Analysis
A seed round used to describe a stage: pre-product, pre-revenue, small check, high ownership. In 2026 it describes only the first two.
Three examples, all from the last twelve months:
- Ineffable Intelligence -- $1.1B seed at $5.1B (co-led by Sequoia and Lightspeed, with Nvidia, DST Global, Index, Google and the UK's Sovereign AI Fund): reinforcement-learning lab in London, incorporated November 2025, no product. Europe's largest seed ever.
- Thinking Machines -- roughly $2B seed at $12B (led by Andreessen Horowitz, with Nvidia, GV, Lightspeed and Conviction), now in talks with Accel for $1B at $40B: founded 2025 by Mira Murati, ships the Tinker fine-tuning platform.
- AfterQuery -- [$3.2B valuation](https://techcrunch.com/2026/09/01/afterquery-reportedly-becomes-y-combinators-fastest-ever-unicorn-now-valued-at-3-2b/): Y Combinator's fastest-ever unicorn, in expert data for model training.
“It was targeting a $50 billion valuation in late 2025 and is now negotiating at $40 billion, against a run rate above $100 million.”
What is actually being priced
Not traction. What these rounds buy is a founder with a specific research track record and the ability to recruit a bench that no other company can assemble. David Silver built AlphaGo and AlphaZero; Mira Murati was OpenAI's CTO. The valuation is a hiring instrument -- it sets the strike price low enough, relative to the expected next mark, that senior researchers will leave nine-figure comp packages at Google and OpenAI to join.
That is a coherent thesis, and it explains a structure that otherwise looks reckless. It also explains today's news that Ineffable handed cofounder titles to six people who joined after incorporation: when the currency is equity and status, you spend both.
What it costs everyone downstream
The problem for the rest of the market is compression. A pre-product mark in the billions resets the reference price for every AI company one tier below it, and it does so without producing a single revenue datapoint anyone can underwrite.
Thinking Machines is the cautionary version. It was targeting a $50 billion valuation in late 2025 and is now negotiating at $40 billion, against a run rate above $100 million. That is a down-ask, not a markup, and it happened inside a year.
For seed funds writing $2 million to $5 million checks, none of this is accessible -- there is no ownership to be had in a $1.1 billion round -- and its main effect is to make the founders they do back believe those are the terms. The gap between a $5 billion research bet and a $15 million pre-seed has never been wider, and the vocabulary has never been more identical.