# Seattle's Ascend raises a deliberately small $17M third fund for the earliest AI and deep tech startups

> The firm writes first checks of $250,000 to $750,000 and is leaning into physical AI, which its partners say needs more than software startups do.

- URL: https://devisdead.ai/news/seattle-s-ascend-raises-a-deliberately-small-17m-third-fund-for-the-earliest-ai-
- Published: 2026-11-11T20:22:38Z
- Author: DevIsDead Newsroom
- Section: Startups & Money
- Tags: venture capital, pre-seed, physical ai, robotics
- Originally reported by Unite.AI (https://www.unite.ai/ascend-announces-17m-fund-iii-for-inception-stage-ai-and-deep-tech/).

Seattle venture firm Ascend has announced a $17 million Fund III for inception-stage AI and deep tech companies, according to [Unite.AI](https://www.unite.ai/ascend-announces-17m-fund-iii-for-inception-stage-ai-and-deep-tech/). Founders, family offices and funds of funds put money in.

## Small on purpose

Founding general partner Kirby Winfield says the size is a choice: Ascend wants to know its founders personally and do the hands-on early work itself. He calls pre-seed "still a craft business," and says the firm looks for founders good enough that Ascend's own people would leave their jobs to work for them.

- **Typical first check:** $250,000 to $750,000.
- **Portfolio:** more than 70 companies, 15 of them acquired, per its website.
- **Track record:** Winfield calls Funds II and III top-decile, though the announcement gives no return figures or method.

Winfield founded Ascend in 2019 after building and selling four tech companies. Since its second fund in 2022 it has hired partner and chief of staff Jen Haller and principal Nate Bek, who focuses on defense, robotics, applied AI and AI infrastructure, and expanded beyond Seattle to San Francisco, New York and elsewhere.

## The physical AI bet

In a July essay, Bek and Ben Eisinger of Plug and Play Seattle framed the city's history in acts: ships and planes, then cloud and enterprise software, and maybe next, intelligent machines. Portfolio examples include Overland AI, building off-road autonomy so one operator can run several vehicles, and Sabi, working on a sensor-packed cap meant to turn thoughts into text.

The essay argued hardware startups need things software companies don't: industrial space, freight and airfield access, open land for vehicle tests, deep water for subsea work, early capital, and lighter permitting. Whether a small, first-check fund can carry founders with needs that heavy is the open question.

**[Scoreboard](/scoreboard) update:** Draw. Investors still bet on people first, and the people are building machines.

Stay undead.