Google has scaled back Meta’s access to its Gemini artificial intelligence models, citing an imbalance between demand and available computing resources. This move, reportedly in place since March, has hindered Google’s ability to fulfill Meta’s full request for AI processing power, according to a report by the Financial Times. The reduction in capacity has disrupted some of Meta’s internal AI initiatives, reflecting broader industry challenges in meeting soaring demand for AI services.
Meta ranks among Google’s top clients for AI infrastructure, but its unusually high consumption of Gemini model capacity made it particularly susceptible to these restrictions. In response, Meta has advised its staff to adopt more efficient use of AI tokens, the metric that governs access to generative AI resources, aiming to curb excessive consumption amid the shortage. This struggle for adequate computing power is not unique to Meta; other Google customers have also encountered limits, though none as significantly affected.
The tension highlights an ongoing bottleneck confronting major tech companies as they expand AI capabilities. Despite heavy investments in data centers and advanced semiconductor technologies, industry leaders confront a persistent shortfall in AI processing capacity. Alphabet’s recent quarterly earnings revealed that Google Cloud revenue climbed to $20 billion, yet Chief Executive Sundar Pichai acknowledged that capacity constraints have capped growth and contributed to a backlog in cloud services.
The current limitations on Meta’s Gemini AI access underscore the growing pressure on cloud infrastructure resources amid the generative AI race. Companies are accelerating investments to meet future needs, but the gap between demand and supply remains a critical challenge stalling innovation and operational scaling within the AI sector.