'10,000 times faster than a human scientist' New AI tool designed ultra-efficient heat-to-electricity generators at lightning speed, a breakthrough that could slash the cost of energy harvesters and help enable cheaper, high-performance home heat pumps

Date:
Tue, 28 Apr 2026 00:20:00 +0000

Description:
New AI-designed thermoelectric generators reach 9.3% efficiency and could
make waste heat easier to harvest.

FULL STORY ======================================================================Copy link Facebook X Whatsapp Reddit Pinterest Flipboard Threads Email Share this article 0 Join the conversation Follow us Add us as a preferred source on Google Newsletter Subscribe to our newsletter AI tool speeds thermoelectric generator design while matching leading prototype performance TEGNet cuts simulation time from thousands of seconds to fractions of one Cheaper waste-heat harvesters could follow, although manufacturing still has to prove itself Researchers in Japan have built an AI tool that can design thermoelectric generators far faster than standard simulation methods, pointing to cheaper ways of turning waste heat into electricity.

TEGNet was developed by Takao Mori and colleagues at Japans National
Institute for Materials Science (NIMS) and the University of Tsukuba. In the paper published in Nature , it predicted generator performance with more than 99% accuracy while using only 0.01% of the computing time needed by
commercial finite-element solvers. Article continues below You may like Could 'thermodynamic computing' unlock the true possibilities of AI? These studies think so, get ready for better image generation and much more The AI heat trap: why data centers must rethink thermodynamics Microsoft and Nvidia team up to use AI to reduce nuclear project bottlenecks Acting as a fast emulator Thermoelectric generators turn heat differences directly into electricity, with no turbines or moving parts.

They already power spacecraft, remote sensors and some isolated infrastructure, but cost and middling performance have kept them out of wider use in factories, refineries, vehicles and electronics.

Designing thermoelectric generators is slow because researchers have to balance materials, geometry, temperature conditions, electrical resistance
and heat flow.

A conventional solver has to solve coupled physics equations again and again, which can take days or weeks for broad searches. Are you a pro? Subscribe to our newsletter Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed! Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors By submitting your information you agree to the Terms & Conditions and Privacy Policy and are aged 16 or over.

TEGNet learns from those simulations, then acts as a fast emulator. The paper says a typical material simulation took about 2,237 seconds in COMSOL, while TEGNet produced the same type of result in about 0.25 seconds.

The researchers used the AI to improve two types of generators, one built
from stacked layers of different materials and another made from paired semiconductor materials that work together to produce electricity.

Lab-built prototypes reached conversion efficiencies of 9.3% and 8.7%, respectively, putting them among strong reported results for that temperature range. What to read next Re-engineered balsa wood can store heat and produce electricity in the dark Antimatter plans global AI network with 1,000 micro data centers by 2030 The post-transformer era has an answer to AIs energy crisis

That still doesnt make thermoelectrics a cure-all. Heat-to-electricity conversion is capped by basic thermodynamics, and these devices need enough temperature difference to be useful.

The interesting part is cost. Mori told IEEE Spectrum that estimated costs suggest an industrially competitive power-generation cost could be possible for the first time in thermoelectric history.

TEGNet also identified designs that could use simpler fabrication and, in
some cases, avoid bismuth telluride, a common but expensive thermoelectric material.

That could help waste-heat harvesters and high-performance home heat pumps become cheaper, although real-world manufacturing still has to prove the numbers. Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.



======================================================================
Link to news story: https://www.techradar.com/pro/10-000-times-faster-than-a-human-scientist-new-a i-tool-designed-ultra-efficient-heat-to-electricity-generators-at-lightning-sp eed-a-breakthrough-that-could-slash-the-cost-of-energy-harvesters-and-help-ena ble-cheaper-high-performance-home-heat-pumps


--- Mystic BBS v1.12 A49 (Linux/64)
* Origin: tqwNet Technology News (1337:1/100)