Extropic.ai logo

Extropic.ai

Paid

Thermodynamic computing for ultra-efficient AI

4.6
Type
Saas
Founded
2022
Company
Extropic

About Extropic.ai

Extropic is building thermodynamic computing hardware (thermodynamic sampling units, or TSUs) that is inherently probabilistic and radically more energy-efficient than GPUs. Their platform includes the XTR-0 prototype, which enables low-latency communication between Extropic chips and traditional processors, and an open-source Python library called THRML that allows developers to create and simulate thermodynamic algorithms. Founded in 2022 with a team from Google, IBM, Apple, and Microsoft, Extropic targets probabilistic AI workloads and aims to disrupt data center computing.

Key Features

Thermodynamic sampling units (TSUs) inherently probabilistic
XTR-0 prototype platform for low-latency chip-to-processor communication
Open-source THRML Python library for thermodynamic algorithm development and simulation
Radically more energy efficient than traditional GPUs
Denoising Thermodynamic Models research paper available

Pros & Cons

Pros
  • Inherently probabilistic hardware perfectly suited for probabilistic AI
  • Open-source software (THRML) lowers barrier to entry for developers
  • Strong team with experience from Google, IBM, Apple, and Microsoft
  • Backed by notable investors and advisors from the AI and tech industry
Cons
  • Hardware still in prototype stage (XTR-0, X0 Z1) – not yet commercially available
  • Requires adoption of a new computing paradigm (thermodynamic/probabilistic)
  • Limited software ecosystem beyond the THRML library
  • Only suitable for probabilistic workloads, not general-purpose computing

Best For

Probabilistic AI workloads (e.g., sampling, inference)Ultra-efficient AI algorithm developmentEnergy-efficient data center computingResearch in thermodynamic computing and probabilistic algorithms

Alternatives to Extropic.ai

FAQ

What is thermodynamic computing?
Thermodynamic computing is a new computing paradigm that uses thermodynamic sampling units (TSUs) which are inherently probabilistic, aiming to be radically more energy efficient than traditional digital hardware like GPUs.
What is THRML?
THRML is Extropic's open-source Python library that enables developers to create thermodynamic algorithms and simulate running them on TSUs.
When was Extropic founded?
Extropic was founded in 2022.
Who are the investors in Extropic?
Investors and advisors include Amjad Masad (Replit), Aidan Gomez (Cohere), Balaji Srinivasan, Naval Ravikant, Aravind Srinivas (Perplexity), Ari Emmanuel, Scott Belsky, Garry Tan (YC), and Logan Kilpatrick (Google DeepMind).