DeepL, the AI-powered language translator, raises over $100M at a $1B+ valuation

Investors are showing a lot of interest in artificial intelligence firms, and (thanks to GPT and OpenAI) particularly ones that let people communicate with one another. The most recent of these is now announcing a significant amount of funding. DeepL, a startup that competes with Google, Bing, and other internet tools by offering immediate translation-as-a-service to both businesses and consumers, has announced a financing at a €1 billion value (about $1 billion at today's exchange rates). headquartered in Cologne, Germany Jaroslaw Kutylowski, the CEO and founder of DeepL, stated in an interview that the company doesn't want to focus on the total amount it has raised; yet, as we were researching this story, we heard a variety of numbers. One investor who was approached about the investment disclosed to TechCrunch that DeepL wanted to seek $125 million. On the other hand, a story from November that contained a rumor about the funding claimed that it was in the neighborhood of $100 million. This month's funding came to an end early. The startup is also withholding more financial information, but according to the investment source, DeepL's yearly run rate, which was $50 million at the end of last year, was multiplied by 20 to arrive at a $1 billion valuation. This is a somewhat bullish multiple in the current fundraising environment, but it reflects the company's growth, which the investor highlighted is currently at 100%, and the fact that DeepL is close to breaking even and turning a profit. The list of investors is more conclusive: According to DeepL, IVP, a new backer, is heading the round with participation from Bessemer Venture Partners, Atomico, and WiL. Benchmark and btov have also contributed to the company's past investors. Since it has mostly worked with smaller and medium-sized firms up until now, DeepL typically offers translation as a service to businesses rather than to people. Some of those have a great deal of scale potential: For instance, Mastodon's translation is powered by DeepL. The startup intends to continue down this path, but it also intends to use the funding to broaden its reach to include larger businesses and to develop new services, such as a monolingual (same-language) writing enhancer in the vein of Grammarly that is currently in closed beta and will soon be made available to the public. Additionally, it will keep funding R&D. The company's model was initially trained using a database of more than a billion translations and queries, as well as a technique for double-checking translations by looking for related online snippets. Then, this is kept on a supercomputer that not only offers translations but also keeps learning and developing. Currently, according to Kutylowski, between 60 and 70 percent of the company's employees are engineers, and they are working to develop new technology on a variety of timescales, from short-term projects with a commercial focus to medium- and long-term "moonshots" and advancements in language modeling for its own sake.