(Bloomberg) -- A Silicon Valley startup backed by the US Defense Advanced Research Projects Agency is seeking at least $70 million more from investors in its quest to to develop an ultra-efficient chip for artificial intelligence technology, according to documents reviewed by Bloomberg. 

The company, EnCharge AI Inc., joins the growing technological race to build better chips to power the current wave of AI technology. Computing power has become a scarce and precious resource in the AI world, helping send chipmaker Nvidia Corp.’s market cap soaring past $2 trillion and spurring OpenAI boss Sam Altman to seek billions for chipmaking endeavors. 

The US government is also interested, backing startups in addition to spending billions to incentivize American manufacturing. EnCharge was “born out of the support of Darpa and the DoD,” or Defense Department, said Chief Executive Officer Naveen Verma. The startup spun out of Verma’s work at Princeton University, where he’s still who is still a professor of electrical and computer engineering. “Even as we were spinning out, Darpa was helpful in getting us our first investors.”

Verma said that the startup is seeking strategic investors in its upcoming round. He declined to comment further on the funding talks or the company’s valuation. 

EnCharge is one of a number of companies working on so-called in-memory computing, an until-now largely academic undertaking that could substantially decrease the energy used by chips for AI processes. In addition to startups, large companies including Taiwan Semiconductor Manufacturing Co., Intel Corp. and Samsung Electronics Co. are also researching the technology. The chips aim to save power by processing data where it’s stored, instead of transporting it to separate locations.

Read More: AI Energy Crisis Boosts Interest in Chips That Do It All

In addition to the mounting demands of private companies, the US government also has pressing needs for fast chips, Verma said. For example, the ability to run military applications of AI where power is limited, like in remote regions and aboard aircraft, is an increasingly critical defense requirement.

EnCharge’s first generation of chips are not designed for OpenAI, Anthropic and other companies training AI models, but for applications that use those models to make predictions.

Verma said the 50-person company already has customers evaluating its chip and technology, and plans to begin releasing products manufactured by TSMC next year.

Darpa’s recent grant to EnCharge increases total Defense Department funding to the startup to more than $23 million, with Raytheon Technologies’ venture arm and other private investors plowing in roughly twice that amount since EnCharge was quietly founded three years ago in Santa Clara, California.

Competitors to EnCharge include Netherlands-based Axelera AI BV and d-Matrix Corp., which has raised more than $160 million. 

©2024 Bloomberg L.P.