Etched raises $120 million to build chip to take on Nvidia in AI

All copyrighted images used with permission of the respective copyright holders.

Harvard Dropouts Take On Nvidia With $120 Million Bet on AI Chips

A two-year-old startup founded by Harvard dropouts is making a bold play in the rapidly growing artificial intelligence (AI) chip market. Etched, headquartered in Cupertino, California, has raised $120 million in venture funding to develop its own chip, Sohu, specifically designed to train and deploy AI models using transformers, the architecture powering AI breakthroughs like ChatGPT.

The company’s ambitious goal is to challenge the dominance of Nvidia, the current leader in the AI chip market. Nvidia’s stock has skyrocketed in recent months, reaching a market capitalization of over $3 trillion, fueled by the phenomenal growth of AI applications. But Etched’s founders believe the future of AI rests in ASICs (Application-Specific Integrated Circuits), which are customized chips designed for specific tasks, offering higher efficiency and lower cost compared to Nvidia’s general-purpose GPUs (Graphics Processing Units).

Here are the key takeaways from this aggressive entry into the AI chip arena:

  • Betting on Transformers: Etched is focused on transformers, a crucial AI architecture, believing it will define the future of AI. This makes them a high-risk, potentially high-reward bet.
  • Efficiency through Specialization: Sohu aims to achieve superior efficiency by focusing on a single task, transformers, sacrificing some flexibility for significant performance improvements.
  • David vs. Goliath: Etched faces an uphill battle against Nvidia, a behemoth with a huge market share and vast resources.
  • The Stakes Are High: With AI’s exponential growth, the demand for specialized chips is expected to skyrocket, prompting a wave of new entrants like Etched.

Taking On The King

Etched’s founders, Gavin Uberti, Robert Wachen, and Chris Zhu, recognized the potential of AI chips while working on compilers and witnessed firsthand the growing need for specialized hardware. Dropping out of Harvard in 2022, they assembled a team of experienced chip designers and are now pushing the boundaries of AI chip development.

While Etched’s $120 million Series A funding might seem substantial, it pales in comparison to Nvidia’s revenue, which is measured in billions annually. The company is developing its Sohu chip, which they claim is 10 times faster than Nvidia’s GPUs, attributed to its simplified architecture and specialized focus on transformers. Sohu’s unique design allows it to dedicate more space to computing power and reduce the reliance on memory, leading to faster processing and reduced latency.

The Challenges and Opportunities

Etched faces several challenges in its quest to displace Nvidia. The chip industry is notoriously capital-intensive, with long development cycles and a limited number of manufacturing partners, such as TSMC (Taiwan Semiconductor Manufacturing Co.), which is currently building Etched’s chip. The company also needs to convince potential customers to invest in custom chips, requiring them to trade some flexibility for enhanced performance.

But the AI chip market offers tremendous potential. Venture capitalists have invested billions of dollars in AI semiconductor companies, recognizing the incredible growth driven by advanced AI models. As AI technology advances and models become more complex, the demand for specialized chips will only increase. This creates a fertile ground for startups like Etched to break into the market.

Etched’s Strategy

The company’s strategy hinges on being the first mover in the specialized AI chip market. They aim to bring Sohu to market sooner than their competitors, securing early adopters and establishing a foothold in a rapidly evolving industry. Etched is confident that their specialized approach will provide a compelling alternative to Nvidia’s GPUs, particularly in scenarios where performance and efficiency are critical.

A key advantage for Etched is the increasing adoption of transformers for AI tasks. As AI models become more complex and the computational demands grow, transformers are expected to remain a dominant architecture. If Etched can deliver on its promise of faster and more efficient chips specifically designed for transformers, they could attract customers seeking to optimize their AI deployments.

The Future of AI Chips

Etched represents the growing wave of startups entering the AI chip market, taking on established giants like Nvidia. The competition is fierce, but the potential rewards are substantial. The future of AI depends on the development and availability of powerful, efficient chips, and startups like Etched play a crucial role in pushing the boundaries of innovation. If they can deliver on their promises, they could disrupt the AI chip market and reshape the landscape of AI technology.

This contest between established behemoths and ambitious startups underscores the dynamic nature of the AI revolution. The AI chip market is poised for explosive growth, and companies like Etched are racing against time and resources to stake their claim in this rapidly evolving landscape. The success of these startups, like Etched, will not only influence the development of future AI models but also shape the future of computing as a whole.

Article Reference

Brian Adams
Brian Adams
Brian Adams is a technology writer with a passion for exploring new innovations and trends. His articles cover a wide range of tech topics, making complex concepts accessible to a broad audience. Brian's engaging writing style and thorough research make his pieces a must-read for tech enthusiasts.