Seattle's Ascend Closes $17M Fund III for Inception-Stage AI

By Daniel Okafor

Seattle venture firm Ascend announced a $17 million Fund III on November 11, 2026, a deliberately small vehicle aimed at inception-stage AI and deep technology companies. The fund's backers include funds of funds, family offices and founders. The size is the strategy, not a constraint.

Founding general partner Kirby Winfield said the firm wants to know founders personally and do the early-stage work itself. "We believe pre-seed is still a craft business," he wrote. He added that Ascend looks for people its own team would quit their jobs to work for.

The firm writes initial checks of $250,000 to $750,000. Its website lists more than 70 portfolio companies and says 15 of them have been acquired.

Why keep the fund small?

The answer lies in how inception-stage investing works. Ascend treats the first check as a craft: few partners, deep relationships, hands-on support before a company has traction. A $17 million pool forces discipline. It rules out the larger rounds that later-stage firms chase and keeps the firm anchored to its founding role.

Winfield started Ascend in 2019 after building and exiting four technology companies. The firm raised its second fund in 2022, then hired Jen Haller and Nate Bek, sharpened its thesis and portfolio construction, and expanded from Seattle into San Francisco, New York and other markets.

Haller serves as partner and chief of staff. Bek is a principal focused on inception-stage defense, robotics, applied AI and AI infrastructure. The firm invests across the United States while keeping Seattle as its base, and its mandate spans AI, software and frontier technology, from intelligent infrastructure to robotics and autonomy.

Winfield described both the second and third funds as top-decile performers. The announcement did not include return figures or a benchmarking methodology to support that characterization.

What is the 'Seattle's Third Act' thesis?

An earlier essay explains how physical AI fits the mandate. In "Seattle's Third Act," published July 25, 2026, Nate Bek and co-author Ben Eisinger, ventures lead at Plug and Play Seattle, mapped the region's economic history in three stages. The first act was maritime and aerospace. The second was cloud computing and enterprise software. The prospective third act is intelligent machines built from software-enabled hardware.

Ascend's portfolio shows how that thesis crosses sectors:

• Overland AI is developing off-road autonomy intended to let one operator command multiple vehicles. Ascend classifies it as a robotics and hardware investment.
• Endurance Energy is developing geothermal systems that draw on seafloor hydrothermal resources. It sits within the firm's climate and energy category.
• Sabi is developing a wearable cap containing 100,000 sensors, with the goal of turning thoughts into text. Ascend also classifies it as robotics and hardware.

Bek and Eisinger argued that physical-AI companies need resources software startups never required. Their list includes high-bay industrial space, freight and airfield access, open terrain for vehicle testing, and deepwater for subsea tests. Their essay also named three requirements for a stronger regional ecosystem: early capital for capital-intensive hardware, greater founder risk-taking, and policies that reduce permitting and other operating obstacles.

Can a craft model fund heavy hardware?

Fund III pairs two ideas that sit in tension. On one side sits a deliberately small, relationship-driven investment model built for speed and conviction. On the other sits a mandate that includes capital-intensive machines and infrastructure — companies that need factories, test ranges and subsea equipment before revenue.

The firm's national scope sharpens the question. The Seattle thesis now informs a firm investing across the United States rather than defining a geographic limit. The central test is whether Ascend's close, first-check approach can deliver early conviction to physical-AI founders whose companies carry resource demands conventional software startups do not.