Another infrastructure breakthrough is brewing in the artificial intelligence industry. Australian data center operator Firmus Technologies, in partnership with Nvidia, has announced the large-scale deployment of up to 170,000 AI accelerators at its Batam campus. This is one of the most ambitious projects in the Asia-Pacific region, which will fundamentally reshape the landscape of high-performance computing.

According to the agreements reached, equipment deliveries will begin in 2027–2028. This is not just about the current generation of chips, but an entire lineup of future platforms: from Grace Blackwell to Vera Rubin and Vera. Such a time horizon indicates that the parties are laying the groundwork for years to come, focusing on the most advanced architectures.

The financial side of the deal is equally impressive. Firmus expects to receive between $25 and $30 billion under take-or-pay agreements for computing capacity during the first six years of operation. This is a classic industry mechanism that minimizes risks for the operator and ensures stable demand for Nvidia's future generations of accelerators.

The Batam campus, being built with the participation of DayOne, is positioned as one of the largest AI infrastructure projects in the Asia-Pacific region. Given the current shortage of computing resources for training and inference of large language models, such capacity will become a critically important asset for the entire region.

My analysis: This deal is a clear marker of how the market is changing. We are seeing a shift from scattered data centers to giant clusters financed under the "take or pay" model. Nvidia is essentially hedging its future sales, while Firmus gains access to the most advanced chips without the risk of downtime. There is an important signal here for the crypto industry: if such capacity is used for mining or decentralized AI, we could see explosive growth in performance in these segments.