Google denied Meta’s request for more compute capacity in March and imposed limits on its Gemini AI model usage, causing disruptions to Meta’s internal AI projects.
Google restricted Meta’s access to its Gemini AI models earlier this year after the social media company requested additional compute capacity and was turned away, according to a Financial Times report published on June 28. The development has caused disruptions and delays across Meta’s internal AI projects and highlights a deepening compute capacity crunch that is affecting some of the largest technology companies in the world.
What Happened Between Google and Meta
In March, Meta approached Google seeking more computing capacity to support its expanding internal AI workloads. Google declined, informing Meta it could not fulfill the full Gemini capacity the company had requested, and proceeded to impose usage limits on Meta’s access to its models.
Meta is reported to have been among the hardest hit of Google’s enterprise customers, though the restrictions have had a broader impact across several other businesses that rely on Google’s infrastructure for AI compute. The episode illustrates how acute demand for AI processing power has become, even among companies with the resources to pay for premium access.
Google Cloud Is Under Strain — Even by Its Own CEO’s Admission
The constraints affecting Meta are not entirely surprising given that Google’s own leadership has acknowledged capacity problems. During the company’s last earnings call, CEO Sundar Pichai stated that computing power limitations held back what would otherwise have been higher growth for Google Cloud, with the resulting backlog nearly doubling quarter on quarter. Google Cloud revenue reached $20 billion in the first quarter, a figure Pichai suggested could have been higher with more available compute.
The situation reflects a broader industry dynamic: massive capital investment in chips, data centres, and AI infrastructure is struggling to keep pace with the speed at which demand for AI services is growing.
Meta Tells Staff to Use AI More Efficiently
The access restrictions have prompted a notable shift in how Meta is approaching internal AI usage. Earlier this year, the company was among several big tech firms that actively encouraged employees to use AI tools as extensively as possible — a trend widely referred to as tokenmaxxing. Meta went as far as tying employee performance evaluations to their usage of AI tools.
That approach has now been reversed. Following the Gemini capacity restrictions, Meta has encouraged staff to be more efficient with AI tokens, the units used to measure how much AI is being consumed across workflows. The shift signals a more measured approach to AI adoption inside one of the industry’s most enthusiastic advocates for AI-first operations.
Microsoft Also Pulls Back on Claude Code
Meta is not the only major technology company reassessing its AI tool usage. Microsoft’s Experiences + Devices division — the unit responsible for Windows, Microsoft 365, Outlook, Microsoft Teams, and Surface hardware — was instructed to wind down its use of Anthropic’s Claude Code by the end of June, according to a report by The Verge.
While part of the decision is intended to redirect developers toward Microsoft’s own Copilot CLI tool, financial considerations also played a role. Together, the moves at Meta and Microsoft suggest that even well-resourced technology companies are beginning to treat AI tool usage as a cost to be managed rather than a resource to be maximised without limit.
Apple’s Gemini Partnership Adds Another Layer of Pressure
Adding further context to Google’s capacity constraints, Apple has also partnered with Google to use Gemini models as part of the next generation of its Siri AI voice assistant. The partnership, announced in January 2026, means Google is simultaneously serving multiple major technology companies with significant AI compute needs.
Notably, the Gemini models used by Apple are distinct from the publicly available versions, a detail that was not initially clear when the partnership was announced. With Apple, Meta, and a range of other enterprise customers all drawing on Google’s AI infrastructure, the pressure on available compute resources becomes easier to understand.
What This Signals for the AI Industry
The Google-Meta capacity dispute is a useful reminder that access to AI infrastructure is not unlimited, regardless of budget. As frontier AI models grow in size and complexity, and as companies integrate AI more deeply into their operations, the physical constraints of data centre capacity, chip supply, and energy availability are becoming meaningful limiting factors.
For companies that have built internal processes around continuous AI access, supply-side restrictions introduce a new kind of operational risk — one that financial investment alone cannot immediately resolve. The coming months will likely see more companies diversifying their AI infrastructure partnerships rather than depending heavily on a single provider.
Frequently Asked Questions
Google was unable to meet Meta's request for additional Gemini compute capacity in March due to surging demand across its enterprise customer base. Google CEO Sundar Pichai acknowledged in an earnings call that compute constraints were limiting Google Cloud's growth and causing backlogs to nearly double quarter on quarter.
Tokenmaxxing was a trend where companies encouraged employees to use AI tools as heavily as possible, sometimes linking usage to performance reviews. After Google restricted its Gemini access, Meta reversed course and began encouraging staff to use AI tokens more efficiently rather than maximising consumption.
Apple partnered with Google in January 2026 to use Gemini models to power a more personalised version of Siri. With Apple, Meta, and numerous other enterprise clients all drawing on Google's AI infrastructure simultaneously, the available compute capacity has come under significant strain, contributing to the restrictions placed on Meta.
