Google Limits Meta's Gemini Access Due to Global AI Compute Shortage
View original source →Google has been rationing how much access to its Gemini AI models it provides to Meta, one of its largest enterprise AI customers, because it does not have enough computing capacity to meet full demand. The development, confirmed by CNBC and Forbes, has forced Meta to reduce AI usage across its organization.
Google notified Meta in approximately March 2026 that it could not fulfill the full Gemini capacity Meta had contracted to purchase. The shortfall disrupted several of Meta's internal AI projects, including automated safety processes and content moderation pipelines. Meta instructed staff to use AI tokens more efficiently and is accelerating its shift toward Muse Spark, its internal AI model.
Key points:
• Google is paying SpaceX $920 million per month for access to approximately 110,000 Nvidia GPUs housed in xAI's data centers — capacity described internally as a 'bridge' while building owned infrastructure
• Anthropic has a separate arrangement at $1.25 billion per month
• DRAM prices are projected to surge 40-50% in Q3 2026, driven by massive memory requirements of long-context AI models and agent-based systems
• Several other Google enterprise clients have experienced similar capacity limitations
• Meta's response — accelerating adoption of its own internal Muse Spark model — previews the enterprise AI market's next competitive dynamic
Why It Matters: Compute scarcity is now a first-order strategic variable in enterprise AI planning. If Google cannot meet demand from a single client like Meta, smaller enterprises face meaningful risk of capacity throttling during peak demand. Organizations should negotiate capacity guarantees into enterprise AI contracts rather than relying on availability-based pricing.