Extropic AI: Building the next era of computing

This AI startup is building a chip that will push AI compute past Moore’s Law

They invented a way to do AI much more power efficiently and faster than current hardware

They recently raised a $14.1M seed round led by Kindred Capital

So what’s the startup and how does it work?

Here’s the story of Extropic AI:

Extropic AI was founded by Guillaume Verdon and Trever McCourt in August 2022 but only came out of stealth mode in March 2024.

Both founders come from a background in quantum computing. Guillaume did a Masters and PHD in Math and Quantum Information at the University of Waterloo. He then worked as an ML research scientist at Alphabet (Google), for 3 years before starting Extropic.

Trevor did a PHD in electrical engineering at MIT. He then interned at Google as a quantum AI researcher and worked as a graduate research assistant at MIT.

Guillaume tried to discredit the idea for Extropic for 6 years, but couldn't convince himself that it wouldn’t work...

So what’s Extropic?

The mission of Extropic is to build the ultimate substrate for AI compute - increase energy efficiency and speed for AI darastically.

The founders figured out they could do this by building a processor that uses noise from the heat and jitteriness of its electrons as an asset rather than a liability.

Since the electrons jiggle so fast, generative AI is done so much faster and more energy efficient because you use as few electrons as possible.

They call this ‘Thermodynamic computing’.

The platform Extropic is currently working on uses as few electrons as possible for power efficiency. However, later platforms will use more electrons which are going to be even hotter.

According to the founders, the chip will accelerate AI compute way past Moore’s Law - the number of transistors on a microchip doubles every 2 years, while its cost and size halves.

Today's algorithms spend around 25% of their time moving numbers around in memory, which limits the speedup achievable by accelerating specific operations.

Extropic's chips accelerate probabilistic algorithms by running them physically, unlocking a new era of AI acceleration well beyond what was previously thought achievable.

According to the litepaper: “Extropic’s superconducting chips are entirely passive, meaning we only expend energy when measuring or manipulating its state. This likely makes these neurons the most energy-efficient in the universe. These systems will be highly energy efficient at scale: Extropic targets low-volume, high-value customers like governments, banks, and private clouds with these systems.”

There’s an infinite hunger for AI compute - the biggest companies in the world are fighting to build the best AI models the fastest, so this could be a gold mine. 

Extropic has an impressive team with experts in Physics and AI, with previous experience at companies like Google AI, AWS, Meta and Nvidia.

In December 2023, Extropic raised $14.1M in seed funding led by Kindred Capital. Notable angel investors that also participated include Aaravind Srinivas (Perplexity), Garry Tan (YC), Naval Ravikant, Tobias Lutke (Shopify).

The litepaper can be found on Extropic's website at https://www.extropic.ai/future