Google caps Meta's Gemini AI usage over compute limits

Google has imposed restrictions on Meta's access to its Gemini AI models due to an overwhelming demand for computing resources that surpasses what Google can provide. A report from the Financial Times indicates that Meta sought to acquire additional access to Google’s AI capabilities to bolster its own AI initiatives. However, in March, Google announced it was unable to offer the full computing resources required. This limitation has hindered several of Meta's internal AI projects, causing delays in their development timelines. Concurrently, Google's own clients have also felt the impact of the restricted capacity. In response, Meta has instructed its staff to optimize their use of AI tokens to avoid excessive consumption. The situation underscores a significant challenge within the tech industry: an escalating shortage of AI computing power. The demand for GPUs and large language models continues to surge, outstripping available supply. Major technology firms are investing heavily—billions of dollars—in AI chips and data centers to enhance their computing capabilities. Yet, despite these substantial investments, they still find themselves struggling to meet the increasing need for AI services. In the first quarter, Google Cloud reported a revenue of $20 billion, reflecting robust business growth. However, CEO Sundar Pichai noted that the cloud division could have experienced even greater growth if more computing capacity had been available. "We are compute-constrained in the near term," Pichai remarked, adding that Cloud revenue would have been higher if demand could have been fully met. This scenario may open doors for other cloud service providers, such as AWS, Azure, and Oracle, to capitalize on the high demand for AI computing. As businesses increasingly deploy chatbots, coding assistants, and AI-driven agents, they are also grappling with rising costs associated with AI usage. Tech giants like Microsoft have begun to impose limits on their AI services in response to these escalating compute expenses.