---
title: "Arcee AI Crosses $1B: Open-Weight Models Built for ~$20M"
description: "Arcee’s Series B, led by Vista Equity, Cambium, and Emergence, values the open-weight lab above $1 billion. Fortune sources put the raise at least at $150M. The company says its 2025 Trinity lineup, including a 400B MoE model, cost about $20M to build."
date: 2026-09-16T00:00:00.000Z
source: https://venturecapitaltracker.com/2026-arcee-ai-150m-series-b-1b-vista-open-weight
---

# Arcee AI Crosses $1B: Open-Weight Models Built for ~$20M

> Arcee’s Series B, led by Vista Equity, Cambium, and Emergence, values the open-weight lab above $1 billion. Fortune sources put the raise at least at $150M. The company says its 2025 Trinity lineup, including a 400B MoE model, cost about $20M to build.

**[Arcee AI](/startup/arcee-ai)** announced a **Series B** on **September 16, 2026** that values the San Francisco open-weight lab at **more than $1 billion**. The company **did not disclose** the round size. **Fortune** reported a **$1 billion pre-money** mark and, citing a source familiar with the matter, said the raise was **at least $150 million**. **Vista Equity Partners**, **Cambium Capital**, and **Emergence Capital** co-led; **AI10 Ventures**, **Hitachi**, **IAG**, Microsoft’s **M12**, **P7**, and **Wipro** joined ([company blog](https://www.arcee.ai/blog/arcee-ai-raises-series-b-to-build-the-future-of-american-open-models), [Fortune](https://fortune.com/2026/09/16/arcee-ai-trained-four-models-for-20-million-now-its-worth-1-billion/)).

Meta stopped releasing Llama models in early 2025. Arcee bet most of a ~$30M bankroll on building open weights from scratch — and claims the entire 2025 Trinity lineup, including a **400B MoE** model, cost about **$20 million**.

## Key facts

| Field | Detail |
| --- | --- |
| Company | [Arcee AI](/startup/arcee-ai) (San Francisco) |
| Round | Series B; amount **company-undisclosed**; Fortune **≥$150M** |
| Valuation | **>$1B** (company); **$1B pre-money** (Fortune) |
| Leads | Vista Equity, Cambium Capital, Emergence Capital |
| Efficiency claim | ~**$20M** for 2025 model lineup incl. Trinity Large (company / Fortune) |
| Prior pivot | Founded 2023 on post-training; full pretrain push after Meta’s Llama pause |

## Who uses the product — and for what job

**Users:** enterprises, developers, researchers, and public institutions that want frontier models they can **run, adapt, and own** without sending proprietary data to a closed API.

**Job:** fill the U.S. open-weight gap after Meta exited Llama releases — permissively licensed models plus the “boring” stack (customize, evaluate, deploy, monitor).

Trinity Large is positioned as a U.S.-developed, permissively licensed frontier MoE model. Genesis-Science-1 with the **Department of Energy** is the institutional proof path.

## Why now

- Closed labs dominate U.S. AI; open weights have been China-led (DeepSeek, Z.ai GLM). McQuade told Fortune the goal is to **catch China**, not chase Poolside.
- Hugging Face’s agreed Nvidia acquisition ([memo](/2026-nvidia-hugging-face-12-9b-acquisition)) tightens the open-stack map.
- Enterprises want on-prem / VPC control after agent workloads move sensitive data through models.

## Why Vista / Cambium / Emergence — portfolio fit

Vista is a **software PE / growth** franchise that can push Arcee into portfolio IT stacks — Fortune notes Arcee will work extensively with Vista companies. Cambium and Emergence add deep-tech and enterprise SaaS pattern recognition.

**Strategic rationale:** raise from capital that buys **enterprise distribution** and long-duration ownership, not only another model-lab brand check. We do **not** list Vista, Cambium, or Emergence as `/fund/` entities — names are press-sourced.

| Dimension | Fit |
| --- | --- |
| Stage | Series B / unicorn mark after product proof |
| Thesis | Open weights as controllable enterprise infrastructure |
| Risk | Model quality vs closed frontier; capital efficiency must hold at larger scale |

## What remains undisclosed

- Exact **raise amount** (company silent; Fortune floor only)
- Post-money precision beyond “more than $1B”
- ARR, customer count, and named commercial logos
- How much of the round is primary vs secondary

The unresolved question: whether a lab that trained four models for ~$20M can stay capital-efficient once it races Chinese open labs at full Series B burn.

**By:** [Venture Capital Tracker](https://venturecapitaltracker.com/editorial-policy)

**Editorial note:** AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.