The proposed transaction, announced from Roubaix on 11 June 2026, remains subject to the completion of talks and customary closing steps. Financial terms were not disclosed. The deal would bring Gladia’s voice AI engineers and platform into OVH Groupe, adding speech-to-text building blocks to a portfolio that already includes cloud infrastructure, AI tools and, after the March acquisition of Dragon LLM, specialised generative models for regulated customers.
Founded in Paris in 2022, Gladia provides an application programming interface that can transcribe conversations in real time and in batch across more than 100 languages. The company says its tools are used by more than 300,000 developers and 2,000 enterprise customers, including HeyGen, Livestorm, Attention, Circleback, Method Financial, Recall. ai and Leexi. Its platform is marketed for meeting assistants, contact centres, sales enablement, media subtitling and voice agents, with features such as speaker detection, entity capture, sentiment analysis, summaries and redaction of personal information.
For OVHcloud, the rationale is strategic rather than merely additive. Speech-to-text has become a core layer in voice agents because spoken interaction typically relies on a chain of systems: audio capture, transcription, large language model reasoning and text-to-speech response. Control over the first stage can affect latency, accuracy, privacy and downstream reliability. That makes voice AI infrastructure more valuable as enterprises move beyond chatbots into calls, meetings, clinical dictation and customer service workflows.
The talks also fit OVHcloud’s broader attempt to differentiate itself from US hyperscalers by emphasising data sovereignty, European infrastructure and control of its technology stack. The group says it operates more than 500,000 servers across 46 data centres on four continents and serves 1.6 million customers in more than 140 countries. Its public positioning rests on an integrated model that covers server design, data centre operations and network orchestration, allowing it to pitch predictable pricing and jurisdictional control to organisations handling sensitive data.
AI has moved higher on that agenda. In March, OVHcloud bought Dragon LLM, a Paris-based specialist in model fine-tuning and vertical generative AI, and launched an AI Lab to develop sovereign large language model services. In April, it said the lab would target demand for “verticalised agentic AI” and highlighted defence, mid-market corporate contracts and improved GPU economics as pillars of its transformation plan. The Gladia talks would add voice as a key modality alongside text and model customisation.
The commercial backdrop is mixed. OVHcloud reported revenue of €555.3 million for the first half of fiscal 2026, up 5.5 per cent on a like-for-like basis, with adjusted EBITDA of €227.2 million and a 40.9 per cent margin. Public Cloud remained the main growth driver, while the group maintained full-year targets of 5 to 7 per cent organic growth, higher adjusted EBITDA margin and positive levered free cash flow.
Voice AI is one such higher-value segment. Demand has been lifted by customer-service agents, meeting transcription tools and multilingual business software. San Francisco-based Deepgram raised $130 million at a $1.3 billion valuation in January to expand voice AI models and pursue acquisitions, while players such as AssemblyAI, Speechmatics, ElevenLabs, Google, Microsoft and Amazon compete across transcription, speech synthesis and conversational systems. Gladia’s appeal lies in its multilingual pitch and European data-residency positioning.
Topics
Technology