RE:AI, the sovereign AI cloud unit of Singtel Digital InfraCo, signed a memorandum of understanding with WEKA, a global AI data and memory infrastructure provider, to jointly build sovereign AI infrastructure across Singapore and the broader Asean region.
The partnership pairs Singtel's RE:AI cloud platform, Paragon network orchestration system and end-to-end managed services with WEKA's NeuralMesh, a software-defined storage product that continuously feeds GPUs with data. The goal is to cut idle GPU time and remove storage bottlenecks that slow AI workloads.
Singtel will also incorporate NeuralMesh as the high-performance data foundation for its GPU-as-a-Service, offered through a Centre of Excellence for Applied AI with a leading AI technology partner.
"Modern AI is no longer limited by compute power alone. The real bottleneck is how quickly data can reach GPUs, making high-speed data architecture just as critical as the GPUs themselves," says Bill Chang, chief executive of Singtel Digital InfraCo.
WEKA's NeuralMesh includes a feature called Augmented Memory Grid, which the company claims extends GPU memory capacity by 1,000 times, cuts the time taken to generate a first output by up to 20 times, and delivers 10 times more concurrent users from the same GPU footprint. WEKA counts 30% of the Fortune 50 among its customers.
Under the deal, the two companies will offer a Sovereign AI Factory as a managed service targeting governments and enterprises across Singapore and Asean.
See also: Jeff Bezos’ AI start-up Prometheus valued at US$41 bil in funding round
Use cases span national digital services, financial risk management, healthcare diagnostics, public safety, and smart transport and utilities operations.
Customers can choose from three configurations depending on their regulatory and operational needs: a single-tenant Dedicated Sovereign Pod for highly sensitive workloads, a Regulated Multi-Tenant Sovereign Zone providing isolated, shared environments for regulated industries, and a Hybrid Sovereign Model that links on-premises systems with sovereign cloud capacity.
Demand for AI environments that meet data residency requirements, regulatory alignment and operational autonomy has been rising across the region. Agentic and reasoning-based AI models are intensifying that pressure, requiring microsecond latency, faster and more reliable access to large datasets, and consistent throughput that conventional storage architectures struggle to deliver.
See also: Trust will decide whether Singapore’s AI bet pays off
"Sovereign AI in the Asean region is not simply about where data is stored. It's about developing future-ready infrastructure that can support exascale AI innovation across the entire ecosystem," says Liran Zvibel, co-founder and chief executive of WEKA.
Since Singtel operates as both a connectivity and cloud provider, AI models built inside sovereign data centres can be extended to edge locations for real-time, ultra-low-latency applications including traffic and video analytics, airport and port operations, and industrial automation. The platform also supports agentic AI systems designed for complex, multi-step workflows across large data volumes.
