Nvidia has signed an agreement with Firmus Technologies, an Australian data-center developer it has invested in, giving the company priority access to its newest artificial-intelligence chips as Firmus expands cloud infrastructure across the Asia-Pacific region. The deal, reported by Reuters, ties one of the world’s most supply-constrained chip makers to a builder of remote, renewable-powered data centers in a market where demand for AI computing is growing faster than capacity.
Firmus has raised more than $500 million, with Nvidia among its investors, and is known for a network of facilities it calls SuperSites: large-scale data centers built in remote parts of Western Australia and powered by solar, wind and other renewable sources. The company has pitched itself as a solution to two problems at once, the shortage of AI compute and the enormous energy appetite of the machines that provide it. The new agreement deepens that relationship, according to people familiar with its terms, by formalizing Firmus’s place in the queue for Nvidia’s latest GPUs.
The practical effect is that Firmus can offer its customers access to hardware that most cloud providers cannot guarantee. Nvidia’s newest accelerators are sold out for quarters in advance, and the allocation process has become a central fact of the AI economy: who gets the chips, in what order, and on what terms. A formal agreement with Nvidia puts Firmus ahead of rivals that must wait for general availability, a significant advantage in a market where enterprises are willing to prepay for capacity and where a guaranteed delivery date is itself a product.
The timing reflects the state of the Asia-Pacific market. Demand for AI computing in the region has been rising quickly, driven by local model developers, enterprises and government programs, but most of the advanced capacity sits in the U.S. Cloud providers have been building in Singapore, Japan and Australia, yet power constraints and land costs have slowed the pace. Firmus’s model, building where energy is cheap and abundant rather than where populations are dense, is designed to sidestep those constraints, and the company has signed long-term power agreements with renewable developers to lock in supply.
Analysts said the deal is also a statement about Nvidia’s strategy. The company has been investing directly in data-center builders and cloud providers around the world, a pattern that gives it influence over where its chips are deployed and who can access them. In addition to Firmus, Nvidia has backed a series of infrastructure companies in Asia and the Middle East, part of an effort to create a network of partners that can serve regional demand without depending on the big U.S. clouds. Each investment expands the pool of buyers for its hardware while building a web of relationships that competitors would struggle to replicate.
The arrangement has drawn questions as well. Some industry executives have asked whether Nvidia’s investments in its own customers create conflicts, particularly when those customers compete with the hyperscalers that buy the most chips. Nvidia has said its investments are designed to expand the market for AI computing, not to favor one customer over another, and that allocation decisions are based on demand and readiness. Firmus has said its agreements are commercial in nature and that it remains free to work with other chip suppliers.
For Firmus, the deal is validation of a long bet. The company was founded on the argument that the AI boom would eventually collide with the physical limits of energy and land, and that the winners would be operators that planned for that collision. Its SuperSites were dismissed by some early investors as too remote and too expensive to build; the Nvidia agreement, and the demand it signals, has shifted that assessment, according to people close to the company. The company’s pitch to customers now starts with the Nvidia relationship and the certainty it provides.
The agreement also puts Firmus in position to serve a wider range of customers. Priority access to Nvidia’s latest hardware makes the company more attractive to global enterprises that have struggled to secure capacity, and to local governments that want AI infrastructure built onshore. Firmus has said it plans to expand beyond Australia, and the Nvidia agreement gives it a credential to take into new markets where sovereign data requirements are becoming a deciding factor in procurement. Several Southeast Asian governments have announced national AI plans in the past year, and each one needs someone to build the underlying infrastructure.
What remains to be seen is how much capacity Firmus can actually deliver. Building data centers at the scale required to satisfy AI demand takes years, and the company’s first major facilities are still coming online. The agreement with Nvidia guarantees access to chips, but the machines need buildings, power and cooling to be useful, and the remote locations that make energy cheap also make construction slower. For now, the partnership gives Firmus something almost as valuable as capacity itself: a place in the line, and the credibility that comes with it.
