SHENZHEN – November 06, 2026 -- Tencent Cloud Enterprise (TCE) has launched a private cloud that starts at eight servers for full-stack deployments and three servers for AI-only entry, aimed at regulated industries that want to run agentic AI inside their own boundaries. The company announced the Full-Stack Intelligent AI Suite and Lite Deployment at the 2026 Tencent Global Digital Ecosystem Summit, alongside joint research with Gartner and Omdia.
Four-layer AI suite keeps data, model weights and inference traffic inside the customer boundary
The suite spans a converged compute, storage and networking foundation; a model service layer built on TokenHub, a private MaaS platform for LLM services; an agent runtime layer with a cloud-native sandbox, egress control and identity credentials; and an agent application layer. Customers can enter at any layer and expand upward.
TokenHub offers multi-tenant isolation and token-level metering down to individual users and API keys. TCE reports measured single-accelerator throughput above 250,000 TPM, cache hit rates above 90% and throughput gains of up to 220% on alternative silicon.
Unified training and inference scheduling raises heterogeneous compute utilization by up to 60%, according to the company. The suite runs on x86 and ARM processors and multi-vendor accelerators, and has been validated at 10,000-GPU scale. It delivers RTO of 2 minutes or less and RPO of 0.
"Full-stack intelligence is the capability we build. Sovereign AI is the boundary we keep," said Kaibo Sha, Vice President, Tencent Cloud.
Lite Deployment delivers a full-stack cloud from three master, three block storage and two network nodes
Lite Deployment targets staged budgets and pilot-first projects. The eight-server configuration keeps core IaaS capability intact, and the AI-only entry point starts from three servers with private model services out of the box. The architecture scales from a single availability zone to metro dual-AZ active-active and then to cross-region disaster recovery.
It is delivered as software on existing hardware, with no capital-heavy refresh and no minimum spend. One of the largest futures brokerages by brokered turnover and client assets runs core trading systems on a TCE private cloud across heterogeneous chip architectures. It started AI on three servers and extended as volumes grew.
Gartner guide projects over 25% of regulated enterprises will repatriate GenAI workloads by 2028
TCE and Gartner jointly released A Practical Guide to Building Sovereign AI Cloud Platforms. It frames sovereignty as a spectrum resting on three pillars: data, operations and technology. It sets out adoption paths for financial services, healthcare, public service cloud and telecommunications.
Gartner research cited in the guide projects that strict data sovereignty and privacy rules will compel over 25% of enterprises in regulated industries to repatriate primary GenAI workloads on-premises by 2028. Through 2029, dependence on proprietary foundation models will lock over 50% of GenAI adopters into a single public cloud provider's ecosystem. Only 17% of organizations report enough talent to support AI adoption.
Omdia projects $160 billion in telco-led AI data center investment from 2026 to 2031
The Omdia whitepaper, Telco Cloud and AI Transformation: Global Market Opportunities and Case Studies, puts that figure at 9% of combined fixed and mobile capex. It cites mobile ARPU declining roughly 2% annually while OTT players capture more than 70% of network traffic value.
Omdia identifies three trends: full-stack industry cloud platforms, national cloud infrastructure, and AI monetization beyond GPU-as-a-Service toward MaaS and token-based business models. It recommends that operators act now on the sovereign AI window and choose software-independent, hardware-open full-stack partners with proven white-labelling capability.
Thai provider launched two TCE data centers in eight weeks after leaving VMware
The Thai digital infrastructure provider now runs a white-labelled, metered cloud business. Other customers presented at the session were Shanghai Xuhui District, on a multi-chip government cloud with tidal compute scheduling, and Shenzhen Second People's Hospital, on a hospital-wide foundation for agentic AI. A large, diversified Hong Kong organization is migrating from VMware Tanzu to a cloud-native hybrid cloud on TCE.