Company grows compute capacity by 164% and commits USD 195 million in infrastructure investment to meet surging demand for enterprise and sovereign AI workloads across the region.
SINGAPORE — Freyr AI, an NVIDIA Cloud Partner and AI infrastructure service provider delivering GPU-as-a-Service and Token-as-a-Service, today announced a major expansion of its colocation infrastructure across Southeast Asia, reinforcing its position as one of the region’s fastest-scaling neocloud operators.
The expansion grows Freyr AI’s colocation compute capacity by 164% across two Southeast Asia sites, driven by rising demand from enterprise and sovereign AI clients seeking reliable, high-density GPU infrastructure closer to their operations. The buildout includes the deployment of NVIDIA B300 and GB300 servers, built to NVIDIA reference design specifications and supported by NVIDIA SpectrumX networking, high-density fiber connectivity, and purpose-built facility fitout — representing an initial infrastructure investment of USD 35 million.
Building on this momentum, Freyr AI has committed a further USD 160 million toward its next phase of expansion, targeted for delivery in the first quarter of 2027. Combined, the two phases represent USD 195 million in committed infrastructure investment across the company’s Southeast Asia footprint.
In addition, Freyr AI will continue to invest further and recently convened a Vera Rubin Workshop with NVIDIA and NEXTDC at NEXTDC’s facility in Kuala Lumpur, where the companies conducted an in-depth technical review of Vera Rubin deployment feasibility. Subject to the successful completion of this evaluation, Freyr AI expects to complete Vera Rubin deployment and go live in Q1 2027.
Freyr AI was the first NVIDIA Cloud Partner to receive Thailand Board of Investment approval, securing BOI status in January 2026 for a data hosting project in Samut Prakan. Freyr AI acknowledges that the Thailand BOI’s designation underscores the importance of GPU-as-a-Service infrastructure to Thailand’s digital economy agenda.
The expansion reflects a broader shift in the region’s AI infrastructure landscape, as enterprises and governments across Southeast Asia move to secure sovereign and localized compute capacity rather than relying solely on hyperscale providers based outside the region. Freyr AI’s model — built on colocation partnerships and strict adherence to NVIDIA reference architecture rather than ground-up construction — allows it to bring GPU capacity online faster than infrastructure-heavy megaprojects still years from completion. All this while ensuring that the performance and interoperability of the system are consistent with NVIDIA’s certified standards.
“We’re building capacity ahead of demand, not chasing it,” said Dapeng Liu, CEO of Freyr AI. “This expansion reflects capital that is already committed and infrastructure already underway, not projections on a roadmap. By following NVIDIA reference design across our deployments, our clients get validated, production-ready performance from day one, not a custom build they have to de-risk themselves. Being the first NVIDIA Cloud Partner to receive BOI approval in Thailand reflects the trust we’ve built with regulators as we scale AI infrastructure across the region.”
Freyr AI’s colocation strategy focuses on strategically located sites across Southeast Asia, giving clients low-latency access to GPU infrastructure without the multi-year timelines associated with sovereign-scale data center construction. The company continues to evaluate additional colocation opportunities across the region as part of its broader expansion strategy.
About Freyr AI
Freyr AI (Freyr Technology AI Pte. Ltd.) is a Singapore-headquartered NVIDIA Cloud Partner and global AI infrastructure platform that integrates high-performance infrastructure, a low-latency network, and an accessible compute marketplace, delivering GPU-as-a-Service and Token-as-a-Service. The company partners with leading colocation providers across the region to deliver high-density, NVIDIA-powered compute infrastructure built to NVIDIA reference design standards, with speed and reliability.