AI21 Labs, a startup headquartered in Tel Aviv, specializing in the development of a diverse array of AI text-generation tools, recently completed a Series C funding round, amassing a substantial $155 million. Leading this funding effort were prominent investors such as Walden Catalyst, Pitango, SCB10X, b2venture, Samsung Next, and Amnon Shashua, who is both the founder of Intel-owned Mobileye and a co-founder of AI21 Labs. Notably, Google and Nvidia also joined in the investment.
This financial infusion elevates AI21 Labs’ total funding to an impressive $283 million, and it confers a valuation of $1.4 billion upon the company—a slight increase from the previously reported figure of $1.2 billion by The Information.
AI21 Labs first emerged from stealth mode in October 2020, marked by the release of their inaugural AI writing tool, Wordtune. Yoav Shoham, one of AI21 Labs’ co-CEOs, conveyed their trajectory, stating, “Since then, we’ve only managed to grow, more than doubling our workforce with plans to add even more employees.”
The inception of AI21 Labs dates back to 2017, when it was co-founded by Amnon Shashua, Yoav Shoham, and Ori Goshen, who serves as the startup’s other co-CEO. Shoham, a professor emeritus at Stanford, previously demonstrated his entrepreneurial acumen by selling two companies to Google: the time management app Timeful and the social network friend organizer Katango. Meanwhile, Goshen is a serial entrepreneur with a track record of co-founding and leading various tech companies in Israel, including the telecommunications analytics firm Crowdx.
AI21 Labs’ flagship product is AI21 Studio, a flexible pay-as-you-go developer platform designed for the creation of customized text-based business applications, utilizing AI21’s exclusive text-generating AI models, including the advanced Jurassic-2 model. The startup also offers access to Wordtune, a multilingual reading and writing AI assistant similar to Grammarly.
Customers can seamlessly integrate AI21 Labs’ platform through APIs to cater to specific generative AI needs, such as summarization, paraphrasing, and grammar and spelling correction. Significantly, the startup’s models support an expanding selection of languages, encompassing Spanish, German, Italian, and Dutch, among others.
AI21 Labs operates within the competitive landscape of generative AI, pitted against an array of both established entities and emerging startups. Giants like Google, AWS, and Microsoft offer tools analogous to AI21 Studio, while startups including Cohere, OpenAI, and Anthropic also provide robust offerings in this space, albeit with varied funding resources. Notably, OpenAI boasts a staggering $11.3 billion in cumulative funding, with Anthropic and Cohere securing $1.6 billion and $435 million, respectively.
However, Shoham asserts that AI21 Labs distinguishes itself through its solutions’ superior characteristics, notwithstanding surface-level similarities and a comparatively modest R&D budget. He highlights that their solutions are developed utilizing some of the world’s most extensive and sophisticated large language models, providing users with enhanced control. Additionally, their models are trained on current data, a notable advantage compared to text-generating models trained on outdated information, which struggle to provide accurate responses to queries about contemporary events.
Shoham, the spokesperson for AI21, emphasized the seamless integration and the production of dependable, trustworthy, and precise outcomes from AI21’s AI systems. He also highlighted the substantial challenge posed by the quality and quantity of training data in the industry, detailing the company’s dedication to utilizing current, reliable, and dependable data while continually enhancing their models through additional training data.
I can’t personally verify the accuracy of these claims since I haven’t recently evaluated AI21 Labs’ products. Nevertheless, AI21 Labs appears to be making strides despite its disadvantages.
Shoham, who did not disclose the exact size of AI21 Labs’ customer base but hinted at its inclusion of “several” Fortune 100 companies, reported that Wordtune alone boasts over 10 million users. Additionally, AI21 Labs was among the inaugural partners for Bedrock, Amazon’s generative AI app development platform.
With this latest funding injection, AI21 Labs intends to accelerate its research and development endeavors, striving to achieve its objective of advancing AI to the next level, enabling reasoning across various domains. The company also plans to forge more partnerships across the entire tech ecosystem and expand its workforce of 200 employees, with a particular focus on research and business development roles.
Furthermore, AI21 Labs is bolstering its leadership team, as evidenced by the recent appointment of Pankaj Dugar, a former executive from Google and Databricks, as its Senior Vice President of go-to-market and General Manager of North America.
The financial support is crucial, given the substantial costs associated with developing large language models. OpenAI, for instance, reportedly expended $540 million in the previous year to create ChatGPT. AI21 Labs itself estimates that training a text-generating model with 1.5 billion parameters can cost as much as $1.6 million. This figure doesn’t even account for the expenses related to hosting these models.
OpenAI was purportedly spending up to $700,000 daily to maintain the infrastructure supporting ChatGPT. A rough estimate suggests that operating a model the size of GPT-3, an older OpenAI-developed model, on a platform like AWS could cost around $87,000 annually.
While AWS is a partner of AI21 Labs, it’s unclear if volume discounts apply, but it’s safe to assume that hosting such models entails substantial overhead costs.
Shoham emphasized responsible growth, emphasizing their commitment to delivering the most current and reliable AI solutions to their customers.