Share with your CTO
Indosat Ooredoo Hutchison is betting that Southeast Asia’s AI buildout needs sovereign, regional infrastructure, not just cloud access from hyperscalers, and it’s structured a new venture called Zankore to capture that opportunity. Ooredoo Group leads as capital provider, Nokia handles networking, and NVIDIA supplies the compute stack, specifically GB300 NVL72 systems, targeting 200 megawatts of capacity by the first half of 2027 and up to 1 gigawatt over time. Former Ericsson CTO Ulf Ewaldsson will run the venture.
What this means for your business
If your enterprise has AI workloads that can’t comfortably land on US-headquartered hyperscaler infrastructure, whether for data residency, latency, or regulatory reasons, Zankore’s 2027 timeline is the relevant pressure point. The GB300 NVL72 is NVIDIA’s highest-density current-generation system, so this isn’t a low-end regional play. CTOs planning GPU procurement or co-location strategy in Southeast Asia now have a credible third option forming alongside AWS and Azure, and the question is whether Zankore’s integrated stack justifies waiting 18 months rather than committing to existing capacity now.
The structural move here is worth watching carefully. Indosat just carved out its fibre assets into a separate entity for roughly $650 million, then immediately channeled proceeds toward 5G and AI infrastructure. That’s capital recycling at speed, trading lower-margin passive assets for a position in higher-margin AI compute. Telecom operators running this playbook, monetize the pipes, buy the brains, are essentially arguing that the connectivity layer is commoditizing and the differentiation shifts to full-stack AI delivery. Zankore’s architecture, compute plus networking plus power management integrated into NVIDIA’s DSX AI Factory design, is the physical expression of that thesis.
NVIDIA’s presence in this deal does more than supply hardware. It signals that NVIDIA is actively shaping regional AI infrastructure consortia, lending its ecosystem credibility to anchor deals that might otherwise struggle to attract enterprise customers. CTOs who assumed their NVIDIA relationship runs through a hyperscaler or a direct reseller should recognize that NVIDIA is now effectively a co-founder of competing regional infrastructure plays. If Zankore lands even two or three large Indonesian or APAC government contracts before 2027, it reframes whether your current cloud-first AI architecture remains the path of least resistance or simply the path of most familiarity.
Concept deep-dive: AI Factory
NVIDIA’s “AI Factory” describes a data center architecture where compute, networking, cooling, and power management are designed and optimized together from the start, rather than assembled from separate vendor choices. The analogy is a purpose-built manufacturing line versus a workshop with mismatched tools. For enterprise buyers, it matters because integrated design typically means higher utilization rates and simpler scaling, but it also means deeper lock-in to NVIDIA’s stack at the infrastructure layer, well below the application software most CTOs think of as the lock-in risk.
Based on reporting from Indosat launches AI infrastructure venture with Ooredoo, NVIDIA and Nokia, originally published 2026-08-06 10:12:00.
