AI infrastructure startup Infinity secures $15 million in seed funding to develop Ignition, its autonomous AI agent that enables any AI chip to achieve production-ready inference performance in days instead of years.
San Francisco, USA – Infinity.inc, an emerging AI infrastructure research company focused on making AI chips inference-ready through advanced software automation, has announced the successful closure of a $15 million seed funding round at a $100 million post-money valuation.
The investment round attracted participation from Touring Capital, Principal Venture Partners, executives from leading semiconductor companies, AI researchers from OpenAI and Anthropic, and several prominent angel investors.
The newly raised capital will support the expansion of Infinity’s engineering team, accelerate the development of its AI-powered research platform, strengthen partnerships with semiconductor companies including d-Matrix, and further enhance its flagship autonomous AI agent, Ignition.
Infinity Aims to Solve One of AI’s Biggest Infrastructure Challenges
As artificial intelligence adoption continues to surge, demand for high-performance AI chips is growing rapidly. However, many advanced semiconductor companies struggle to deploy their hardware efficiently because building optimized inference software typically requires months—or even years—of engineering effort.
Infinity is addressing this problem by creating an automated software layer capable of generating optimized inference code for virtually any AI accelerator.
According to founder and CEO Jeremy Nixon, the future AI ecosystem will not be determined solely by hardware performance but by how effectively AI models can run across different chip architectures.
“The next generation of AI leadership will belong to those who can enable any chip to run state-of-the-art AI models at exceptional speed,” Nixon said.
Ignition: Infinity’s Autonomous AI Research Agent
At the heart of Infinity’s platform is Ignition, an autonomous AI research system designed to automatically write, test, optimize, and improve low-level compute kernels responsible for AI inference performance.
Instead of relying entirely on large engineering teams, Ignition continuously generates optimized software while engineers provide architectural guidance and strategic oversight.
Its capabilities include:
- Automatic generation of inference kernels for new AI hardware
- Continuous optimization through recursive self-improvement
- Performance tuning based on real-world execution feedback
- Support for multiple hardware architectures and proprietary instruction sets
- Faster deployment of production-ready inference libraries
This dramatically shortens the software development lifecycle traditionally required for new AI chips.
Proven Performance Improvements
Infinity has already demonstrated significant performance gains using Ignition across multiple AI workloads.
Among its reported achievements:
- Improved inference throughput for the Qwen3-8B model by approximately 34% compared to the popular vLLM framework after just one day of automated optimization.
- Increased inference speed from roughly 1,400 tokens per second to over 20,000 tokens per second, representing nearly a 14x improvement during internal testing.
- In collaboration with d-Matrix, Ignition reached approximately 92% of the theoretical peak performance of the company’s Corsair AI chip within only 10 hours of first accessing the hardware.
- Successfully enabled three frontier AI models—Qwen3, Qwen3.5, and Gemma4—to run end-to-end on the new architecture within ten days.
These milestones illustrate how AI-generated software can significantly reduce deployment timelines for next-generation semiconductor platforms.
Reducing Dependence on Traditional AI Software Ecosystems
For years, NVIDIA’s CUDA ecosystem has dominated AI inference software, creating a substantial competitive advantage for NVIDIA hardware.
While competing chip manufacturers often develop powerful hardware, many struggle to build equally optimized software environments.
Infinity’s platform seeks to bridge this gap by automatically generating production-quality inference stacks, enabling semiconductor companies to unlock the full performance potential of their hardware without years of software development.
Industry analysts expect AI inference to account for nearly two-thirds of AI compute spending in 2026, making software optimization increasingly important across the semiconductor industry.
Strategic Partnership with d-Matrix
Infinity’s collaboration with d-Matrix demonstrates the commercial potential of its technology.
According to Sid Sheth, Founder and CEO of d-Matrix, AI-driven software generation could dramatically reduce the time required to commercialize new AI computing architectures while accelerating revenue generation for hardware manufacturers.
The partnership showcases how automated inference optimization can help emerging AI chip companies compete more effectively in the rapidly evolving AI hardware market.
Investors See Strong Market Opportunity
Investors believe Infinity is addressing one of the most significant bottlenecks in AI infrastructure.
Songyee Yoon, Founder and Managing Partner at Principal Venture Partners, emphasized that making advanced AI more accessible requires infrastructure capable of supporting a broader range of hardware platforms.
Similarly, Samir Kumar, General Partner at Touring Capital, noted that AI hardware companies constantly struggle to keep their software aligned with rapidly evolving neural network architectures.
He believes Infinity’s autonomous approach to inference optimization has the potential to transform how semiconductor companies maximize real-world chip performance.
Founded by Former Google Brain Researcher Jeremy Nixon
Infinity was founded by Jeremy Nixon, a former Google Brain researcher and co-founder of AGI House, a well-known San Francisco AI research and developer community responsible for launching hundreds of AI startups and projects.
Nixon has become a recognized voice in artificial intelligence, with his work and commentary featured by leading publications including The New York Times and Forbes.
Founded in August 2025, the company is headquartered in San Francisco and continues to expand its AI infrastructure research efforts.
Infinity Plans to Expand Engineering Team
Following its successful seed funding round, Infinity is actively recruiting engineers specializing in AI infrastructure, hardware enablement, inference optimization, and recursive self-improving systems.
The company aims to further accelerate development of Ignition while expanding partnerships with additional semiconductor manufacturers beyond its existing collaboration with d-Matrix.
About Infinity.inc
Infinity.inc (Infinity Artificial Intelligence Institute) is an AI infrastructure research company focused on building software that enables AI chips to efficiently execute modern inference workloads.
Its flagship autonomous AI agent, Ignition, automatically generates, tests, and optimizes inference software for new semiconductor architectures, reducing development timelines from years to just days.
Backed by leading venture capital firms, semiconductor executives, and AI researchers, Infinity is working to simplify AI hardware deployment and help unlock the full potential of next-generation AI accelerators.