The new distributed system provides enterprise-grade AI computing at up to 70% lower cost by intelligently combining cloud and distributed resources
Today, Theta Labs announced the beta release of the hybrid edge cloud architecture of Theta EdgeCloud Network, a computing platform that combines traditional cloud-based GPUs with a distributed network of edge nodes operated by over 30,000 communities. The platform is designed to provide cost-effective access to high-performance computing resources for AI model training and inference, video processing, financial modeling, and other GPU-intensive tasks. This new release introduces a new decentralized GPU market that remains competitive and transparent across the platform.
Hybrid architectures address fundamental challenges in modern computing. The increased costs and limited availability of specialized hardware required for AI and machine learning tasks. Traditional cloud providers charge premium rates for GPU access, often price small research teams, startups and academic institutions. By integrating distributed computing resources from community members with traditional cloud infrastructure, the platform offers similar capabilities at significantly reduced costs.
Dynamic GPU Market for Efficient Computational Routing
Theta Edgecloud is a decentralized market that connects supply and demand for GPU computing power. Anyone with an idle GPU can provide resources and earn rewards, but it provides developers and AI teams with a scalable, cost-effective platform for running containerized workloads.
The platform allows customers to choose the best infrastructure for a wide range of computing tasks. For example, training large AI models that require substantial GPU VRAM can be directed towards powerful cloud or datacenter-based GPUs. In contrast, tasks such as inherently parallelizable burst model inference workloads are distributed to game machines operated by many communities, providing flexible and cost-effective alternatives.
To ensure a fair and dynamic pricing model, Theta EdgeCloud allows node operators (supply side) to set up their own hourly rental rates. On the other hand, users (demand side) can select nodes that meet performance requirements and budget constraints when launching workloads. This market-driven approach helps keep GPU pricing competitive and transparent across the platform.
The system includes a backup mechanism with automatic failover logic to reassign work when community devices go offline, ensuring reliable completion of computing tasks across heterogeneous networks of GPU type.
Power India to power major academic and enterprise customers
Currently, the platform serves customers such as Stanford University, Seoul National University, Kaisto, Yongpay University, University of Oregon, Michigan State University, and NTU Singapore. Enterprise clients include major sports teams such as the NHL’s Las Vegas Knights, the NBA’s Houston Rockets, the global eSports team’s Fly Quest, and the Evil Genius.
The beta release includes features that existing customers request, including persistent storage for AI model training, improved job prioritization, and a developer API interface for recruitment and analytics.
“The reality is that GPU costs are banned for many organizations doing research on important AI,” said Jieyi Long, CTO at Theta Labs. “The university says they’ll have to wait several months to scale back projects and access affordable computing resources. By tapping on unused GPU powers sitting on gaming computers and workstations around the world, they can offer the same calculation capabilities at a fraction of cost. Limit their ambitions.”
Technical capabilities
The hybrid architecture supports containerized computing tasks such as AI model training and inference, video encoding and transcoding, 3D rendering, financial simulation, and scientific computing applications. The platform offers over 80 petaflops of distributed GPU computing power through a combination of Google Cloud and Amazon web services, distributed edge networks and cloud partnerships.
EdgeCloud client nodes are lightweight software packages that allow community members to provide idle GPU capacity. Advanced job containerization ensures high efficiency calculations across a variety of GPU types and specifications. For more information about EdgeCloud client software, see this link.
About Theta Labs
Theta Labs is a leading provider of distributed cloud infrastructure for AI, media and entertainment with a global network of 30,000 distributed distributed edge nodes and a native blockchain. Supported by Samsung, Sony, Bertelsmann Digital Media Investments and Creative Artists Agency, Theta is one of the top 10 Depin blockchains in Coingecko’s market capitalization and Binance.com’s top AI tokens. Theta’s Enterprise Balidator and Governance Council is made up of global market leaders such as Google, Samsung, CAA, and Binance.
Recently launched Theta EdgeCloud is the first hybrid cloud edge edge computing AI platform with over 80 petaflops of on-demand distributed GPU computing power. Edgecloud currently counts 25 global customers, including top five universities in Korea, the Las Vegas Knights of the NHL, the Houston Rockets of the NBA, and the global eSports teams, including the global eSports team.
