source: techcrunch ai: ai chip startup etched defies skeptics, hits $10.3b valuation from big-name investors
level: business
etched, founded by three harvard dropouts in 2022, closed a $300 million series c round led by sequoia, reaching a $10.3 billion valuation. the round included andreessen horowitz, sk hynix, jane street, and diffusion capital. this follows a $500 million round in december at a $5 billion valuation. the company says it has manufactured its first chips, with systems being tested by clients and $1 billion in orders already booked.
the startup designs full systems for ai inference, not just chips. its approach splits inference into prefill and decode stages. a custom prefill chip runs at lower voltage to pack in more transistors and reduce heat. for decode, etched built a new memory and interconnect technology called cluster-scale memory, letting many chips share a fast, low-latency memory pool. the systems can run any ai model, including mixture of experts and non-transformer architectures like mamba.
etched faced skepticism for building chips tailored to transformer models, but the concept is gaining traction, with google reportedly working on a similar idea. early access was limited to investors and customers, helping attract backers like andrej karpathy and geoffrey hinton after private demos. the company now has 400 employees and operates a 2 megawatt data center, with a new 10mw facility in milpitas. co-founder robert wachen says scaling remains a challenge but the team is committed to the long road ahead.
why it matters: specialized inference chips could lower costs and speed for ai services, challenging nvidia's dominance in the growing market for transformer-based models.
source: techcrunch ai: ai chip startup etched defies skeptics, hits $10.3b valuation from big-name investors