AI21 Labs Secures $155M Funding, Achieving a $1.4B Valuation in Generative AI Sector

AI21 Labs, a Tel Aviv-based startup specializing in text-generating AI tools, has successfully secured $155 million in a Series C funding round. The funding was led by Walden Catalyst, Pitango, SCB10X, b2venture, and Samsung Next, with participation from co-founder Amnon Shashua, also known for Mobileye, and support from tech giants Google and Nvidia. This round brings AI21 Labs’ total funding to $283 million and values the company at $1.4 billion, surpassing an earlier report by The Information which cited a valuation of $1.2 billion.

“AI21 Labs emerged from stealth mode in October 2020 with the launch of our first AI writing tool, Wordtune,” said Yoav Shoham, co-CEO of AI21 Labs. “Since our inception, we have experienced significant growth, more than doubling our workforce, and we plan to expand even further.”

Founded in 2017 by Shashua, Shoham, and fellow co-CEO Ori Goshen, AI21 Labs boasts a leadership team with deep expertise in technology and entrepreneurship. Shoham, a Stanford professor emeritus, previously sold two companies to Google, while Goshen has launched several tech ventures in Israel.

AI21 Labs’ flagship offering is AI21 Studio, a pay-as-you-go platform that enables developers to create customized text-based business applications using AI21’s advanced text-generating models, including the robust Jurassic-2 model. The company also markets Wordtune, a multilingual AI assistant designed to enhance reading and writing, similar to services like Grammarly.

Through APIs, businesses can leverage AI21 Labs' platform for various generative AI applications such as summarization, paraphrasing, and grammar correction, supporting multiple languages including Spanish, German, Italian, and Dutch. AI21 Labs faces competition from numerous startups and established players in the rapidly evolving generative AI market.

While Google, AWS, and Microsoft provide tools similar to AI21 Studio, startups like Cohere, OpenAI, and Anthropic also compete in this space, often with significantly larger funding rounds—OpenAI alone has raised $11.3 billion. Despite this competitive landscape, Shoham asserts that AI21 Labs’ offerings stand out due to their development on some of the largest and most sophisticated large language models, providing users with "more refined control" compared to many generative AI solutions that may lack up-to-date training data.

“The AI21 systems are designed for easy integration, yielding reliable and accurate results,” Shoham emphasized. “The key challenge in this industry is the quality and volume of training data. We invest significantly in using current data and continuously train our models to ensure their reliability.”

While I haven’t personally assessed AI21 Labs' products recently, the startup appears to be gaining momentum. Shoham notes that Wordtune alone has attracted over 10 million users, and AI21 Labs has partnered with Amazon as a launch partner for Bedrock, its generative AI app development platform.

The new funding will enable AI21 Labs to enhance its R&D initiatives and strive to develop AI solutions capable of reasoning across diverse domains. The company plans to broaden its partnerships within the tech ecosystem and to grow its workforce from 200, emphasizing roles in research and business development. AI21 Labs is also strengthening its executive team with the recent addition of Pankaj Dugar, a former executive from Google and Databricks, who will serve as the SVP of go-to-market and GM of North America.

With the capital-intensive nature of developing large language models, securing funding is crucial. Reports indicate that OpenAI spent $540 million last year on ChatGPT development, while training a 1.5 billion parameter text-generating model could cost up to $1.6 million. For perspective, AI21 Labs’ prior model, Jurassic-1, contained 178 billion parameters.

On top of development costs, hosting expenses for these models can be significant. At one point, OpenAI was reportedly paying as much as $700,000 daily for ChatGPT's infrastructure. The estimated annual cost to run a model similar to GPT-3 on services like AWS is around $87,000.

Since AWS partners with AI21 Labs, it remains uncertain whether any cost savings apply, but operational costs are unlikely to be trivial. “We are committed to responsible growth, allowing us to continually enhance our world-class AI solutions for our customers,” said Shoham.

Most people like

Find AI tools in YBX