Arcee AI Inc. has raised an undisclosed Series B round that lifts the open-weight model developer's valuation above $1 billion, the company said. One source told Fortune the round was worth at least $150 million. For businesses that want to run AI on their own infrastructure, the deal signals that capital is still available for alternatives to closed models.
Who backs the round
The round was led by Vista Equity Partners, Cambium Capital and Emergence Capital. A10 Ventures, Hitachi, IAG, Microsoft Corp.'s M12, P7 and the AI consultancy Wipro Ltd. also participated. The company is based in San Francisco and was founded in 2023, after co-founder and Chief Executive Mark McQuade left Hugging Face Inc., which is now being acquired by Nvidia Corp. Before building its own models, Arcee worked on post-training and fine-tuning, supplying infrastructure and tools to companies adapting existing AI models.
Arcee develops open-weight models, which publicly release their core numerical parameters so anyone can download, modify and run them on their own infrastructure. Its Trinity family was released over the past year; the most powerful, Trinity Large, has 400 billion parameters and 13 billion active for each token. The company spent about $20 million developing those models last year, covering compute infrastructure, data, operations and the salaries of developers and engineers. According to McQuade, Arcee had only about $30 million available when it decided to build its own models, and almost 70% of its total capital went into Trinity.
What this means for business
Most U. S. AI developers keep model code private and often dictate where models can run, while most Chinese developers have embraced open weights and produced powerful, accessible systems. McQuade said his goal is for Arcee to catch up with China's leading AI developers, arguing that the U. S. is far ahead in closed source but dropped the ball on open source. He added that organizations should not have to choose between frontier-model capabilities and control over the technology at the center of their work. The new money will train a next generation of Trinity models already under development and expand work with the U. S. Department of Energy and 17 national laboratories on Genesis-Science-1, an open-weight model for scientific computing and research. Arcee will also keep investing in infrastructure and products for companies that customize, test and deploy open-weight models.
For companies adopting AI, the practical effect is a wider set of vendors whose models can be hosted in-house, which matters where data cannot leave a controlled environment. A small company can download weights and run them on rented GPUs without negotiating a license; a large enterprise gains a second source alongside closed providers and more room to bargain on terms. The same choice appears in regulated work: banks, manufacturers and public-sector teams often need to keep inference inside their own perimeter, and open weights make that possible without giving up model quality.
What the round does not settle is total cost of ownership. Running a 400-billion-parameter model requires hardware, engineers and ongoing evaluation, so the license price is only one line in the budget. Buyers should ask the vendor who maintains the model after release, how updates and security patches arrive, what license governs commercial use, and whether support exists for fine-tuning on proprietary data. Monti Saroya of Vista Equity Partners said Arcee showed a rare combination of technical execution and capital efficiency, and that enterprises increasingly look for AI systems they can control, customize and deploy on their own terms.
The marker to watch is the next Trinity generation: if Arcee ships it at a cost and timeline comparable to the $20 million and one year behind the current family, the open-weight approach becomes a repeatable business model rather than a single bet. That would give corporate buyers a credible third option between closed U. S. models and Chinese open-weight systems.
