Australia's Firmus Technologies has struck an AI access deal with Nvidia that will reshape how businesses access artificial intelligence infrastructure, potentially determining which companies can afford to compete in the rapidly consolidating AI economy.
Under the arrangement, Firmus will purchase Nvidia infrastructure and resell Nvidia-powered cloud services to 'AI Native' customers and other businesses. The deal creates a revenue-sharing model where Nvidia collects both upfront product sales and a portion of ongoing cloud revenue tied to those sales.
Who Controls AI Access
The partnership illustrates how access to cutting-edge AI capabilities is increasingly mediated through a small number of hardware providers and their chosen partners. Firmus will serve as a gateway between Nvidia's dominant AI chips and businesses seeking to build AI-powered services, a structure that concentrates market power at multiple levels of the technology stack.
Companies wanting to develop AI applications will now depend on relationships with intermediaries like Firmus, rather than building their own infrastructure or accessing AI resources through more distributed channels. This model favors established players with capital to purchase Nvidia's expensive hardware upfront.
The Economics of AI Infrastructure
Nvidia's dual revenue approach—collecting payment when hardware sells and again when that hardware generates cloud revenue—represents a significant shift in how technology infrastructure gets monetized. The chip maker isn't just selling products; it's claiming a stake in the ongoing commercial activity those products enable.
For Firmus, the deal provides access to Nvidia's market-leading AI accelerators, which have become essential for training and running large language models and other computationally intensive AI systems. But the arrangement also locks the Australian company into a revenue-sharing structure that gives Nvidia ongoing financial claims on its business operations.
Market Concentration Concerns
The partnership reflects broader consolidation in the AI industry, where a handful of companies control the specialized chips required for advanced AI work. Nvidia holds an estimated 80-95% share of the AI accelerator market, giving it substantial leverage in negotiating terms with cloud providers and resellers.
Smaller businesses and researchers seeking AI access will face costs shaped by this multi-layered intermediation, potentially limiting who can participate in AI development. The structure favors deep-pocketed enterprises over startups, universities, or public-interest projects with limited budgets.
Why This Matters:
The Firmus-Nvidia deal exemplifies how AI infrastructure is being organized around profit-sharing arrangements that concentrate both market power and revenue streams among a small number of corporate players. As AI becomes essential to everything from healthcare to education to scientific research, these access arrangements will determine which institutions can afford to use the technology and on what terms. The revenue-sharing model means costs get passed down to end users, potentially pricing out smaller competitors, academic researchers, and public-sector organizations. Without regulatory frameworks ensuring broad, affordable access to AI infrastructure, the technology risks becoming a tool that reinforces existing economic advantages rather than democratizing capability. How governments respond to this consolidation—whether through antitrust enforcement, public investment in alternative infrastructure, or requirements for open access—will shape whether AI serves broad social benefit or primarily enriches a narrow group of technology giants and their partners.