Nvidia has been the focal point of recent Wall Street discussions, growing exponentially in value since the conclusion of 2022 and surpassing a $3 trillion market cap this month.

A startup established two years ago by Harvard ex-students has secured $120 million in financing to develop a competitive processor and challenge Nvidia in the field of artificial intelligence.

Located in Cupertino, California — known as Apple’s headquarters — Etched is in the process of creating a processor named Sohu. This chip is designed for training and implementing AI models utilizing “transformers,” the pivotal architecture behind advancements such as OpenAI’s ChatGPT.

CEO Gavin Uberti, a co-founder, expressed that as AI progresses, the majority of the technology’s energy-intensive computational needs will be met by specialty, embedded chips named ASICs. These chips excel at executing specific AI models they were designed for, unlike Nvidia’s more versatile but costlier general-purpose GPUs.

“We are placing a large bet on AI,” stated Uberti in an interview. “If transformers become obsolete, we’re in trouble. But if they continue to be relevant, we could become one of the largest companies in history.”

Etched and its co-founders understand the high stakes of their endeavor, going up against some of the most well-funded and competitive corporations globally. While $120 million is a substantial amount for a series A funding round, it’s essentially what Nvidia earns in revenue in just twelve hours. Nvidia’s sales have more than tripled annually for the last three quarters, exceeding $26 billion most recently.

According to estimates, Nvidia holds more than 80% of the AI chip market. Etched is part of a consortium of startups attracting investments to capitalize on the expanding opportunity. Primary Venture Partners spearheaded this funding round, with backing from Peter Thiel, Stanley Druckenmiller, and Cruise founder Kyle Vogt.

Uberti and Chris Zhu commenced work on a chip enterprise after Uberti completed a summer internship focused on compilers. This experience exposed him to the fundamental hardware concepts that formed the basis of Etched.

The duo departed from Harvard in 2022, later joined by Wachen, a former college roommate of Uberti. They rapidly started recruiting industry experts in chip technology. The business settled in Cupertino and presently employs 35 personnel. The company offers housing allowances to new recruits.

“The emergence of ChatGPT, the meteoric rise of Nvidia’s stock, along with the prevalence of transformer models, placed us in an opportune position,” Uberti remarked.

Etched is gearing up to introduce its Sohu chip to the market, with the founders aiming to showcase progress later in the year. The startup is actively seeking clients, with technology firms displaying eagerness to explore new AI processors.

For this venture to succeed, companies that invest billions in GPUs must witness substantial savings in developing custom chips tailored to their specific AI models and be inclined to accept trade-offs in flexibility.

By channeling their focus on transformers, which facilitate predictable data flow from the chip to memory, Etched’s Sohu chip can allocate less memory space and more transistors to boost raw computing capabilities, Uberti explained.

In addition to efficiency gains, the chip’s singular larger core reduces inefficient calculations compared to Nvidia’s GPUs. This development could have a similar impact to the introduction of ASICs in 2013, which decreased the demand for Nvidia GPUs in bitcoin or ether mining operations.

The Etched founders anticipate a surge in the demand for chips to operate these models, especially once they are employed to support AI software millions of times per minute.

They assert that by embedding the AI architecture directly into the chip, their device can minimize response latency, unlocking new applications such as AI agents and real-time conversational interfaces. Etched asserts its chips deliver more than a tenfold speed increase over Nvidia’s GPUs due to their streamlined architecture and singular use case.

However, competing with some of the most valuable corporations globally, like Nvidia, poses a formidable challenge. These tech giants boast significant development teams and financial resources essential for manufacturing and enhancing chips regularly.