Indian startup Ziroh Labs, together with IIT Madras researchers, has developed an affordable system that can execute big artificial intelligence (AI) models without costly computing chips made by firms such as Nvidia Corp. The new framework, called Kompact AI, was created in collaboration with the Indian Institute of Technology, Madras, based on a Bloomberg report.

Ziroh Labs says that its platform enables AI to run on the central processing units (CPUs) in everyday computing devices, instead of the highly coveted and expensive graphics processing units (GPUs) that have been at the core of the AI revolution. The startup says it can optimize top AI models to run on personal computers. In a demonstration event earlier this week, the research team demonstrated their product operating on an ordinary laptop featuring an Intel Xeon processor and successfully querying models like Meta's Llama 2 and Alibaba's Qwen 2.5.

Whereas other businesses have used CPUs for some inference workloads, Ziroh Labs points out that its approach produces high-quality results. The technology has already been tested, according to reports, by US chip makers Intel and Advanced Micro Devices.

Why is CPU usage in AI models significant? An increasing number of AI developers have pointed to improvements in efficiency that allow them to use fewer chips, especially after the viral success of China's DeepSeek, which reportedly created a competitive AI model at a fraction of what similar models cost from US developers. Ziroh Labs' method is focused primarily on the inference process, which is running AI systems once trained.

"It's going to have a very profound market impact in the years to come," said William Raduchel, former chief strategy officer for Sun Microsystems, who is a tech advisor to the startup and addressed the conference remotely.

Similar to developers elsewhere, Indian developers have struggled with the cost of and access to high-end Nvidia chips required to create and sustain AI products. The shortage of GPU threatens the speed and magnitude of local AI research and deployment. "The AI divide is real because only those with high-end, costly GPU-based resources can access, create, and deploy powerful AI," said V. Kamakoti, director at the Indian Institute of Technology, Madras. "We're proving that you don't require a revolver to kill a mosquito," he added.