Today we're announcing Phinity, an applied research lab for autonomous chip design and discovery.
We've raised $5.2 million in seed funding led by Uncork Capital, with participation from Moxxie and angels including Jeff Dean, following our pre-seed led by Pear through PearX. We're already working with some of the largest foundation model labs, have reached 8-figures in annualized run rate, and we're just getting started.
Our mission is to build fully autonomous chip design so AI can help us search beyond the limits of human engineering and discover computational systems we would never find on our own.
The limits of human search
Progress towards AGI is constrained by physical computation. The more efficiently we implement intelligence, the more capability we can extract from the same compute, energy, and silicon. There are still enormous gains to find across the full stack, from models and algorithms down to computer architecture and silicon itself.
But in hardware, the barrier for experimentation is exceptionally high. Silicon development is a one-shot bet that takes 9 to 24 months, with separate teams carrying out design, verification and physical implementation. By the time that chip reaches customers, the workloads it was optimized for may already have changed. With hundreds of millions of dollars and months of delay at stake if they get it wrong, teams must be cautious, risk-averse and preserve flexibility.
The result is an economic limit on exploration. Chip design is an enormous optimization problem, but today we search the design space only through a sequential human process, limited by human time and resources.
Toward autonomous chip design
Our team is building the infrastructure for a closed-loop system where agents can design, evaluate, and revise computational systems using real physical feedback, from architecture discovery through tapeout. What takes human teams months to learn can now become part of the design loop itself. The Phinity system produces verified designs an order of magnitude faster with better power, performance, and area than experienced hardware engineers working from the same specifications with frontier coding agents.
A new era of custom silicon
As Phinity's agents take on more of the work of chip design, they'll be able to find architectures no human team could have reached on their own.
We believe chip design will follow Jevons Paradox: as custom chips become faster and cheaper to build, demand for them will explode. Once a chip can be designed in months instead of years, we move from a world where a handful of companies design chips to one where every important product can have silicon built for exactly what it needs to do. AI will design the best chips to run itself, and specialized intelligence will be practical in places where power, cost, and size make it impossible today.
Imagine if we could build a new kind of computer for every problem humanity has never had enough compute to solve. A machine built specifically to simulate a living cell atom by atom to replace years of wet-lab trial and error with compute. A spacecraft whose onboard silicon is designed specifically to reason, experiment, and make discoveries for itself at the edge of the solar system, where Earth is hours away. A brain implant with circuitry built around neural signals, decoding thousands of neurons in real time without generating enough heat to damage the brain. An AI model running on hardware designed precisely for its own architecture, then using that advantage to design an even better AI model and the chip that comes after it.
Today, we build custom silicon only for problems valuable enough to justify years of work by hundreds of engineers. What happens when designing a new chip becomes almost as cheap and fast as writing software?
We're building towards a future where we can explore computational systems beyond the reach of human search. We'll be able to discover entirely new architectures and turn them into the silicon that powers the next generation of intelligence.
Come join us
We founded Phinity because we believe the way we design chips has to change if we want to keep pushing the frontier. We came to chip design from post-training. Sonya built synthetic data pipelines for NVIDIA's chip design and code reasoning LLMs and did graph ML research at the AWS AI Lab; Aadi is a previously-exited founder who's worked on synthetic data generation and post-training.
Along the way, we've brought together an exceptional group of engineers and researchers, including a former leader of Intel's custom silicon group, a former ASIC Lead at Kepler Computing, former NVIDIA architects who have led multiple tapeouts on advanced nodes, and AI researchers who have trained LLMs for chip design and formal verification. Collectively, they've shipped AI accelerators across the full spectrum, from power-efficient automotive NPUs to 3D-stacked datacenter silicon with HBM.
We're a lean, deeply technical team and always on the lookout for top talent. If you want to build the future of autonomous chip design, reach out to us at hello@phinity.ai.
– Sonya and Aadi