Extropic AI Unveils Silicon P-Bit Results, Targeting 1,000x AI Energy Efficiency

Paris, France – Extropic AI, a startup pioneering a novel physics-based computing paradigm, recently presented its silicon probabilistic bit (p-bit) results in Paris, with the findings also submitted for peer review. The announcement, made by Extropic AI co-founder Guillaume Verdon, known as "Beff – e/acc" on social media, highlights the company's significant progress in developing hardware designed to revolutionize artificial intelligence computation. According to Verdon's tweet, "> That's right. Here's a preview of our silicon pbit results. Huge effort by our team @Extropic_AI. These were recently presented in Paris and submitted to peer review."

Extropic AI, co-founded by former Alphabet Inc. quantum computing researchers Guillaume Verdon and Trevor McCourt, is developing chips that harness thermodynamic fluctuations within electronic circuits. Unlike traditional digital bits that are fixed as 0s or 1s, Extropic's p-bits fluctuate between states based on controllable probabilities, leveraging what engineers typically consider noise as a computational asset. This approach aims to deliver substantial energy efficiency improvements, potentially up to three to four orders of magnitude more efficient than current hardware.

The company's innovation centers on controlling thermodynamic effects in conventional silicon, a departure from earlier thermodynamic computing efforts that relied on superconducting electronic circuits requiring extreme cooling. This use of standard silicon is expected to make Extropic's technology more scalable and mass-manufacturable. The demonstrated functionality includes controllable bit-state transitions, laying the groundwork for a new class of energy-efficient computing platforms.

Extropic's probabilistic hardware is particularly well-suited for Monte Carlo simulations, a computational method critical in fields such as finance, biology, and advanced AI. The company believes its chip architecture is ideal for powering sophisticated AI models like OpenAI's o3 and Google's Gemini, which heavily rely on probability and inference. This positions Extropic as a potential challenger to established players like Nvidia in the high-performance computing and AI hardware market.

The startup secured $14.1 million in seed funding in December 2023, led by Kindred Ventures, with participation from several other venture capital firms. This capital injection is intended to accelerate the development of Extropic's computational paradigm and advance the field of generative AI. The company aims to address the escalating energy demands of AI, offering a more sustainable path for future computational growth.