---
title: "Google Couldn't Meet Meta's Demand for Its Gemini AI Models, FT Reports"
description: "Google told Meta it could not supply all the Gemini AI capacity the social-media giant wanted to buy, the Financial Times reported — a constraint that disrupted some of Meta's AI work and underscored how scarce computing power has become, even between fierce rivals."
category: "Markets"
category_url: https://boursel.com/category/markets
author: "Marcus Feldman"
published: 2026-06-28T05:43:00.000Z
updated: 2026-06-28T05:43:00.000Z
canonical: https://boursel.com/article/google-couldnt-meet-metas-demand-for-its-gemini-ai-models-ft-reports
tags: ["google", "meta", "gemini", "ai", "cloud-computing"]
---
# Google Couldn't Meet Meta's Demand for Its Gemini AI Models, FT Reports

Google told Meta it could not supply all the Gemini AI capacity the social-media giant wanted to buy, the Financial Times reported — a constraint that disrupted some of Meta's AI work and underscored how scarce computing power has become, even between fierce rivals.

Google has put limits on how much of its Gemini artificial-intelligence models Meta can use — but the reason looks less like a competitive snub than a supply problem. Around March, Google told Meta it "could not meet the full Gemini capacity the company had sought to purchase," [the Financial Times reported](https://www.investing.com/news/economy-news/google-limits-metas-use-of-its-gemini-ai-models-ft-reports-4764066), as relayed by Investing.com. Meta was "particularly impacted due to its exceptionally high demand," even though other Google Cloud customers hit similar limits.

## What was actually restricted

The shortfall delayed some of Meta's internal AI projects and pushed its managers to tell staff to "be more efficient with AI **tokens**" — the units that measure how much computing work an AI system does to answer a request, per the FT. Neither company commented on the report.

A bit of vocabulary helps. **Frontier models** are the most capable AI systems, like Google's Gemini or OpenAI's GPT. Companies that don't run their own top-tier model often rent access to someone else's over the cloud, paying by usage. Meta was one of those renters — and it wanted a lot.

## A supply crunch, not a lockout

The framing matters. This was not, on the FT's account, Google cutting off a rival out of spite; it was Google running short of the physical capacity — data-center space, power, networking and AI chips — to serve everyone's surging demand at once. Google's own cloud business has been capacity-constrained for months even as orders pile up: Google Cloud booked roughly $20 billion in a single quarter while its order backlog ballooned, the company has said. With supply tight, Google appears to have spread it across customers — and Meta's unusually large appetite put it first in line to feel the pinch.

## Why Meta was buying a rival's AI at all

It can seem strange that Meta — which is pouring enormous sums into its own AI and builds the open-source Llama models — would lean on a competitor's system. According to earlier reporting by [The Information](https://www.theinformation.com/), Meta staff discussed with Google Cloud the idea of using Google's Gemini and open Gemma models, fine-tuned on Meta's advertising data, to sharpen the ad-targeting that powers Facebook and Instagram. Meta pushed back on that characterization, saying it builds its "own industry-leading, proprietary ad targeting and recommendation systems" and only "regularly evaluate[s] third-party tools for benchmarking."

Whatever the use, the volume Meta sought was large enough to strain Google's supply — a sign that, for now, Meta's own models still leave gaps it fills with rivals' technology.

## The bigger picture

Google and Meta are arch-rivals in digital advertising, which makes Google being Meta's would-be AI supplier structurally awkward: one company depends on the other for the very infrastructure it needs to compete. Today the constraint is industry-wide and physical, not a deliberate choice. But as computing capacity expands and providers regain flexibility over who gets served, whether to sell scarce AI power to a direct rival at scale becomes a real strategic decision — not just a logistical one. For Meta, the episode is a reminder of the risk in relying on a competitor's cloud, and a spur to close the gap with its own models faster.

## Sources

- [Google limits Meta's use of its Gemini AI models, FT reports](https://www.investing.com/news/economy-news/google-limits-metas-use-of-its-gemini-ai-models-ft-reports-4764066)