Here is a fun thought experiment: What if the company that already makes 90% of AI chips also controlled the software that decides how those chips get used?

That is not a hypothetical anymore. Nvidia quietly acquired SchedMD last December, giving it ownership of Slurm, the open-source workload scheduler that runs on roughly 60% of the world's supercomputers. Meta uses it. Anthropic uses it. Government labs that forecast hurricanes and simulate nuclear weapons use it. And now it belongs to the company whose hardware it is supposed to manage neutrally.

Reuters reported this week that AI specialists and supercomputer operators are raising alarms. Their concern is straightforward: a vendor that controls workload scheduling software has enormous leverage over how efficiently competing hardware performs. Nvidia promised to keep Slurm open-source and vendor-neutral. But promises and roadmaps are different things.

TechInsights analyst Manish Rawat told InfoWorld that Nvidia could shape the development roadmap to prioritize GPU-aware scheduling and topology optimizations that favor its own hardware. Integration timelines already show faster CUDA ecosystem support compared to AMD's ROCm or Intel's oneAPI. He called it a "best-supported path effect." That is a polite way of saying Nvidia's competitors will always be one step behind on the software their customers depend on.

The timing matters. AMD and Intel are both spending billions to challenge Nvidia's AI chip dominance. Chinese chipmakers just hit 41% domestic market share. Custom silicon from Google and Amazon is gaining ground. The one thing all of these alternatives need? Software that treats them fairly. Slurm was supposed to be that neutral ground.

Dr. Danish Faruqui, CEO of Fab Economics, said the risk is real: "Nvidia now controls Slurm's official development roadmap and code review process, which could influence how quickly competing chips are integrated." The open-source license means anyone can fork the code. But forking Slurm is like forking Linux: technically possible, practically a nightmare when 60% of the world's supercomputers run the original.

This is the playbook Big Tech has run for decades. Control the platform, control the ecosystem. Google did it with Android. Apple did it with the App Store. Nvidia is doing it with the invisible software layer that decides which chips get priority on the world's most powerful computers.

Watch this one closely. If Nvidia's competitors cannot prove Slurm is treating them fairly, the AI chip race stops being about who builds the best hardware. It becomes about who controls the software that picks the winner.