TLDR: Extropic has introduced a new class of ‘thermodynamic’ or ‘probability chips’ designed to drastically reduce the energy consumption of artificial intelligence. The company claims its innovative hardware could be thousands of times more energy-efficient than current GPUs, offering a potential solution to the escalating energy demands of the AI industry.
Boston-area startup Extropic has unveiled a groundbreaking approach to artificial intelligence hardware, introducing ‘thermodynamic computing’ chips that promise to tackle the AI industry’s burgeoning energy crisis. The company’s first working chips, known as Thermodynamic Sampling Units (TSUs), are designed to be significantly more energy-efficient than conventional GPUs, with claims of up to 10,000 times less energy consumption for certain generative AI tasks.
The rapid expansion of AI has led to an ‘AI Energy Wall,’ where the energy required to train and operate large models is becoming a fundamental barrier to future scaling. Industry analysis projects that U.S. data centers are on track to consume a staggering 426 terawatt-hours by 2030, more than doubling their 2024 usage of 183 TWh. This insatiable demand has fueled a multi-billion-dollar infrastructure arms race, with tech giants like Amazon, Apple, OpenAI, and Meta investing heavily in massive, power-hungry data centers.
Extropic’s innovation lies in a fundamental rethinking of chip architecture. Instead of relying on deterministic processing, their chips utilize ‘probabilistic sampling’ and ‘p-bits’ (probabilistic bits) that represent uncertainty. This approach harnesses the inherent electronic noise in silicon, a phenomenon typically combated in traditional computing, allowing the chips to ‘think’ in probabilities rather than binary absolutes. This paradigm shift enables the hardware to focus on outcomes, leading to radical energy savings.
Extropic officials emphasized the urgency of the problem, stating, “With today’s technology, serving advanced models to everyone all the time would consume vastly more energy than humanity can produce.” The company’s XTR-0 development platform, backed by a $14.1 million seed round led by Kindred Ventures, has already produced its first working hardware. This prototype is currently undergoing testing by a select group of early partners, including frontier AI labs, weather startups, and government representatives, signaling strong early engagement and potential for real-world applications.
Also Read:
- AWS Unveils Trainium3 at re:Invent 2025, Promising a New Era for AI Scaling
- Meta Leverages Advanced AI to Streamline Operations and Drastically Reduce Carbon Emissions
If successfully scaled, Extropic’s thermodynamic chips could drastically cut the operational costs and energy footprint of AI data centers, paving the way for more sustainable and widespread AI adoption across various sectors, from scientific research to advanced weather forecasting. This breakthrough challenges the current GPU-centric AI infrastructure and offers a credible path toward a more energy-efficient future for artificial intelligence.
