Groq currently operates data centers across the U.S., Canada, Europe, and the Middle East, with new sites planned in Asia
In sum – what to know:
Post-2025 expansion – Groq plans to exceed its 2025 buildout of 12 data centers, with new sites coming online across Asia and other regions next year.
Surging AI demand – Chairman Jonathan Ross said demand continues to exceed capacity, with particular traction in India due to faster, lower-power Groq chips.
Colocation growth – Groq relies on providers such as Equinix, DataBank, and Bell Canada to scale its footprint efficiently across multiple continents.
AI chip startup Groq is to establish more than a dozen new data centers next year, after building 12 facilities in 2025, according to the firm’s chairman Jonathan Ross. The executive told The Wall Street Journal the company’s infrastructure footprint is set to grow at a fast pace, as demand for its AI inference chips continues to outpace projections. “Every time we over-compensate by trying to build more, people surprise us by needing even more than we built,” he said.
The U.S. firm specializes in delivering infrastructure for AI inference — the process of running trained AI models — using its proprietary Language Processing Unit (LPU) hardware. The company currently operates data centers across the U.S., Canada, Europe, and the Middle East, with new sites planned in Asia. Ross cited particularly strong growth in India, where the company’s chips have seen high adoption.
The AI company relies on colocation partners such as Equinix, DataBank, and Bell Canada to support its distributed footprint. In July, Groq had expanded its data center operations to Europe with the opening of a facility in Helsinki, Finland. The company said that the new data center, developed in collaboration with Equinix, aims to serve growing demand for AI inference capacity across the region by bringing computing resources closer to European customers.
In August, Saudi Arabian AI venture Humain, backed by Saudi Arabia’s Public Investment Fund (PIF) and Groq, announced the deployment of OpenAI’s new open-source models – gpt-oss-120B and gpt-oss-20B – in Humain’s sovereign data centers in Saudi Arabia. Humain noted that the models run on Groq’s high-speed inference platform and are fully hosted within the country.
The announcement builds on the strategic partnership between Humain and Groq established in May 2025, aimed at expanding AI infrastructure and services in the region.