AI21 Labs, a Tel Aviv-based startup developing a range of text-generating AI tools, raised $155 million in a Series C funding round led by Walden Catalyst, Pitango, SCB10X, b2venture, Samsung Next and Amnon Shashua, the founder of Intel. Mobileye, owned by Mobileye, and co-founder of AI21 Labs. Google and Nvidia also participated.
The tranche, which brings AI21 Labs’ total raised to $283 million, values the company at $1.4 billion, slightly higher than the figure reported earlier today by The Information (1.2 billion dollars).
“AI21 Labs came out stealth in October 2020 with our first launch of [AI writing tool] Wordtune,” Yoav Shoham, one of the co-CEOs of AI21 Labs, told TechCrunch via email. “Since then, we have only grown, more than doubling our headcount and planning to add even more employees. »
AI21 Labs was founded in 2017 by Shasuha, Shoham and Ori Goshen, the startup’s other co-CEOs. Shoham, a Stanford professor emeritus, had previously sold two companies to Google, time-tracking app Timeful and social media organizer Katango. Goshen is also a serial entrepreneur, having co-founded and led several Israel-based tech companies, including telecommunications analytics firm Crowdx.
AI21 Labs’ flagship product is AI21 Studio, a development platform for building custom text-based business applications from AI21’s proprietary text-generating AI models. The startup is also selling access to the aforementioned Wordtune, a multilingual AI reading and writing assistant.
Customers can leverage AI21 Labs’ platform through APIs for specific generative AI use cases, such as summarizing, paraphrasing, and grammar and spelling correction. The startup’s templates support an increasing number of languages, including Spanish, German, Italian, and Dutch.
AI21 Labs competes with a growing list of startups and incumbents in the hot field of generative AI.
Google, AWS, and Microsoft offer tools comparable to AI21 Studio, as do startups like Cohere, OpenAI, and Anthropic (and to a lesser extent marketing-focused vendors like Jasper, Regie, and Typeface). And AI21 Labs is at a disadvantage when it comes to funding; OpenAI has raised $11.3 billion to date, while Anthropic and Cohere have raised $1.6 billion and $435 million, respectively.
But Shoham says AI21 Labs’ solutions are superior in several ways, even if they appear similar on the surface and don’t benefit from a higher R&D budget.
For one thing, says Shoham, they’re developed on “some of the largest and most sophisticated large language models in the world” and offer “more refined control” than many generative AI apps on the market. Moreover, they are trained on up-to-date data, says Shoham, unlike text generation models trained on older data, which cannot accurately answer questions about current events.
“AI21’s AI systems are easy to integrate and generate reliable, trustworthy and accurate results,” Shoham said. “The quality and quantity of training data is a major challenge for the industry. We have worked hard to ensure that we use up-to-date data, constantly training our models on additional data and ensuring that the data used is reliable and trustworthy.
Having not tested AI21 Labs products recently, I cannot speak to the veracity of these claims. But overdone or not, AI21 Labs seems to be gaining traction.
Shoham — who wouldn’t reveal the size of AI21 Labs’ customer base, except that it includes “several” Fortune 100 companies — says Wordtune alone has more than 10 million users. And AI21 Labs was a launch partner for Amazon’s Bedrock generative AI application development platform.
Shoham says the latest round of funding will allow AI21 Labs to accelerate its R&D efforts and “achieve its goal of developing the next level of AI,” with “reasoning capabilities across many domains.” AI21 Labs also plans to pursue more partnerships with companies “across the tech ecosystem” and grow its workforce by 200 people, with a primary focus on roles in research and business development.
This will certainly require money, given the capitalistic nature of developing large language models.
According to The Information, OpenAI spent $540 million last year to develop ChatGPT. And AI21 Labs’ own research estimates the expense of developing a text generation model with 1.5 billion parameters (i.e. the variables the model uses to generate and parse text) at $1.6 million. For reference, Jurassic-2’s predecessor, Jurassic-1, contained 178 billion parameters.
This does not take into account the hosting costs to serve the models.
OpenAI reportedly at one point paid up to $700,000 a day to keep the infrastructure hosting ChatGPT operational. Simple math puts the cost of running a model the size of GPT-3 and an older model developed by OpenAI at $87,000 per year on a service like AWS.
AWS counts AI21 Labs as a partner, so it’s unclear if there’s some sort of volume discount in play. But I’d bet it’s not a trivial surcharge.
“We have been focused on responsible growth, which allows us to continue to grow and expand, providing our customers with the most up-to-date and trusted AI possible,” Shoham said.