Indosat’s AI fiber play needs Zankore’s anchor

Indonesia’s AI infrastructure build has split into two bets: a data center venture announced last month with up to 1GW of planned AI capacity, and the fiber spine meant to carry traffic between those data centers. Infra Fiber Teknologi (IFT), the Indosat Ooredoo Hutchison spin-off, unveiled that spine this week as RAIA Grid, an open-access platform that embeds machine learning, AI, and agentic AI into how its 86,000km of fiber is planned, deployed, and maintained.

IFT was created last month when Indosat spun off its fiber business with Indonesian investment firm Arsari Group, and the network it inherited already covers the archipelago: more than 86,000km of backbone, domestic subsea, and access fiber. Run by an independently governed, open-access vehicle rather than a telco cost center, RAIA Grid is the project that gives that independence a product.

The company calls RAIA Grid “an AI-powered infrastructure platform by design,” where machine-learning models help decide where the next route goes, how much to spend on it, and when demand is about to outstrip capacity. The same layer automates deployment, runs predictive maintenance, and moves the network toward more autonomous operations, all built for the traffic that follows compute: 5G, fiber-to-the-home, and data center-to-data center connections.

The strategic point is open access: RAIA Grid will sell connectivity to telcos, data centers, cloud providers, hyperscalers, and enterprises, rather than feeding one carrier’s consumer book. “RAIA Grid is built with a clear strategic ambition: to become the digital superhighway connecting Indonesia’s increasingly complex digital ecosystem, from data center to data center, and ultimately connecting Indonesia to the world,” said Hendri Mulya Syam, president director of RAIA Grid.

Indosat CEO Vikram Sinha framed RAIA Grid as more than connectivity, tying it directly to Zankore by Indosat, the data center venture the carrier announced last month with the Ooredoo Group, Nokia, and Nvidia and aiming for up to 1GW of AI data center capacity. “The opportunity ahead is bigger than connectivity. RAIA Grid is building the highway that connects data centers, cloud platforms, businesses and AI infrastructure,” Sinha said.

The pairing is what makes the fiber worth pricing like infrastructure. An open-access fiber build without a committed tenant is merchant risk: the operator pays for the duct and waits for tenants. Zankore’s 1GW capacity pipeline is the anchor that lets RAIA Grid borrow at infrastructure prices, and the AI layer is what may let it command a premium for less downtime and faster delivery.

The pattern is consistent with what this publication has argued across the AI buildout: the bottleneck has shifted from land and chips to electrons, and hyperscaler-anchored assets are the only digital infrastructure that clears infrastructure pricing. Zankore, with Nokia and Nvidia attached, is the anchor here; RAIA Grid is the route to that anchor.

The risks are in the timing. Indonesia’s geography makes fiber a natural monopoly, but it also leaves the network a patchwork of backbone, subsea, and access routes with different competitive dynamics. RAIA Grid’s machine-learning layer cannot make tenants appear before the data center capacity is live, and the open-access model invites rivals; other Indonesian infrastructure groups can build around it, and hyperscalers can decide to own their own fiber. And because approval queues, not capital, now decide where hyperscale computing gets built, Indonesia’s permitting environment will shape RAIA Grid’s route map as much as the AI models do.

Indosat has taken a cost center off its balance sheet, given it a long-term capital partner in Arsari Group, and wrapped the pitch to speak directly to the hyperscaler procurement cycle. The test comes when the anchors arrive: if Zankore’s 1GW materializes, IFT has the route to serve it and the software to manage congestion; if the data center build slips, RAIA Grid is a very smart 86,000km toll road waiting for traffic.