OpenAI just put its name on Firmus’s 900-megawatt push. Now we find out what that signature actually buys. Quick context: big AI labs and cloud platforms have been locking in compute years ahead instead of treating it like an operating expense. Even before Firmus, AWS, Alibaba, ByteDance, DeepSeek, OpenAI, and Anthropic all had buildout plans underway. Anthropic also reportedly has a $35 billion obligation with Lambda centered on Hut 8’s Beacon Point campus in Texas, with Nvidia supplying the GPUs and holding the lease. This is AI Daily Briefing. OpenAI finally gives one of these giant builds an anchor customer, and Indonesia has a huge target of its own. Stay with us—because the hardware dates are where these promises get awkward. From ET Datacenters:
Australian artificial intelligence (AI) infrastructure company Firmus Technologies has secured a multi-year agreement with OpenAI to provide dedicated AI compute capacity from two of its AI Factory sites in Malaysia. This marks Firmus’s expansion into a fourth Asia-Pacific market, following its existing footprint in Australia, Singapore, and Indonesia.
On the Big Tech AI lease front, OpenAI is now Firmus’s anchor customer for Malaysian AI factories. Two sites, a multi-year commitment, and Firmus says its contracted book is above 900 megawatts. Okay—this is the bar. OpenAI is an actual tenant, not a logo floating over a rendering, and Firmus has tied the build to Vera Rubin NVL72 hardware. Seven planned AI factories would put Malaysia alongside Firmus’s footprint in Australia, Singapore, and Indonesia. NVIDIA shows up across these regional compute deals, and OpenAI gets a direct path into the inference stack. But Vera Rubin is still a forward silicon bet. Great, OpenAI has reserved the seats; now NVIDIA has to deliver the plane on time. Here's TelecomDrive:
The collaboration is designed to meet the region’s rapidly growing demand for secure and scalable AI computing, while advancing Indonesia’s position as a regional AI hub. With a deployment target of 1 gigawatt (GW) of NVIDIA DSX AI Factory capacity, Zankore by Indosat lays the foundation for one of the region’s largest AI infrastructure platforms which will enable the next phase of agentic and enterprise AI adoption at scale.
Zankore’s headline is one gigawatt of NVIDIA DSX capacity. What matters more is the usable date: about 200 megawatts of GB300 NVL72 targeted for the first half of 2027. Put that on a deployment calendar. And Indonesia follows the Malaysia deal: Nokia and NVIDIA are inside Zankore, while Vera Rubin is planned for Firmus. Southeast Asia is where sovereign-compute plans are starting to meet actual silicon. I’ll give Indosat this: with Nokia, NVIDIA, Ooredoo, and a 200-megawatt H1 2027 target, it’s a more testable pitch than a giant capacity number floating in space. Now it comes down to shipping the GB300s on time, keeping the power steady, and finding work to fill them. Also watch who owns the inference layer after all this comes online. A sovereign facility running someone else’s hardware, networking, and operating stack can get complicated very quickly. If your team needs a briefing like this for your own industry, Lantern can make a private daily podcast covering your competitors, market, or beat, delivered through a private feed for your whole team. Try it free for 14 days at lantern podcasts dot com slash briefings.
Every story is linked in the show notes. Check out the ones that caught your attention when you have a minute. Thanks for listening. We’ll be back tomorrow. This is a Lantern Podcast.