Back to Blog

Can This New Chip End AI’s Energy Problem? Extropic’s Bid to Beat GPUs at Their Own Game

Image from Wired’s Article on Extropic’s Innovative Thermodynamic Chip

TL;DR FAQ: Can Extropic’s Thermodynamic Chips Solve AI’s Energy Crisis?

▼ Q: What is Extropic’s thermodynamic chip and why is it revolutionary?

A: Extropic’s thermodynamic chip uses thermal noise to process probabilistic AI workloads far more efficiently than traditional GPUs, potentially lowering the massive electricity costs of AI and speeding up complex models that rely on randomness and sampling.

▼ Q: Why does AI use so much electricity today?

A: AI models, especially large ones, require intense computations with billions of calculations, which strains data centers, increases power consumption, and spikes cooling needs—driving up operational costs and environmental impact.

▼ Q: What makes Extropic’s approach different from NVIDIA chips?

A: A: While NVIDIA chips specialize in digital math and graphics processing, Extropic’s chips harness physical randomness through thermodynamics. This could make them much better at probabilistic AI tasks and dramatically cut energy use for specific AI workloads.

▼ Q: Which types of AI models benefit from thermodynamic chips?

A: Generative AI (like language models and image generators), simulations, robotics, and any model heavy on sampling or uncertainty could run faster, cheaper, and with richer outputs if powered by thermodynamic chips instead of standard GPUs.

▼ Q: Why were some promising AI projects abandoned before?

A: Large AI models got abandoned or scaled back because their compute needs were so high that running them wasn’t sustainable, affordable, or scalable within existing hardware and data center infrastructure.

▼ Q: How big could Extropic become if their chips outperform GPUs?

A: If Extropic’s chips deliver as promised, the company could follow the path of NVIDIA, whose focus on AI fueled a $5 trillion valuation. Extropic could become a central technology player and reshape the future of AI hardware.

▼ Q: What other inventions does this compare to?

A: This breakthrough is similar to the transistor, steam engine, printing press, and internet—each dramatically lowered barriers and sparked new eras of technology due to improved efficiency, scalability, or connectivity.

Artificial intelligence is exploding, and with it, data centers are burning through more power and water than ever. AI models are becoming larger and more complex, while the costs of electricity and hardware keep rising. Some of the most promising AI ideas, huge language models and detailed simulations, have been pushed aside because they simply demand too many resources to run.

But a startup called Extropic may have just changed everything. Their new chip, inspired by the physics inside electrical circuits, might solve AI’s biggest problem: how to run powerful models without draining our grids or killing research budgets.

What Makes Thermodynamic Chips Different?

Most computer chips (including the ones made by giants like NVIDIA) rely on strict math, battling against natural randomness in electricity to keep our calculations precise. Extropic’s thermodynamic chips flip the script, they use that randomness on purpose. For certain kinds of AI, like those that need to make guesses or run through lots of scenarios, these chips can blast through the work hundreds or even thousands of times more efficiently than today’s best GPUs.

Imagine a chip that could run giant language models and creative AI more quickly and without sending your electric bill sky-high. That’s how big this breakthrough could be.

Why AI Needs This Now

AI’s growth means more energy, more water, and more money spent just to keep computers running and cool. Some bold models, like those promising smarter answers or mind-blowing images, were dropped because the costs were just too high.

  • Product teams have “walked away” from frontier models that simply cost too much to operate at scale.
  • Scientists have simplified creative AI, trading quality for speed and lower energy use.
  • Smaller labs and companies never even get to try cutting-edge ideas, because the cost of running the models is out of reach.

If chips like Extropic’s take off, a lot of lost potential comes roaring back.

Big Changes, Big Comparisons

Our team sees this moment like some of the biggest inventions ever:

  • Transistor: Kicked off the digital revolution and made modern electronics possible.
  • Steam Engine: Powered trains and factories, launching the Industrial Age.
  • Printing Press: Made sharing ideas fast and cheap for everyone, not just the rich.
  • Internet: Connected the world, opening up business and learning on a global scale.
  • AC Electricity: Brought reliable power to homes and factories, making modern life possible.

If thermodynamic chips collapse the cost of uncertainty and massive AI workloads, they could do for AI what these inventions did for their eras: turn wild ideas into everyday tools.

Could Extropic Be the Next NVIDIA?

Just look at NVIDIA, a company with a $5 trillion valuation because their chips power today’s AI revolution. If Extropic succeeds at making smarter and much cheaper chips for the next wave of AI models, their business could follow the same rocket path. We could see billion- or even trillion-dollar numbers, and a possible “platform shift” where new inventions take root everywhere.

Why We’re Excited

Honestly, our whole team is buzzing about what this means.

  • More powerful AI tools could become available to more people, students, researchers, small businesses.
  • Energy-heavy AI models that were shelved could get a second chance.
  • Smarter, greener technology could reshape everything from medicine to transportation.

If sampling and simulation get cheap, entire industries will rethink what’s possible.

For Additional Reading

If you want to go deeper on AI’s energy crisis, sustainability, and future chips, check out some of our other posts:

Sources:

Recruiting redefined; built for high-tech,
high-growth teams

Hire Talent
Job Board