Artificial Intelligence

Extropic Unveils Thermodynamic 10,000x Energy Saving AI Chips

Published OCT 30 2025

Story

Image credit: Unsplash

The Facts

Extropic, a Boston-area startup founded by former Google quantum computing researchers, has developed thermodynamic sampling units that use probabilistic bits rather than traditional processing methods, claiming energy efficiency gains of up to 10,000 times those of current graphics processing units (GPUs).
Guillaume Verdon on X
WinBuzzer
Wired
The XTR-0 development platform consists of a central processing unit, an FPGA chip and two sockets for thermodynamic sampling units. Early testing partners include weather modeling companies and government representatives who are evaluating the probabilistic computing approach.
Wired
Extropic
SciTech Era on X
Extropic's thermodynamic sampling units harness electron fluctuations to model probabilities directly in silicon, using probabilistic circuits that sample from distributions rather than performing the traditional matrix-multiplication operations required by current AI systems.
WinBuzzer
Extropic
The Rundown AI
This comes as U.S. data centers consumed 183 terawatt-hours of electricity in 2024 — over 4% of the country's total electricity consumption — and this figure could reach 426 TWh by 2030, according to International Energy Agency estimates.
Pew Research Center
Digital Watch Observatory
IEA
Natural gas supplies over 40% of electricity for U.S. data centers as of 2024, while renewables provide about 24%, nuclear power around 20% and coal approximately 15%. Natural gas is projected to remain the largest energy source through 2030.
Scientific American
Pew Research Center
IEA
Data center expansion could lead to an 8% increase in average U.S. electricity bills by 2030. High-demand markets are potentially seeing increases exceeding 25%, as utilities pass infrastructure upgrade costs to consumers through rate adjustments.
Scientific American
Pew Research Center
Axios

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-optimist narrative

Thermodynamic computing is a revolutionary breakthrough that harnesses thermal noise as a computational resource instead of fighting it. This paradigm shift eliminates the massive energy waste of digital systems that artificially generate pseudo-randomness, delivering true efficiency gains for AI and optimization tasks.

Viggo Hansen on LinkedIn

E1 Ventures

Techno-skeptic narrative

Extropic's XTR-0 is recycled stochastic computing dressed in thermodynamic buzzwords — a tiny FPGA toy for niche RNG tasks. Lacking scalable benchmarks, it ignores analog pitfalls like chip variability and verification nightmares. Amid GPU dominance, this unproven prototype risks vaporware status, fooling investors with sci-fi flair.

Y Combinator

Cynical narrative

AI has hit an energy wall. The AI boom mirrors the unsustainable telecom bubble of the late 1990s, with massive investment in technology that can't deliver profits fast enough. Companies are already installing polluting generators because the grid lacks capacity for their data centers, threatening economic collapse when reality hits.

Paul Krugman on Substack