Etched's $10.3B AI Chip Valuation
AI Chip Startup Etched Secures $300M Series C to Challenge Nvidia at $10.3B Valuation
Anonymous
Key Points
1
SK Hynix is participating in the funding round, a notable move given that SK Hynix supplies the HBM3E memory used in Etched's hardware.
2
There are discussions of a follow-on round targeting about a $20 billion valuation, led by Jane Street, according to a Wall Street Journal report.
3
Etched has manufactured its first chips through TSMC and already booked roughly $1 billion in orders, signaling rapid initial client traction.
4
The founders are Harvard dropouts who started from a garage-like setup and have grown the team to about 400 employees with a 2-megawatt data center.
5
Etched describes its hardware as architecture-agnostic, capable of running a range of AI models beyond transformers, including non-transformer designs and MoE models like Qwen and Mamba.
Etched, a San Jose-based AI chip startup, has raised $300 million in a Series C round at a $10.3 billion post-money valuation, according to TechStartups. The round was led by Sequoia, with participation from a16z, Jane Street, SK Hynix, and Diffusion. It marks the highest valuation ever for a Sequoia-led Series C, per TipRanks.
Etched builds chips designed specifically for AI inference — the process of running a trained model to generate outputs. Unlike Nvidia's general-purpose GPUs, Etched's hardware is purpose-built to run transformer-based AI models faster and cheaper. The company emerged from stealth less than a month ago and already claims roughly $1 billion in orders.
Harvard Dropouts Who Took On Nvidia From a Garage
Etched was founded by Gavin Uberti and Chris Zhu, two Harvard dropouts who built the company from a garage-like setup, according to TechStartups. The team has since grown to about 400 employees. Etched also runs a 2-megawatt data center and has manufactured its first chips through TSMC, the world's leading chipmaker.
The startup is now building out serious infrastructure. Plans include an 80,000-square-foot facility near its San Jose headquarters and a 10-megawatt site in Milpitas. That second location will house a new product lab and an in-house assembly line for circuit boards.
SK Hynix Joins a Round Stacked With Strategic Investors
SK Hynix's participation stands out. The South Korean memory giant supplies the HBM3E chips that go inside Etched's hardware, according to The Next Web. Having a key supplier invest directly signals strong alignment between the two companies as Etched scales production.
Jane Street, the quantitative trading firm, also joined the round. That firm is now reportedly in talks to lead a follow-on round targeting a valuation of around $20 billion, according to TipRanks. If that round closes, it would nearly double Etched's current valuation in a matter of months.
Chips That Go Beyond Transformers, Makers Say
Etched calls its hardware "architecture-agnostic." That means the chips are not locked into one type of AI model. The company says its hardware can run transformer models, mixture-of-experts models like Qwen, and newer designs like Mamba, per The Next Web.
That flexibility matters. The AI model landscape shifts fast. A chip that only runs one model type risks becoming obsolete. Etched is betting that broad compatibility — combined with raw inference speed — gives it a durable edge over both Nvidia and other AI chip startups.
$1 Billion in Orders Before Large-Scale Deployments Even Begin
Etched says it has already booked roughly $1 billion in orders, according to MarketScreener. That traction is striking for a company that just came out of stealth. The fresh $300 million will go toward fulfilling those orders and ramping up production capacity at its new facilities.
Demand for fast, cheap AI inference is exploding. Every major tech company is racing to deploy large language models at scale. Etched is positioning itself as the infrastructure layer that makes that possible — without relying on Nvidia. Whether it can deliver at that scale is the key question ahead.
Sentiment
Mostly PositiveNEGATIVENEUTRALPOSITIVE
Publishers
22
Articles
76
Reach
98
A
AnonymousAuthor