Google has reportedly limited Meta’s access to its Gemini AI models after the Facebook parent company asked for more computing capacity than Google could provide, according to a Financial Times report cited by Reuters.
The restriction has reportedly disrupted and delayed some of Meta’s internal AI projects, showing that even the world’s biggest technology companies are struggling to secure enough compute power for their artificial intelligence ambitions.
The report comes at a time when Google, Meta, Microsoft, Amazon and other major technology firms are spending heavily on chips, data centres and cloud infrastructure. But the latest development suggests that money alone is not solving the AI capacity problem quickly enough.
According to the report, Meta had sought to purchase more access to Google’s Gemini AI models. Google reportedly told Meta around March that it could not provide the full capacity Meta wanted.
This is important because Meta is not a small AI customer. It is one of the largest technology companies in the world and is spending heavily to build its own AI models, assistants and infrastructure. If a company of Meta’s scale can still face compute limits, it shows how tight the AI infrastructure market has become.
The restrictions reportedly affected some of Meta’s internal AI work. Meta employees were also asked to use AI tokens more efficiently. AI tokens are the units used to measure how much text or data an AI model processes during prompts and responses.
The report highlights a bigger issue across the AI industry. Companies are racing to build smarter models, but those models require enormous computing power to train, test and run.
For years, the AI conversation focused mainly on model quality, product launches and chatbot features. Now, the real competition is also about access to chips, data centres, electricity and cloud capacity.
Google has its own AI products to support, including Gemini, AI Overviews, AI Mode and enterprise AI services through Google Cloud. At the same time, it is also selling AI infrastructure to outside customers. That creates pressure because Google must balance its own internal AI demand with the needs of cloud clients.
Meta, meanwhile, is also trying to scale its own AI ecosystem. The company has been investing heavily in AI infrastructure and has raised its 2026 capital expenditure forecast. Its long-term goal is to build advanced AI systems across Facebook, Instagram, WhatsApp, advertising products and future AI assistants.
Google Cloud has been one of Alphabet’s fastest-growing businesses, helped by strong demand for enterprise AI tools and infrastructure. But strong demand also creates a problem: customers want more AI compute than cloud providers can always deliver immediately.
This is why the reported Gemini restriction matters. It shows that even Google, one of the biggest AI and cloud companies in the world, may have to ration access when demand becomes too high.
In simple terms, Big Tech is not only competing to create better AI models. It is also competing to secure the physical infrastructure needed to run them.
Meta has made AI one of its biggest strategic priorities. The company wants AI to power better recommendations, advertising tools, smart assistants, creator tools and future products.
But the company has also faced delays in its AI roadmap. Earlier reports said Meta had delayed the rollout of a new AI model known internally as “Avocado.” That delay was linked to performance concerns and showed that Meta was still trying to close the gap with rivals such as Google, OpenAI and Anthropic.
Using Gemini access could help Meta test, compare or support some internal AI work while its own models continue to improve. But if that access is limited, Meta may need to rely more heavily on its own infrastructure or seek capacity from other cloud and AI partners.
The reported Google-Meta restriction comes as AI infrastructure spending across the technology sector continues to rise sharply.
Meta has said it expects 2026 capital expenditures, including finance lease payments, to be between $125 billion and $145 billion. Alphabet has also increased its AI infrastructure spending plans, with Google Cloud revenue growing strongly because of demand for AI products and enterprise infrastructure.
These numbers show how expensive the AI race has become. Companies are no longer spending only on software teams. They are spending on servers, chips, data centres, power supply, networking and cooling systems.
Yet the latest report suggests that even huge spending plans are not enough to remove capacity shortages immediately.
The situation also shows a new reality in the AI market: access to powerful models may depend not only on contracts, but also on available compute.
For customers, this means AI services may become more expensive, more carefully limited or harder to scale during periods of high demand. For AI companies, it means infrastructure planning is becoming just as important as model development.
Google limiting Meta’s Gemini access also raises an interesting competitive question. Google and Meta are rivals in advertising, AI and consumer technology. At the same time, Big Tech companies often buy infrastructure or services from one another when needed.
That creates a complicated relationship where companies can be both partners and competitors.
The larger message is clear: the AI boom is entering an infrastructure-first phase.
Building a powerful chatbot or AI assistant is no longer just about having talented researchers and good data. Companies also need enough chips, energy, servers and cloud capacity to serve millions or billions of users.
The reported Gemini limits on Meta show that even the biggest players are not immune to AI supply constraints. As demand for AI keeps growing, compute access may become one of the most important advantages in the technology industry.
Most investors building around AI stocks are chasing growth, and that...
Artificial intelligence is changing how we use video. It used to be a...
If you look at how digital content is made today, everything moves inc...
Let’s be entirely honest about the modern corporate environment: nobod...
AI video generators create ads from product photos by first understand...
Google is officially retiring Veo as the default video model inside th...
Discussion