AlphaTON Capital has announced plans to invest significantly in high-performance graphics processing units (GPUs) and data center infrastructure to support Cocoon, a confidential compute open network introduced by Telegram founder Pavel Durov at Blockchain Life. The initiative targets secure, decentralized artificial intelligence (AI) inference, enabling application developers to compensate GPU owners with TON cryptocurrency for processing AI requests. Telegram is set to be the first major user of Cocoon, committed to leveraging the network for confidential AI queries and actively promoting its adoption across its user base.
AlphaTON Capital reports that it will deploy next-generation high-memory GPUs, specifically designed for AI inference workloads, within strategically located data centers. The infrastructure is intended to deliver high uptime and the processing power required for demanding AI models, including architectures like DeepSeek and Qwen. These technical resources aim to meet the requirements of data center operators supporting decentralized, privacy-preserving AI workloads originating from Telegram and other applications.
Brittany Kaiser, Chief Executive Officer of AlphaTON Capital, said, “We recognize the transformative potential of a decentralized, privacy-preserving AI inference network, and we are committed to becoming a cornerstone contributor to this ecosystem. Our investment in enterprise-grade GPU infrastructure will help ensure Cocoon has the computational capacity needed to serve Telegram and other major applications from day one.”
The Cocoon network is scheduled to go live in November 2025, following the onboarding of application developers and GPU providers. AlphaTON Capital states that its participation aligns with its ongoing focus on TON blockchain and Telegram-based AI applications, aiming to play a foundational role in the deployment of confidential AI solutions at scale.
Source: AlphaTON Capital
