Google has limited access to Meta’s Gemini AI model after a surge in demand for large‑language models strained its data‑center capacity. The restriction applies to Google Cloud customers who previously used Gemini for generative‑AI workloads, forcing them to switch to alternative providers or older models. Analysts say the move highlights how compute power has become the sector’s bottleneck in the cloud.

The throttling comes as AI‑driven services consume an estimated 30% more GPU cycles than traditional cloud workloads, prompting providers to prioritize high‑margin clients. Google, which runs the world’s largest AI‑focused infrastructure, reportedly allocated extra chips to its own products, leaving less headroom for external partners. Meta’s Gemini, launched to compete with OpenAI’s GPT‑4, now faces reduced exposure on a key platform.

Investors are watching the capacity squeeze because it could pressure pricing for premium AI compute and reshape cloud‑service revenue streams. Companies that secure dedicated GPU allocations may command higher margins, while those forced onto secondary tiers could see slower growth. The episode underscores that control over hardware, not just algorithms, now drives competitive advantage in the AI market.