Nvidia's AI chips are being challenged by $120 million raised by college dropouts.

• A startup founded by college dropouts has secured $120 million in funding from investors led by Primary Venture Partners to develop an AI chip that competes with Nvidia.
• Gavin Uberti, CEO of the etched startup, stated that they believe that as AI technology advances, the majority of its power-intensive computing demands will be met through the use of customized, hard-wired chips known as ASICs.
• "If transformers disappear, we'll perish, as Uberti stated to CNBC. However, if they remain, we'll become the largest corporation ever."

The stock market has experienced a significant increase in value, with the hottest story on Wall Street rising eightfold since the end of 2022 and reaching a market cap of over $3 trillion this month.

A startup founded by Harvard dropouts has secured $120 million in funding to develop a competitive chip and challenge Nvidia in the field of artificial intelligence.

Apple-based Etched is developing a chip called Sohu for training and deploying AI models using "transformers," the core architecture driving advancements like OpenAI's ChatGPT.

Gavin Uberti, co-founder and CEO, stated that as AI technology advances, the majority of its power-intensive computing demands will be met through customized, hard-wired chips known as ASICs. These chips are specifically designed to execute a single AI model, in contrast to general-purpose graphics processing units (GPUs) from Nvidia, which are more versatile but also significantly more expensive.

"Uberti stated in an interview that the company is taking the largest risk in AI. If transformers become obsolete, the company will fail. However, if they remain relevant, the company will become the largest in history."

Uberti and his co-founders are aware that they are taking a high-risk bet by going up against some of the most well-funded and competitive companies in the world. Although they have raised $120 million in a series A, it is only a fraction of Nvidia's revenue generated in half a day, which has more than tripled annually for three consecutive quarters, reaching $26 billion in the latest period.

Nvidia dominates the AI chip market with over 80% market share, while Etched is among startups attracting capital for the growing opportunity. Primary Venture Partners led the funding round, with Peter Thiel, Stanley Druckenmiller, and Cruise founder Kyle Vogt also investing.

Although Nvidia has a head start and some developers describe a moat, new chipmakers are still pushing forward, mainly because the opportunity is too big to ignore. Some of these chip startups challenging Nvidia include Cerebras Systems, which is developing a physically larger AI chip, and Tenstorrent, which is utilizing RISC-V technology to create AI chips.

"Robert Wachen, Etched's operating chief, stated that the reason we were so enthusiastic about what we were doing, why we dropped out of school, and convinced so many people to leave chip projects is because this is the most crucial aspect to be working on. The entire future of technology will be determined by the infrastructure's ability to handle the scale."

The semiconductor industry has historically been challenging for startups due to the lengthy development cycles, substantial capital needed to manufacture a chip, and the necessity to collaborate with a limited number of manufacturing partners, such as Intel, which is producing Etched's chip.

In 2023, $6 billion was invested in AI semiconductor companies by venture capitalists, a slight increase from the $5.7 billion invested in 2022, as per PitchBook data.

Hard coded

After completing a summer internship working on compilers, Uberti became interested in low-level hardware and later co-founded Etched with Chris Zhu.

In 2022, the pair dropped out of Harvard and added Uberti's college roommate, Wachen. They quickly began hiring chip industry veterans. The company settled in Cupertino and now has 35 employees. It offers housing stipends to new hires.

Uberti stated that when ChatGPT was released and Nvidia's stock price soared, and particularly when every other model that emerged would be a transformer, they were in the perfect location at the perfect moment.

Sohu is set to be launched by Etched, with the founders promising to reveal something later this year. The startup is also actively seeking customers and reports that technology companies are eager to explore new AI chips.

To be successful, companies investing billions in GPUs must achieve substantial savings by designing custom chips for their AI models, even if it means sacrificing some flexibility.

Etched's Sohu chip, by focusing on transformers, can allocate less memory space and more to transistors that determine a chip's computational power, according to Uberti.

The chip's large core reduces the need for a streaming multiprocessor to coordinate calculations among different cores, resulting in increased efficiency.

The impact of specialized AI chips could be similar to how custom chips called ASICs, first introduced in 2013 for mining bitcoin or ether, reduced the demand for Nvidia GPUs, says Uberti.

As the use of AI software increases, the demand for chips to run these models is expected to rise, particularly among the founders of Etched.

By embedding AI architecture onto their chips, they claim to decrease latency, enabling new applications such as AI agents and real-time voice conversations. Etched's chips are reportedly 10 times quicker than Nvidia's GPUs due to their streamlined architecture and single-purpose design.

Nvidia, a company with a large development team and significant financial resources, is among the valuable companies that Etched is targeting.

Etched, with a countdown timer in its headquarters, must accelerate.

Uberti stated that the key to success in the specialized AI chip market and future markets is to be the first to bring products to market.

WATCH: Nvidia's market cap surge