Highlights:

  • Together AI’s funding round was led by Salesforce Ventures, with participation from other notable investors such as Coatue Management, Lux Capital, and Emergence Capital.
  • Together AI asserts that its pricing structure is considerably lower than that of leading cloud infrastructure providers like Amazon Web Services Inc., Google LLC, and Microsoft Corp.

Artificial intelligence startup Together Computer Inc., commonly called Together AI, recently announced that it has secured funding worth USD 106 million, doubling its valuation since its previous raise last fall.

Together AI’s funding round was spearheaded by Salesforce Ventures, with participation from other prominent investors including Coatue Management, Lux Capital, and Emergence Capital. Consequently, the company’s valuation has soared to a staggering USD 1.25 billion.

Less than six months have passed since Together AI concluded its USD 102.5 million Series A funding round, spearheaded by Kleiner Perkins and Nvidia Corp.

Nvidia’s involvement is pivotal to the success of Together AI. Founded in 2022, the startup offers cloud-based access to Nvidia’s graphics processing units, which are highly valued for their capability to fuel advanced generative AI models like ChatGPT.

The inception of Together AI dates back to June 2022, when its Chief Executive Officer, Vipul Ved Prakash, co-founded the venture alongside fellow pioneers Ce Zhang, Percy Liang, and Chris Re. Initially, it started as a straightforward, open-source, cloud-based infrastructure platform tailored to AI workloads.

However, its evolution has led to the developing of proprietary AI development tools, models, and services to assist organizations in seamlessly integrating AI into their business applications. Employing its cloud platform, clients can train, refine, and ultimately deploy AI models into production.

Notably, the startup asserts that its pricing structure is considerably lower than that of leading cloud infrastructure providers like Amazon Web Services Inc., Google LLC, and Microsoft Corp.

Previously, Prakash highlighted that the company achieves notable cost reductions for large language models by meticulously optimizing every facet of its infrastructure stack. In addition to amassing thousands of Nvidia GPUs, Together AI offers virtualization software, scheduling, and model optimizations that collectively contribute to reducing the costs associated with AI workloads.

Together AI’s infrastructure encompasses a network of data centers across the U.S. and Europe, boasting a cumulative compute power of over 20 exaflops. The GPU clusters come in a range of sizes, from 16 modules to a maximum of 2,048 at the upper limit.

Prakash stated that customers can leverage the startup’s infrastructure and development tools to swiftly integrate popular open-source LLMs like Meta Platforms Inc.’s Llama 2 into their business applications. Customers also have the option to bring their models and utilize its platform for training and fine-tuning them. Furthermore, the startup provides consulting services to assist enterprises in designing and constructing tailored LLMs for particular workloads, alleviating the requirement for them to possess extensive AI expertise.

Together, AI also plays a significant role in open-source AI research. It spearheads a project known as RedPajama, dedicated to crafting open-source, chat-based generative AI models intended to compete with OpenAI’s ChatGPT. Additionally, it has developed a tailored version of Llama 2.

Rob Keith, Managing Director at Salesforce Ventures, reported that many enterprises are initially drawn to Together AI due to its GPU infrastructure. However, although many initially came for the GPUs, he said many would “stay for the workflow services and research.”

Prakash mentioned that the funding round secured by Together AI will enable the company to lease additional computing power from cloud infrastructure providers and recruit more researchers to assist in designing more powerful open-source AI models. He further updated on Together AI’s progress, revealing that the company has now amassed over 45,000 registered users.

The CEO said, “We are really becoming one of the default places if you want to use open models or customized models.”