Google Denies Meta’s Request for Computing Power
Google has restricted Meta’s access to its Gemini artificial intelligence models after Meta requested processing capacity that exceeded Google’s available infrastructure. According to a report by the Financial Times, Google informed Meta in March that it could not fulfill the requested volume of compute resources. The limitation has reportedly delayed some of Meta’s internal AI projects.
The Physics of a Silicon Shortage
The decision to limit access stems from a broader industry-wide shortage of computing power. Despite massive capital expenditures on data centers, specialized chips, and energy infrastructure, tech companies are struggling to keep pace with the surge in demand for AI processing.
While multiple clients have faced similar capacity constraints, the Financial Times reported that Meta’s request was particularly large, making the impact on their operations more visible. In response to these shortages, Meta has encouraged its employees to use AI resources more efficiently to manage costs.
Meta Pivots to Internal Models
Meta has been using Google’s Gemini models to support internal workflows, including coding assistance, advertising tools, customer service, and content moderation. To mitigate reliance on third-party providers, Meta has shifted some of these workloads to its own internal models, such as “Muse Spark.”
This pivot coincides with Meta’s long-term infrastructure spending. Mark Zuckerberg has committed to heavy investment in AI hardware, with the company promising to invest up to $600 billion in the U.S. by 2028 to expand its data center capacity.
Google’s Cloud Backlog
The competition for compute power is intensifying across the sector. Google itself is actively expanding its capacity to meet internal and customer demand. Earlier this month, Google reached an agreement to lease computing capacity from SpaceX, a deal reportedly valued at approximately $920 million per month.
During Google’s first-quarter earnings call in April, Sundar Pichai noted that Google Cloud revenue surpassed $20 billion for the first time. Pichai acknowledged that compute capacity is still insufficient in the short term, stating that cloud revenue would have been higher if the company had been able to meet customer requests. He further highlighted that Google’s total cloud backlog—contracts signed but not yet fulfilled—reached over $460 billion, nearly double the figure from the previous quarter.
Operational Realities
- Capacity Limits: Google restricted Meta’s access to Gemini in March after Meta requested processing power beyond Google’s current availability.
- Operational Impact: The restriction caused delays in some of Meta’s internal AI projects, prompting the company to shift tasks toward its own models like Muse Spark.
- Industry Shortage: The incident underscores a systemic lack of compute capacity, which continues to affect major tech firms despite investments in data centers and power infrastructure.
- Google’s Growth: Google reported cloud revenue of over $20 billion in Q1, with a backlog of unfulfilled cloud contracts of over $460 billion.