Skip to main content

Extropic

Category: AI Chips / Semiconductors

A pioneering thermodynamic computing platform that harnesses natural thermal fluctuations to run probabilistic AI models with extreme energy efficiency. Extropic was founded in 2022. The company is led by Guillaume Verdon. Based in Austin, USA. Team size: 11-50. Total funding raised: $14.1M. Latest round: Seed, $14.1M, 2023-12, led by Kindred Ventures. Key investors include Kindred Ventures, Buckley Ventures, HOF Capital, OSS Capital, Valor Equity Partners, Marc Andreessen, Chris Dixon.

Founded
2022
Headquarters
Austin, USA
Team size
11-50
Total funding
$14.1M

Value proposition

Achieves 10,000x to 100,000x greater energy efficiency for generative AI inference compared to traditional GPUs by utilizing physics-based randomness rather than suppressing it.

Products and solutions

XTR-0 (First-generation silicon-based thermodynamic processor), Thermodynamic Sampling Unit (TSU) - Core hardware primitive for probabilistic sampling, thrml (Thermodynamic Hypergraphical Model Library) - Open-source Python/JAX framework, Denoising Thermodynamic Model (DTM) - Optimized architecture for diffusion-based generative AI

Unique value

Replaces deterministic digital bits with 'probabilistic bits' that treat electronic noise as a computational resource rather than a limitation, allowing for passive operation where energy is primarily spent on measurement.

Target customer

Frontier AI research labs, weather modeling startups, sovereign governments, financial institutions (Monte Carlo simulations), and private cloud providers.

Industries served

Artificial Intelligence (Generative AI & LLMs), High-Performance Computing (HPC), Scientific Research (Meteorology & Physics), Financial Services (Quantitative Analysis), Semiconductors

Technology advantage

Enables a new compute primitive called 'sampling' that sits alongside traditional matrix multiplication (matmul), drastically reducing the power-law costs of scaling AI by aligning AI's probabilistic nature with the underlying laws of physics.

How they differentiate

Extropic utilizes thermodynamic fluctuations in electronic circuits (entropy) as a computational resource for probabilistic sampling, whereas traditional competitors attempt to suppress noise. It claims a 10,000x to 100,000x energy efficiency advantage over GPUs for generative AI inference by moving from deterministic digital bits to probabilistic 'P-bits'.

Main competitors

Normal Computing, Rain AI, Inpro

Key partnerships

Frontier AI research laboratories (Early access partners), Government-linked research institutions, Weather modeling agencies, Venture partners (Kindred Ventures, Buckley Ventures, OSS Capital)

Notable customers

Frontier AI research labs (Early access partners), Weather modeling startups, Government-linked research institutions

Major milestones

Founded in 2022 by former Alphabet/Google X researchers Guillaume Verdon and Trevor McCourt, Closed $14.1M Seed round in late 2023 (announced March 2024), Publicly revealed the XTR-0 development platform, the first silicon-based thermodynamic processor, in 2024, Released 'thrml', an open-source Python/JAX library for thermodynamic hypergraphical modeling, Founder Guillaume Verdon revealed as 'Beff Jezos', the leader of the Effective Accelerationism (e/acc) movement

Growth metrics

Transitioned from stealth to public debut in March 2024; team size estimated between 11-50 employees.

Market positioning

Frontier AI infrastructure hardware disruptor targeting the sampling bottleneck in Generative AI.

Geographic focus

North America (primarily USA and Canada)

Patents and IP

Extensive patent portfolio covering thermodynamic sampling architectures and physics-based AI acceleration; includes prior IP by CEO Guillaume Verdon in Quantum Machine Learning and Signal Processing from his tenure at Google X.

About Guillaume Verdon

Guillaume Verdon is a mathematical physicist and entrepreneur who previously served as a Research Scientist and Quantum Machine Learning Lead at Alphabet’s X (formerly Google X) and Google AI. He is the co-inventor of TensorFlow Quantum and is widely recognized as the founder of the Effective Accelerationism (e/acc) movement under the pseudonym 'Beff Jezos.' He left Alphabet to found Extropic to pioneer thermodynamic computing.

Official website: