Amid growing concerns over the high costs associated with proprietary artificial intelligence models, a U.S.-based start-up, Together AI, is capitalizing on the expanding demand for open-source alternatives that rival the performance of leading technology from established firms. The company announced an $800 million funding round on Wednesday, valuing it at $8.3 billion and bringing its total capital raised to $1.3 billion. The round was led by Prosperity7 Ventures, the investment arm of Saudi Aramco, with additional support from Nvidia, Vista Equity Partners, General Catalyst, and other investors.

Founded in 2022 by Vipul Ved Prakash, a former Apple executive, and four academics, Together AI specializes in developing and deploying open-source AI models. The company, based in San Francisco, leverages advanced innovations, including FlashAttention, an algorithm created by its chief scientist, Tri Dao, to accelerate model processing speeds. Together AI helps businesses access and customize hundreds of open-source AI models by providing hardware rental and optimized software that improve efficiency during model inference, reducing the need for companies to maintain large engineering teams.

Demand for Together AI’s services has surged sharply over the past year, with a reported 10,000-fold increase in tokens processed monthly, and an annualized revenue run rate reaching $1.2 billion last quarter. This growth aligns with a broader industry trend toward open-source AI options, many originating from China. Chinese start-ups have introduced models comparable to those developed by prominent U.S. firms, at a fraction of the cost. According to a JPMorgan analysis, Chinese models can be roughly one-fiftieth as expensive to operate per token, fueling widespread adoption. Additionally, open-source processing on AI marketplaces like OpenRouter increased from 34 percent in January to 65 percent in June of this year, according to a Citi report.

Despite these cost advantages, some experts express caution regarding reliance on Chinese-developed models, citing concerns about potential ties between these companies and the Chinese government, as well as allegations of unauthorized use of American technologies in their development.

Together AI’s ability to provide access to these affordable open-source models makes it an attractive option for businesses seeking to reduce operational costs. Ashwin Sreenivas, co-founder of AI start-up Decagon, noted that integrating Together AI into their workflow has resulted in substantial savings — he estimates that tasks shifted from proprietary models to Together AI’s open-source equivalents cost between one-fifth and one-seventh as much.

With major U.S.-based AI companies like OpenAI and Anthropic preparing for significant public offerings, some businesses are considering paths that do not rely exclusively on these firms’ technology. According to Vipul Ved Prakash, companies are looking ahead to broader choices in the AI market as OpenAI reportedly contemplates delaying its initial public offering.

Together AI plans to allocate much of its recent funding toward research and development, as well as expanding its computational infrastructure. Prosperity7 managing director Abhishek Shukla highlighted the strategic role of investors in facilitating relationships that will support scaling hardware capacity to meet growing demand.