constraints, and vendor lock-in.
ASI:Cloud aims to meet this demand by offering a decentralized alternative that allows developers to authenticate using Web3 wallets without KYC requirements.
The platform supports payments in FET tokens and stablecoins, with fiat payment options planned for the future.
Transparent pricing is a core feature, according to the alliance, allowing teams to better predict costs compared with traditional cloud environments that often add unpredictable fees for bandwidth, storage, and data egress.
Ben Goertzel, CEO of SingularityNET and the ASI Alliance, described ASI:Cloud as a transition from vision to execution.
He said each GPU cycle processed on the platform contributes to the broader goal of sustainable, decentralized intelligence.
Competitive pricing and open-source model access
ASI:Cloud provides OpenAI-compatible inference endpoints across several leading open-source models, including Llama 3.3 70B, Qwen 3 32B, and Gemma 3 27B.
Pricing starts at $0.07 per million input tokens, positioning the service as a lower-cost option for developers running inference workloads at scale.
This contrasts with centralized cloud platforms, where operational instances of Nvidia’s H100 GPU typically cost between $3.90 and $6.98 per hour on providers such as AWS and Azure.
Those rates often exclude additional charges tied to data transfer and storage, which can significantly increase overall spending.
The alliance argues that these cost dynamics have become a major barrier to experimentation and deployment, particularly for smaller teams and independent researchers.
Collaboration amid GPU shortages
ASI:Cloud is built through a collaboration between SingularityNET and CUDOS.
SingularityNET provides the AI infrastructure backend and model optimization, while CUDOS operates the underlying enterprise-grade compute infrastructure across a global network.
Together, they aim to deliver dependable access to high-performance GPUs without the bottlenecks common in centralized cloud environments.
Luke Gniwecki, Head of AI Compute Product at CUDOS, said enterprises are facing real constraints as demand for AI accelerates, citing capacity shortages and vendor lock-in as key challenges.
He said the combined offering is intended to give builders consistent access to compute resources as legacy providers struggle to meet demand.
The timing is notable. Nvidia’s latest GPU allocations are reportedly sold out through 2026, with waitlists extending into next year.
Against this backdrop, the decentralized computing market is projected to grow substantially, reaching an estimated $45 billion by 2035.
Developers and researchers can access ASI:Cloud via asicloud.cudos.org, with documentation available at docs.cudos.org.
The post ASI Alliance brings decentralized GPU infrastructure online with ASI:Cloud appeared first on Invezz
