Welcome back to The Cap Table Newsletter where we break down what's actually happening in startups and private markets.
This week I want to talk about a chip startup that is currently doing something I don't see very often: raising two rounds at two very different prices, at the same time.
Etched just closed a $300 million Series C at a $10.3 billion valuation, led by Sequoia. Simultaneously, the company is also in talks with Jane Street to lead a further round that would value it closer to $20 billion. They're two different raises, running in parallel, on the same company, in the same month.
If you want to start angel investing, apply to our syndicate!
Follow our new and improved Instagram to hear about Trending Deals, Founder Stories, & Investor News!
Subscribe Now to get this newsletter delivered straight to your inbox every other Wednesday morning!
The Cap Table is powered by Apollo.
I was talking to a founder recently about hiring their first GTM person. A few years ago, the answer would've been another SDR. Instead, they were looking for someone who could build the entire revenue engine.
It made sense. AI changed what one person can accomplish. The advantage is no longer adding more people anymore. It's giving the people you already have a better system to work with.
That's why Apollo has become much more than a prospecting tool. It brings research, enrichment, outreach, and execution into one workflow instead of scattering it across half a dozen apps.
Where It Started
In 2023, three Harvard dropouts were pitching a 30-page memo arguing that AI would eventually need chips built specifically for AI models, not general-purpose GPUs. Most investors passed. The company was running month to month, close to out of cash.
It started a year earlier, when Gavin Uberti was on a gap year writing compilers at OctoML and kept hitting the same wall: the chips on the market weren't built for the workloads that actually mattered. He started Zoom calling Chris Zhu, still on campus studying math and computer science. Robert Wachen, who'd been running a nonprofit baking birthday cakes for underprivileged kids while launching startups on the side, joined them. By spring 2023, all three had dropped out, became Thiel Fellows, and moved to the Bay Area.
They raised a $5 million seed round. Then hit a wall. Most VCs didn't buy the thesis, specialized chips for one architecture felt too risky, and they burned through runway with no clear path to the next check. Wachen still tells the early part plainly: He remembers landing in the Bay Area after telling his parents he was leaving school to do a startup, with no office or apartment arranged. He slept on the floor of a friend's unfurnished house. 'I remember staying in my friend's house that they were about to sell, using a towel as a blanket,' he laughs.
What changed the pitch was inference becoming the bottleneck it is today. Training happens once. Inference happens every single time a model runs, and as AI moved from chatbots to agents operating continuously, inference costs started eating companies alive. The thesis that felt too narrow in 2023 suddenly looked like the whole point.
The company kept building quietly after that, closing round after round without announcing any of them, until June 30, when it came out of stealth all at once: $800 million raised across four previously undisclosed rounds, $1 billion in signed customer contracts, and word that its first chips had come back from TSMC successfully manufactured.
The investor list that came with it reads like a who's who: Peter Thiel, Geoffrey Hinton, Andrej Karpathy, Fei-Fei Li, Stanley Druckenmiller, Jane Street, Two Sigma. The team had grown to more than 400 engineers, pulled from Nvidia, Google's TPU group, Broadcom, and SK Hynix, with a former Cypress Semiconductor CTO running engineering and a stated path to gigawatt-scale production by 2027.
But two years earlier, they couldn't close a round.
The Valuation Climb
The valuation mark everyone's measuring this week's number against was set in December, off a $500 million round led by Stripes with Peter Thiel participating at a $5B valuation.
Seven months later, Sequoia is leading the $300 million Series C at $10.3 billion, close to a 2x.
At the same time, Jane Street is reportedly in talks to lead a separate round near $20 billion, which would put the climb closer to 4x.
For a company that only told the public it existed in June, it's growing at an aggressive pace, even by AI infrastructure standards.
The Product
Nvidia's chips are built to do almost anything: train a model, render a video game, run a physics simulation. That flexibility is valuable, but it also means a company running an AI product is paying for a lot of capability it will never touch. Etched's bet is the opposite: build a chip that does exactly one thing, run transformer models at the moment a user hits enter, and trade away everything else for speed.
The number Etched leads with is the one that gets people's attention: a Sohu server can push more than 500,000 tokens per second on a standard AI model, against roughly 23,000 for a comparable Nvidia setup. Call it 20 times faster, by the company's own account, held loosely until an outside customer publishes what it does under real production traffic rather than a controlled demo.
Sohu also launched as a chip that ran transformers and nothing else. The company now says it can handle a wider range of model designs than it originally let on, a meaningfully bigger claim than the one it started with, and the kind of shift worth a plain explanation before anyone writes a check.
My Take
The tradeoff hasn't changed since the founders were pitching a memo nobody would fund. Their Sohu hasn't shipped at real scale yet, but when it does it could take the AI industry by storm since they have a billion dollars in signed contracts.
But Etched isn't alone. Cerebras, Groq, Tenstorrent, and SambaNova are all chasing versions of the same gap, and even Google is reportedly building something similar for Gemini.
What sticks with me is the distance traveled. A team that couldn't close a round two years ago is now being priced by two different lead investors in the same month, at numbers ten billion dollars apart. That's not a good run of press. That's what happens when a thesis nobody wanted to underwrite turns out to be exactly right, right as the market finally needs it. Doesn't mean the risk is gone. It means enough people with real conviction and real capital are betting it's already resolved in Etched's favor.
The Cap Table is powered by Lovable.
Ever since I announced the Two Roads Capital Scout Program, people have been sending me deals from everywhere. Emails, LinkedIn messages, X DMs, even texts after meeting a founder at dinner. I'd remember someone mentioning a company, but not where they sent it. A few deals slipped through simply because there wasn't one place for everything to live.
So I opened Lovable and built exactly what I needed. In one afternoon, I had a Scout Program on our website where people can submit the company, stage, round size, deck, and why they're excited. Every submission lands in my inbox instantly, and each week I get a digest of everything that came in, so nothing gets lost.
A year ago, I would've assumed something like that needed an engineer. Now, if I run into an operational problem, my first instinct is to see if I can build it myself.
👋 That’s all for now friends! See you next week.
In the meantime, Follow our instagram to see the latest founder and VC updates.
💌 Join our subscribers and sign up for this weekly Cap Table Newsletter if you haven’t already!
Also if you are interested in starting to Angel Invest you can apply to our syndicate to see our weekly deal flow!
Disclaimer: The Cap Table DOES NOT provide financial advice. All content is for informational purposes only. The Cap Table is not a registered investment, legal, or tax advisor or a broker/dealer.