Harvard Dropouts Raise $120M to Challenge Nvidia’s AI Chip Dominance with Etched

A team of Harvard dropouts has raised $120 million to challenge Nvidia’s dominance in AI chips with their startup, Etched.

Key takeaways: Etched is betting that as AI advances, most of the technology’s computing needs will be met by customized, hard-wired chips called ASICs, which are more efficient than Nvidia’s general-purpose GPUs:

• Etched’s Sohu chip is designed specifically for “transformers,” the core architecture behind AI models like ChatGPT, offering more than 10 times the speed of Nvidia’s GPUs due to its simpler, single-use architecture.
• By specializing in transformers, Etched can dedicate more chip space to raw computing power and less to memory, resulting in fewer inefficient calculations.
• The startup aims to be first to market with its specialized AI chip to secure a competitive advantage against larger players like Nvidia.

Broader context: Despite Nvidia’s current 80% market share in AI chips and its significant resources, startups like Etched are attracting substantial investments to pursue the massive opportunity in the burgeoning AI chip market:

• Venture capitalists invested $6 billion in AI semiconductor companies in 2023, slightly up from $5.7 billion in 2022.
• Other startups challenging Nvidia include Cerebras Systems, which is building physically larger AI chips, and Tenstorrent, which uses RISC-V technology.
• The success of these startups hinges on their ability to offer significant cost savings and performance gains that outweigh the trade-offs in flexibility compared to Nvidia’s GPUs.

Analyzing deeper: While Etched’s founders are confident in their approach, they acknowledge the high-risk nature of their bet on transformers remaining the dominant AI architecture. The startup faces formidable competition from Nvidia and other deep-pocketed players in the chip industry, which have substantial development resources and manufacturing partnerships.

However, if Etched can deliver on its promise of faster, more efficient AI chips and secure early adopters, it could potentially disrupt the market and unlock new AI use cases. The company’s ability to move quickly and be first to market will be critical to its success in this rapidly evolving space.