Protopia and Rafay partner to enable secure multi-tenant AI infrastructure
Protopia AI and Rafay Systems announced a partnership to enable secure multi-tenancy for enterprise AI infrastructure using their respective 'Stained Glass' and 'Token Factory' technologies. The integration allows operators to securely share GPU capacity for sensitive inference workloads through metered, token-based access.
Key Takeaways
- Rafay’s Token Factory provides serverless, metered access to open-source models without requiring dedicated GPU allocation for each tenant.
- Protopia’s Stained Glass technology transforms plain text prompts into stochastic representations to protect data privacy at the inference layer.
- The integration aims to capture a share of the GPU-as-a-service market, which is projected to reach $7.36 billion in 2026.
- Data isolation allows infrastructure providers to convert governance and security requirements from operational costs into revenue-driving high-margin services.
Why It Matters
Multi-tenancy is shifting from a cost-saving measure to a strategic requirement for scaling enterprise AI. By isolating sensitive context at the software layer, providers can increase GPU utilization without exposing plain text data on shared hardware. For the streaming ecosystem, this facilitates the migration of compute-heavy tasks like real-time metadata generation and personalized content synthesis to third-party 'AI factories' without compromising proprietary viewer data. The combination of token-based metering and stochastic data transformations establishes a blueprint for sovereign AI environments. Watch for adoption rates among neocloud providers, who are using these specialized security layers to compete with hyperscalers on margin and performance.
Additional Context
The push for multi-tenant AI infrastructure arrives amid significant capital moves and technological shifts within the GPU-as-a-service market. Per Fortune Business Insights, July 2026, the global GPU-as-a-service sector is estimated to grow from $8.66 billion in 2026 to $162.54 billion by 2034. This growth is increasingly driven by 'neoclouds' and regional operators seeking to challenge hyperscaler dominance. For example, per Investing News, July 2026, NAVER, NVIDIA, and Brookfield recently announced plans to expand gigawatt-scale, multi-tenant AI cloud infrastructure in South Korea and the U.S., including a proposed expansion of the GAK Sejong data center to 200 megawatts.
Enterprise demand is specifically centering on hardware-based and software-defined 'confidential computing' to bridge the gap between data utility and privacy. Per Gartner's October 2025 reporting, more than 75% of processing operations in untrusted infrastructure are expected to be secured by confidential computing by 2029. This transition is verified by sector-specific momentum; per Forbes, July 2026, the emergence of agentic AI systems—which require continuous, high-volume token consumption—is making hardware-isolated environments essential for enterprises handling sensitive health, financial, or intellectual property data.
Infrastructure providers like Broadcom and VMware are also responding to this trend. Per Broadcom, October 2025, the VCF 9.0 platform now includes multi-tenancy capabilities specifically designed to enable secure private environments for AI on shared infrastructure. These developments underscore a market transition where raw compute capacity is being replaced by managed, secure, and metered AI services as the standard consumption model for large-scale enterprise deployments.
Read full article at siliconangle.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source