Eye for an eye? Think AI for AI: How one startup is using AI to make AI chips more energy-efficient

AI models need chips to generate results. These chips work around the clock, processing loads of information and producing outputs. As a result, they get too hot. In turn, these chips consume a lot of electricity to run the cooling systems, adding to the growing demand for power consumption with the growth of AI.

But a PhD scholar turned entrepreneur is giving AI problems an AI solution. Here's how:

‘Discovered Materials,’ founded by Akash Ramdas and Advaith Sridhar, is trying to find new materials to make chips that help consume less electricity and provide better heat management.

The startup has the backing of $9 million in seed funding from Lightspeed India Partners, Y Combinator, and several other investors. The company page writes, "We are backed by a $9M seed round led by Lightspeed, with participation from Y Combinator and Peak XV, and angels like Paul Graham, Gokul Rajaram and Thariq Shihipar."

Discovered Materials founder Akash Ramdas has a doctorate in materials science from Stanford University. The other founder, Advaith Sridhar, has worked on AI agents at Persona AI and Luma Labs.

Ramdas and Sridhar are using Anthropic AI models to make chip materials. They use AI models trained in physics to simulate how materials would behave in the real world.

ALSO READ: Luis Ceze, a professor, turned his lab's project into a startup. Nvidia bought it, made him Vice President, and he kept teaching

Sridhar told TechCrunch that Ramdas was making approximately 20 guesses per day during his PhD research. He further added that they are now able to make thousands of guesses in a day. They were able to achieve that target by running these agents 24/7 on the cloud.

The founders have claimed that they have discovered materials matching the properties of existing materials that are used by key chipmakers, but they refused to share more details with TechCrunch.

Manufacturing a chip out of the material researched for reducing heat and power consumption might be challenging due to compromised electrical properties or several other factors, report TechCrunch.

The company also plans to patent new material discoveries, their use, and the process of making chips from a particular substance, then license them to chipmakers. The cofounders hope to have patent-worthy materials ready by next year.

ALSO READ: AI's hidden power problem: 'Weeteq' claims to cut the electricity wasted inside data centres, here's how


Sources:

Discovered Materials

TechCrunch

FAQs

1. What is Discovered Materials and what is its primary goal?
   Discovered Materials is a startup founded by Akash Ramdas and Advaith Sridhar, focused on finding new materials for chips that consume less electricity and provide better heat management.
2. Who are the founders of Discovered Materials and what are their backgrounds?
   The founders, Akash Ramdas and Advaith Sridhar, have strong academic and professional backgrounds; Ramdas has a PhD in materials science from Stanford University, while Sridhar has experience working with AI agents at Persona AI and Luma Labs.
3. What advancements has Discovered Materials made in using AI for chip material discovery?
   Discovered Materials has significantly increased its materials discovery output, making thousands of simulations per day using AI models trained in physics, as opposed to the 20 guesses a day made by Ramdas during his PhD.