The Hype Train Is Moving Fast
If you thought the AI gold rush had settled down, Etched would like a word. The four-year-old startup, which only recently closed a massive $700 million round at a $21 billion valuation, is already looking at offers that could push its worth to $40 billion or even $50 billion. Real talk: this speed is wild, but Etched is playing a different game than your average software startup.
Why Everyone Is Obsessed
Ok, but why are investors literally throwing money at them? Etched is trying to do something incredibly expensive: building full AI hardware systems using their own custom-made chips from the ground up.
They aren't just selling a promise; they're aiming to take on the king of the space: Nvidia. They’ve already got real traction, including a $1 billion order book. One of their notable customers is Jane Street, the quant trading firm that actually led their last investment round. For a firm like that, a microsecond advantage in processing speed equals major profit, and Etched claims their tech handles inference—the "thinking" part of AI—faster and cheaper than anything Nvidia currently offers.
The Secret Sauce
Etched’s team is lowkey stacked with talent they’ve poached from the competition. About 15% of their 400-person workforce are former Nvidia engineers. The co-founders, Gavin Uberti and Chris Zhu, met in a Harvard math class, and along with COO Robert Wachen, they dropped out to build this thing.
They’re also moving offline, operating a 10-megawatt datacenter in Silicon Valley and setting up shop in Taiwan to stay close to TSMC for manufacturing. With these new potential funding offers, the company could secure enough cash to fuel operations for the next three and a half years.
Why it matters
We’re seeing a pattern here: buzzy AI companies are splitting funding into rapid-fire rounds to keep their valuations soaring. While a $40 billion-plus valuation is massive, it shows just how much capital is required to actually build the physical "brains" of the AI revolution. If Etched can actually pull off their performance claims at scale, they might just give the current chip giants a serious run for their money.
