It shouldn’t come as much of a surprise that DARPA is hurriedly throwing in with the ongoing AI trend…

In fact, Bloomberg reported this week that the US Defense Advanced Research Projects Agency is looking for an additional $70 million to help further its “quest to develop an ultra-efficient chip for artificial intelligence technology”.

It is backing a company called EnCharge AI Inc., which seeks to build chips specifically suited for AI.

The company is pioneering in-memory computing—a technology mainly theoretical until recently that could significantly reduce the energy consumption of AI processing chips. It started from work conducted at Princeton. 

CEO Naveen Verma, who is also a professor of electrical and computer engineering at Princeton, noted that EnCharge was launched with substantial support from the Defense Advanced Research Projects Agency and the DOD.

The report says that this technology has the potential to perform data processing directly where it is stored, thereby conserving energy by eliminating the need to move data across different components.

EnCharge is not alone in its research; major firms like Taiwan Semiconductor Manufacturing Co., Intel Corp., and Samsung Electronics Co. are also exploring similar advancements.

Currently, the U.S. government, emphasizing a shift towards boosting American manufacturing, has shown significant interest in such innovations, providing over $23 million in funding to EnCharge, with additional private investment nearing double that amount from entities including Raytheon Technologies’ venture arm.

This is part of a broader initiative backing startups and enhancing the domestic production capabilities.

In addition to the technology’s appeal to private sector firms, the U.S. government sees crucial applications in defense, particularly for running AI-driven military applications in power-constrained environments such as remote areas and aircraft, Verma explained.

EnCharge is preparing for its next investment round and is in pursuit of strategic investors, although Verma did not disclose specifics about the funding discussions or the company’s valuation.

The startup has already started deploying its chips, which are tailored not for training AI models like those of OpenAI or Anthropic, but for applications utilizing these models for predictive tasks.

With a workforce of 50, EnCharge anticipates launching products manufactured by TSMC in the coming year and currently has customers evaluating its technology.