TL;DR:
- Extropic, a hardware startup led by former Alphabet Inc. quantum computing researchers, secures $14.1 million in seed funding.
- Kindred Ventures leads the investment, joined by HOF Capital, Julian Capital, Marque VC, OSS Capital, Valor Equity Partners, and Weekend Fund.
- Extropic aims to create a unique chip optimized for large language models (LLMs) using principles of ‘physics-based computing.’
- The company distances itself from quantum computing, citing scalability challenges.
- Extropic intends to leverage computing noise as an asset and reduce electricity consumption for AI models.
- Specific technical details are undisclosed, but the focus on LLMs suggests competition with Nvidia’s H200 data center processor.
Main AI News:
In a groundbreaking development, Extropic, a pioneering hardware startup, has successfully secured $14.1 million in seed funding. This venture is spearheaded by former members of Alphabet Inc.’s quantum computing research team, bringing a wealth of expertise to the forefront of generative AI. Leading this investment is Kindred Ventures, with significant contributions from HOF Capital, Julian Capital, Marque VC, OSS Capital, Valor Equity Partners, and Weekend Fund. Additionally, more than a dozen other notable backers, including top executives from Adobe Inc., Shopify Inc., and various venture-backed artificial intelligence startups, participated in this funding round.
Extropic, founded last year by Chief Executive Officer Guillaume Verdon, who previously led a quantum computing team at Alphabet’s X research unit, is poised to make a significant impact on the AI hardware landscape. The company’s Chief Technology Officer, Trevor McCourt, boasts a storied history as a researcher at the search giant. While at Alphabet, Verdon and McCourt spearheaded the development of a TensorFlow library, a pivotal step toward running AI models on quantum computing chips.
Rumors abound, but it is widely believed that Extropic is on a mission to create a highly optimized chip specifically designed to power large language models (LLMs). Verdon, in a recent blog post, described Extropic’s technology as a “novel full-stack paradigm of physics-based computing” and revealed that it harnesses the principles of “out-of-equilibrium thermodynamics.” This tantalizing glimpse suggests that Extropic’s chip design incorporates concepts from non-equilibrium thermodynamics, an emerging branch of physics focused on studying phenomena like chemical reactions.
Verdon has been explicit in stating that Extropic’s product is not a quantum computing chip. He pointed out that the timelines for achieving scalability with quantum physics-based computers have grown increasingly protracted. Faced with these challenges, the Extropic team sought an alternative path to practical physics-based computing.
One of the primary impediments to commercially viable quantum chips is their susceptibility to computing errors or noise, rendering complex calculations unreliable. Extropic aims to flip the script by developing a system in which “noise is an asset rather than a liability.”
Although Extropic has yet to divulge detailed technical specifications of its technology, Verdon did share some tantalizing insights. One of the company’s key objectives is to reduce the electricity required to operate AI models. Verdon also hinted at the automation of certain coding tasks, suggesting a future where computers autonomously learn representations of the world, rather than relying on imperative programming.
Crucially, Extropic’s focus on running AI models implies that its technology will involve neural networks crunching data through matrix multiplications—a fundamental mathematical process. AI-optimized chips typically include circuits optimized for these operations, as they are integral to data processing. Additionally, these chips boast ample high-speed memory to facilitate efficient data transfer during processing.
If reports about Extropic’s development of an LLM-optimized processor hold true, it will likely find itself in competition with Nvidia Corp. The chip manufacturer recently unveiled the H200, a data center processor tailored for LLMs. With double the onboard memory compared to Nvidia’s previous flagship graphics card, the H200 represents fierce competition in the AI hardware arena.
Conclusion:
Extropic’s significant seed funding and innovative approach to physics-based computing herald exciting potential in the AI hardware market. By prioritizing noise as an asset and energy efficiency, Extropic aims to carve a niche in the ever-evolving landscape of AI hardware, potentially challenging established players like Nvidia. Investors’ enthusiasm suggests a promising future for the company’s endeavors.
