A New Step Against NVIDIA Dominance: Startup Infinity Raises $15 Million

Infinity, a startup focused on AI infrastructure, announced it has raised $15 million in a new funding round. The company is valued at $100 million. The project has attracted investment from major funds like Touring Capital and Principal VC, as well as leading researchers from OpenAI and Anthropic, highlighting the project's technological promise. This is reported by Techcrunch.com reports .

The main goal of the Infinity startup is to create software that makes it easier for AI models to run on various chips. Today, NVIDIA's absolute market dominance is linked not only to its powerful chips but also to its CUDA (Compute Unified Device Architecture) software platform. CUDA allows developers to write applications in popular languages like Python and run them seamlessly on NVIDIA graphics processors (GPU).

A universal alternative to the CUDA system

Currently, major AI frameworks like PyTorch and TensorFlow are optimized specifically for NVIDIA chips. For small startups and researchers, adapting applications to other types of chips—writing low-level kernels—is a complex and expensive process. Infinity aims to solve this by creating a universal inference library capable of running on any chip (SRAM, GPU, or smartphone chips).

Company founder Jeremy Nixon was previously a researcher at Google Brain. According to him, the research agent developed by Infinity, called Ignition, has the ability to write, test, and debug low-level code without human intervention. This system self-optimizes, measuring how fast the hardware is running and automatically rewriting code to increase performance.

Efficiency and market impact

The solutions offered by Infinity significantly save human labor. For example, in one study conducted by the company, a coding process that could take months or even years was completed in a few hours or days using this agent. Currently, the startup is collaborating with chip manufacturers like D-Matrix and is in negotiations with other major cloud platforms.

The company's financial model is also unique: it does not require upfront license fees from customers. Instead, Infinity takes a percentage of the performance gains and cost savings achieved. This reduces financial risks for customers testing the technology. The company currently employs 26 highly skilled engineers and designers.

At a time when interest in AI technologies is growing in the Uzbekistan market, such universal solutions could allow local developers to reduce their dependence on NVIDIA chips and effectively utilize more affordable hardware. According to TechCrunch, these types of projects will serve to democratize AI infrastructure in the future.