Story

AI Researchers Launch Physics-Based Model After Turning Down Bezos-Backed Prometheus

ENTHMSVIIDZHZH-TWJAKOHI
Aug 25, 20262 min read
AI Researchers Launch Physics-Based Model After Turning Down Bezos-Backed Prometheus
Text size
Background

A pair of AI researchers, who were once offered leadership roles at the Jeff Bezos-backed Project Prometheus, have unveiled their own company, Accelerated Understanding Inc., and a new type of artificial intelligence. The model is designed to understand the laws of physics rather than human language, a fundamental departure from mainstream systems like ChatGPT.

A New Approach to AI

Accelerated Understanding's model is built on a technology co-pioneered by founder Anima Anandkumar called neural operators, forgoing the Google-invented Transformer architecture that underpins most large language models. The company claims this physics-centric approach allows its AI to process vastly larger datasets related to physical phenomena.

In an exclusive interview with Reuters, the founders stated their system handled 5 trillion pieces of data in a single test prompt. This represents a data input capacity approximately 5 million times larger than what flagship models from Google and Anthropic can typically process. "The language-centric view of intelligence is humans at the center. Putting physics at the center is a nature-centric view," said Anandkumar, a professor at Caltech.

The Prometheus Connection

Before launching their company, Anandkumar and her husband and co-founder, Benedikt Jenik, were courted to lead Project Prometheus, a venture co-founded by Jeff Bezos. According to an offer letter seen by Reuters, the proposal included a 35% stake in the company, a combined annual salary of $1 million (set to double after three months), and access to over $2 billion in committed financing through a Series B round.

Sample IUX Markets – In-articleAd

The duo ultimately declined the offer to continue building their own company. Prometheus, which aims to automate the manufacturing of complex physical systems, went on to raise a $12 billion Series B in June 2026, according to the report.

Enterprise Applications and Origins

Accelerated Understanding is initially focused on enterprise applications rather than a consumer product. The founders believe their AI's intrinsic grasp of physics can optimize complex industrial processes, with potential uses including:

  • Advanced semiconductor design
  • Robotics and automation
  • Extreme weather prediction
  • Geological data analysis for energy companies

Anandkumar's work on this technology was encouraged by Nvidia CEO Jensen Huang during her five-year tenure as a director at the chipmaker. Huang reportedly presented her research on neural operators at Nvidia's GTC Conference in 2021.

Read next

More on Stocks
Back to latest news

LATEST