# Infinity

> Infinity builds inference for new silicon. Our research agent Ignition generates a full inference stack — kernels, compiler backend, and serving layer — on any AI chip in days, not years, making new accelerators inference-ready without years of hand-tuned software work.

All pages below are served as prerendered static HTML, so no JavaScript is required to read them.

## Research

- [LLM Inference on New Silicon in Days, Not Years: d-Matrix meets Infinity](https://infinity.inc/research/dmatrix-corsair): Ignition generates a full inference stack on the d-Matrix Corsair, running Qwen3, Qwen3.5, and Gemma4 end-to-end in 10 days at up to 92% Speed-of-Light.
- [Performant Full-Model Inference on d-Matrix Corsair: 20x Speedup](https://infinity.inc/research/dmatrix-full-model-inference): Infinity and d-Matrix optimized Qwen3 by 20x tokens/second on a single Corsair card, with 16x longer context.
- [Surpassing vLLM With a Generated Inference Stack](https://infinity.inc/research/qwen3-optimization): Infinity's infy optimization system wrote an inference engine from scratch and autonomously optimized it on Qwen3-8B, delivering up to 34.3% more tokens per second than vLLM with identical parameters.
- [Infinity × d-Matrix Partnership](https://infinity.inc/research/dmatrix-partnership): Partnering to enable and optimize performant full-model inference on d-Matrix Corsair.
- [Research index](https://infinity.inc/research): All research posts and technical deep dives.

## Company

- [Home](https://infinity.inc/): What Infinity does — inference for new silicon.
- [Hardware partnerships](https://infinity.inc/partnerships/hardware): Working with Infinity to make your accelerator inference-ready.
- [Newsroom](https://infinity.inc/press): Announcements and press releases, including our $15M seed funding.
- [Team](https://infinity.inc/team): Who we are.
- [Careers](https://infinity.inc/careers): Open roles.
- [Contact](https://infinity.inc/contact): Get in touch.

## Press Releases

- [Infinity Hires Meta Superintelligence Labs Chip Veteran and the CUDA Educator Who's Taught Two Million Engineers](https://infinity.inc/press/new-hires): October 23, 2026. Aditya Om, a former senior software engineer at Meta Superintelligence Labs, and Elliot Arledge, whose CUDA and LLM coding courses have taught more than two million people, join Infinity’s engineering team to advance automated AI research and inference optimization.
- [Infinity and HP Deliver 2X Performance Gains on NVIDIA's GB300 Grace Blackwell Ultra Desktop Superchip](https://infinity.inc/press/hp-collaboration): October 20, 2026. Infinity and HP doubled concurrent AI workloads and overall token throughput on the tested HP ZGX Fury configuration powered by NVIDIA’s GB300 Grace Blackwell Ultra Desktop Superchip, running NVIDIA Nemotron 3 Super without changing the underlying hardware.