Google Restricts Meta AI
Google limits Meta's Gemini AI access as surging demand creates industry computing bottlenecks.
Google has capped how much of its Gemini AI its rival Meta can use, Financial Times reports, citing concerns that demand has outpaced available computing power. The restrictions have disrupted multiple internal AI projects at Meta and forced the company to tell staff to use fewer AI tokens.
The move highlights a deepening tension in Big Tech: AI demand is growing faster than even the largest companies can build infrastructure to meet it. Google Cloud brought in $20 billion in revenue in Q1 2026 — up 63% year-over-year — yet CEO Sundar Pichai admitted that a lack of compute power held growth back even further, according to Reuters.
Meta Turned to Its Ad Rival for AI Help
Meta has spent years building its own AI models under the open-source Llama brand. But by late 2025, internal tests showed its next-generation model, code-named "Avocado," was falling short of rivals like Gemini 2.5 and OpenAI's latest offerings, according to The Information. That gap pushed Meta into an unusual position: licensing the AI brain of its biggest advertising competitor.
Meta reportedly held talks with Google, OpenAI, and Anthropic as potential partners. Google won that business — but could not fully deliver. By March 2026, Google told Meta it could not provide the full volume of Gemini API access Meta had requested, Financial Times reported. Meta has separately committed $115 billion to $135 billion in capital spending for 2026 to build its own AI infrastructure.
Google's Compute Crunch Is Hitting Everyone, Not Just Meta
Gemini API usage more than doubled between March and August 2025 — jumping from 35 billion to 85 billion monthly requests, according to The Information. That surge strained Google's data centers. In May 2026, Google shifted from prompt-based limits to "compute-based" limits, which track how complex and token-heavy each request is, according to PCWorld.
Google VP Josh Woodward said the company is "capping the amount of quota a single prompt can use" to help users stretch their limits during this tight period. Power users on Reddit pushed back hard, with some calling Gemini Pro and Ultra tiers "barely usable" for complex tasks. Google Cloud's project backlog now stands at $462 billion, per Seeking Alpha — reflecting huge future demand the company cannot yet fulfill.
Is This a Capacity Problem or a Strategic Move?
Google frames the limits as a side effect of "tremendous momentum" — a fair-access measure for a fast-growing customer base. But some analysts see it differently. By slowing Meta's access to top AI tools, Google may be — intentionally or not — putting a brake on a company that competes directly with it for digital advertising dollars, according to Money US News.
Analysts at Arete note that Meta's willingness to rent AI from Google shows a flexible but risky strategy. Instagram ad conversion rates rose 5% thanks to AI tools, but that growth now depends on a competitor's infrastructure. Meta is working to escape that dependence by accelerating production of its own MTIA chips, though that buildout will take time.
Meta Tells Staff to Do More With Less
Rather than telling employees the situation is a setback, Meta is framing the token limits as a push for efficiency. Staff have been encouraged to conserve AI tokens and find leaner ways to run their projects, Financial Times reported. The internal model code-named "Avocado" is among the projects facing delays.
Meta CTO Andrew Bosworth has meanwhile pushed a separate internal effort — called the Model Capability Initiative — to collect better data and improve Llama models from within. The goal is to reduce how much Meta must rely on outside partners like Google. Until that infrastructure is ready, though, Meta remains subject to whatever limits Google decides to set.
Sentiment
NeutralPublishers
26
Articles
133
Reach
159
