AlphaTON Capital Corp. (Nasdaq: ATON) has taken a major step in decentralized AI infrastructure by securing $82.5 million to deploy more than 1,000 Nvidia B200 GPUs. The investment follows the successful November 25, 2025 pilot of Telegram’s Cocoon AI, showing the company’s readiness to scale privacy-first compute for a platform with more than a billion users.
The AlphaTON GPU cluster adds $70 million in new Nvidia hardware to the balance sheet backed by $30 million in equity and $52.5 million in debt amortized over 36 months. This approach builds financial flexibility while creating an asset base designed to generate revenue from dedicated Cocoon workloads and external AI customers through the CUDO Compute marketplace.
The deployment model surrounding the AlphaTON GPU cluster pulls together experienced partners capable of delivering production-grade AI infrastructure from day one. CUDO Compute, a certified Nvidia Cloud Partner, manages orchestration, workload scheduling, monitoring, and commercialization.
Their platform opens access to AI developers and enterprises seeking alternatives to oversubscribed cloud providers. SNET Energy UK Ltd brings sustainable data center engineering to ensure optimal power, cooling, and energy efficiency.
Vertical Data adds financing, GPU sourcing, and deployment support to keep the rollout on schedule. This partnership structure is designed to reduce operational risk while providing a scalable, modular cluster suitable for high-throughput inference and training.
Cocoon AI’s requirements highlight why the AlphaTON GPU cluster is positioned as sovereign compute rather than another cloud vendor extension. Built within Telegram’s enormous ecosystem, Cocoon emphasizes privacy, encrypted data pipelines, and independence from Big Tech oversight.
Traditional cloud providers control infrastructure, access, and content policies, raising concerns around censorship, surveillance, and commercial exploitation of user data. AlphaTON’s approach flips this paradigm by offering dedicated capacity with strict data sovereignty.
The cluster processes workloads in a decentralized environment where ownership, rights, and access remain transparent and enforceable. This aligns with the project’s ethical AI posture and its aim to prioritize user rights over centralized monetization models.
Financial projections for the next five years paint a confident picture supported by detailed assumptions. Management estimates a 59.7% internal rate of return, $59.6 million net present value, and more than 600% return on investment.
EBITDA margins are projected between 64% and 73%, driven by high-demand GPU rentals and priority access workloads for Cocoon. These projections assume 90% utilization and a 20% price decline in the first year, reflecting typical compression in AI compute markets.
If targets are met, total cash returns could exceed $150 million through a combined model of dedicated compute, bare-metal rentals, marketplace engagement, and value-added AI services. The structure gives AlphaTON resilience even if market conditions shift, since debt service remains covered above the required threshold.
The company’s move into large-scale compute mirrors growing global demand. McKinsey expects data center capacity to more than triple by 2030, with most value captured by companies that own physical GPUs rather than resellers or brokers.
Adding the AlphaTON GPU cluster to the balance sheet continues the firm’s strategy of holding digital and physical assets that deliver recurring revenue. This builds on its $27 million digital treasury within the TON ecosystem, staking operations, and validation capabilities.
As AI usage expands across finance, communications, healthcare, entertainment, and national infrastructure, platforms with sovereign compute become strategic rather than optional. AlphaTON’s model gives public investors Nasdaq-listed exposure to decentralized compute without requiring direct crypto participation.
The Cocoon pilot demonstrated that AlphaTON’s architecture can support real workloads. Data flows, inference performance, load patterns, and privacy protections were validated during the test phase.
This cleared the way for closing the financing in Q4 2025. Hardware shipments, installation, and benchmarking are planned for Q1 2026, with full Cocoon and marketplace workloads expected by Q2.
Expansion options remain on the table if demand exceeds baseline capacity or if Telegram scales Cocoon across new geographies. CEO Brittany Kaiser has stressed that the goal is to build ethical infrastructure for Telegram’s massive user base, ensuring AI access unlinked from surveillance capitalism.
Risks remain, including hardware supply constraints, utilization shortfalls, debt timing pressures, and volatility in the broader TON ecosystem. AlphaTON acknowledges these factors but notes that debt coverage ratios remain above 1.0x even under downside scenarios.
The firm’s focus on AI, decentralized compute, and TON-aligned business lines differentiates it from companies that attempt to span unrelated industries. Its strategy positions it as a bridge between public capital markets and decentralized AI infrastructure at a time when global demand is accelerating beyond what centralized cloud providers can supply.
This deal highlights the broader rise of decentralized AI, where privacy, ownership, and independence increasingly challenge the dominance of legacy hyperscalers. The AlphaTON GPU cluster could become a defining example of how compute for billion-user applications transitions away from centralized control and toward sovereign infrastructure built for trust, transparency, and user rights.
Catch the pulse of decentralized AI funding and infrastructure at ainewstoday.org, your gateway to tomorrow’s breakthroughs!
