The chip is placed on a circuit board as wide as a Big Mac burger. The silicon wafer is only the size of a pinky fingernail and is covered with strange components. This is not an ordinary semiconductor transistor or quantum chip superconducting component, but a core component from a new computing paradigm-thermodynamic computing. Unlike quantum computing, which attempts to reduce random thermodynamic fluctuations in electronic components, thermodynamic computing aims to exploit these fluctuations and break through the limitations of traditional binary zeros and ones.
On Verdon's chip, an array of square elements tens of microns wide is eye-catching. These elements will be used to generate "programmable randomness." Combined with traditional computers, they can efficiently simulate uncertainty and play a key role in fields such as weather, financial market modeling, and artificial intelligence. Some academic labs have now built prototypes of thermodynamic hardware, including simple neural networks.
Extropic plans to bring the first chips to market later in 2025. Verdon is very optimistic about its prospects and believes that eventually entire workloads and even language models will be able to run on this hardware. The speed of its development is amazing, and there is a story of "defection" behind it.
Previously, while Verdon was studying for a doctoral degree at the University of Waterloo in Canada, he was recruited by Google to participate in a project using quantum computers to assist artificial intelligence. In early 2022, the team made important progress. Verdon led the development of the artificial intelligence software Tensor-FlowQuantum suitable for quantum computers. The Google hardware team is also improving quantum error correction technology. However, during the same period, many theoretical discoveries in the academic community have cast a shadow on the prospects of quantum computing. For example, Evan Tang's research shows that the acceleration effect of the quantum recommendation algorithm, which has high hopes in quantum machine learning, is far less than expected. This led many people such as Verdon to believe that the quantum computing bubble had burst, and Verdon therefore "defected" to the field of thermodynamics computing. Extropic has attracted many talents who have left the field of quantum computing.
In January 2022, Verdon created an account on the X platform under the name "BasedBeff". With his humorous remarks, he attracted many like-minded people and also caused controversy. In December 2023, Forbes exposed his true identity through voiceprint recognition. Verdon did not escape, but used it to promote Extropic. At present, the company has raised US$14.1 million in seed funding and has about 20 engineers working full-time on research and development.
At Extropic's hardware lab, CTO and co-founder Trevor McCourt, who also came from Google's quantum computing project, joined the company after becoming disillusioned with quantum computing. Although Google's newly released "Willow" quantum computing chip can perform error correction on a 105-qubit scale, it requires extremely low-temperature cooling, and its advantages are still uncertain. When Extropic's chips are delivered to early adopters later this year, they are expected to demonstrate value in the fields of high-tech trading and medical research by running probabilistic simulation algorithms.