The Cologne-based translation provider DeepL is expanding its portfolio with a powerful voice-to-voice feature. Is this enough to keep Big Tech competitors at bay?
The deliberate babble of languages during DeepL’s product demonstration highlights its new voice-to-voice capability. Leonardo Doin, Head of Voice at the Cologne-based AI translator, spoke in Portuguese; CEO Jaroslaw Kutylowski conversed in Polish; attendees could follow the German translation on their smartphones, while an AI voice simultaneously spoke the English interpretation aloud. Despite minor stumbles (like “Fireside Jet” instead of “Fireside Chat”), the demonstration was largely successful, underscoring DeepL’s advanced translation technology and the promise of its new feature.
With the voice-to-voice feature, DeepL is enhancing its product offerings. According to CEO Kutylowski, the company’s core mission remains breaking down language barriers globally. Since its inception in 2017, stemming from the predecessor company Linguee, DeepL has achieved considerable success. Despite facing competition from Big Tech giants like Google and Apple, and more recently chatbot providers like OpenAI, DeepL has maintained its market position.
Since attracting major US investors in 2023 and 2024, which boosted its valuation to $2 billion, DeepL has adopted a more assertive global stance. A US IPO is considered a possibility, perhaps even this year. However, recent insider reports in “Handelsblatt” have alluded to increasing pressure and an “Americanization” of the company. Nevertheless, a tech success story, especially in this competitive market, is unlikely to unfold without growing pains.
DeepL: Its Own Language Model at the Core
Kutylowski acknowledges that competition is fierce. A few months ago, Apple introduced real-time translation capabilities for its AirPods, with reportedly mixed results. DeepL’s new feature is specifically designed for corporate use, targeting global teams in meetings or frontline workers overcoming language barriers in diverse operational environments. “It’s a competitive market,” admits the DeepL CEO, “but it always has been.” DeepL aims to differentiate itself through superior quality, high speed, and low latency.
The voice-to-voice feature has been under development for two years. The company is proud that its underlying AI language models are entirely self-developed and owned: “All the technology is proprietary,” states Kutylowski. This is a significant differentiator, as most new AI providers rely on models from Anthropic or OpenAI. Kutylowski emphasizes the importance of owning the technology to “get the most out of it” and enable the development of the next generation of products and technology.
For Kutylowski, controlling the models signifies genuine technological sovereignty, overriding concerns about whether the software utilizes US cloud providers. DeepL’s recent reliance on Amazon’s AWS subsidiary had previously drawn criticism.
AI Agents as a Lever?
The DeepL CEO prefers not to disclose specific business figures but asserts that the company is doing “great,” contrary to “Handelsblatt” reports. He claims DeepL achieves “one of the best gross margins in the AI market,” comparable to software companies rather than AI providers dependent on high computing costs. In the Software-as-a-Service market, a 75 percent gross margin is considered excellent.
Despite this success, the question lingers whether being an excellent translation provider will be sufficient for DeepL in the long term. Last autumn, the now 1000-employee company announced its entry into new markets, offering AI agents for various corporate applications. Kutylowski offers no specific updates on this front, describing it as a “very dynamic area,” suggesting the company is still searching for a strong product-market fit.
For now, DeepL appears to be refocusing on the language and translation market, a substantial challenge given its powerful competitors. Kutylowski, however, is experienced in facing down Big Tech: “As a tech CEO, you shouldn’t be afraid.” The risk of failure is always present and “should always be present,” he concludes, because otherwise, you are simply not courageous enough.
