Google reportedly curtails Meta’s access to Gemini after capacity request, Financial Times says

Google has reportedly placed limits on Meta’s use of its Gemini AI models after Meta sought additional computing capacity that Google could not provide, according to a Financial Times report carried by Yahoo Finance.

The report frames the change as part of an ongoing, capacity-constrained period in AI development, where demand for model access and the infrastructure behind it can outstrip available compute. In that environment, the supply of faster or larger-scale processing often becomes a negotiating point between major AI users and the organizations providing the models and hosting.

Meta is one of the largest consumers of external AI capabilities in the industry, using advanced models across products and internal research. Google’s Gemini models, by contrast, sit within Alphabet’s AI strategy and are made available to partners and developers through a combination of model offerings and cloud-based deployment options. The Yahoo Finance summary does not specify whether the reported limits relate to output size, request rates, specific Gemini variants, or the amount of compute Meta can reserve.

For Google, putting constraints on a partner’s use can reduce operational strain and protect service reliability for other customers, including cloud clients using Gemini in production settings. It can also shift commercial leverage toward higher tiers or differently priced commitments if capacity is tied to spend or to reserved infrastructure.

The report also suggests a practical mismatch between what Meta wanted and what Google could deliver at the time. The detail that the issue centered on “more computing capacity” is important because it implies the dispute is not primarily about the model’s quality, but about the throughput and infrastructure required to run it at scale.

As of the time of writing, the companies have not provided publicly available documentation in the Yahoo Finance item that spells out the revised terms. That means key specifics, such as when the limits began, how long they will remain in effect, and how they were measured or enforced, are not disclosed in the referenced account.

Alphabet does not discuss individual partner arrangements in every case, and AI model access can be governed by contracts or operational policies that are not always reflected in public statements. Without additional confirmation from company releases, the Financial Times report should be treated as a claim about negotiations and internal capacity planning rather than a confirmed public change to product terms.

What to watch next is whether Google or Meta acknowledges any model access or capacity adjustments, and whether cloud and AI services are updated in ways that indicate a broader pattern of throttling or re-tiering among enterprise partners using Gemini. Any public detail could clarify whether the limits are temporary, whether other customers faced similar constraints, and how Alphabet intends to manage demand as competition for AI compute intensifies.