Google Limits Meta's Access to Gemini AI Over Computing Capacity
Even the world's biggest technology companies are running into the same problem facing the broader AI industry: there is not enough computing power to meet demand.
According
to a report by the Financial Times, Google has restricted Meta's
access to its Gemini AI models after the Facebook, Instagram and WhatsApp
parent company requested more computing capacity than Google was able to
provide. The decision reportedly forced Meta to delay some internal AI projects
and rethink how its engineering teams used AI resources.
The
development offers a rare glimpse into the infrastructure bottlenecks behind
today's AI race, where access to computing power has become as valuable as the
AI models themselves.
The
AI Capacity Crunch
The
report says Google informed Meta around March that it could not provide the
full Gemini capacity the company wanted to purchase because of limited
available infrastructure.
Meta
was reportedly affected more than other Google customers because of the scale
of its demand. In response, the company asked employees to be more conservative
with their use of AI tokens, the units that measure interactions with AI models
and determine how much computing capacity is consumed.
The
situation highlights a growing reality across the industry. While companies
continue investing billions of dollars in AI chips, cloud infrastructure and
data centres, demand for AI computing resources is growing even faster than
supply.
Google
has previously acknowledged these constraints. During its first-quarter
earnings call, Chief Executive Sundar Pichai said computing capacity had become
one of the biggest factors limiting the growth of Google Cloud, despite the
business generating $20 billion in quarterly revenue.
Why
It Matters
Meta
is rapidly expanding AI across its ecosystem, from AI assistants inside
Facebook, Instagram and WhatsApp to the continued development of its Llama
family of large language models. Those ambitions require enormous amounts of
computing power to train models, run AI services and support millions of users
simultaneously.
The
reported restrictions show that even companies building some of the world's
most advanced AI systems still depend on external infrastructure providers to
meet growing demand.
More
broadly, the story reinforces a shift taking place across the AI industry.
Competition is no longer centred solely on who builds the best model, but also
on who can secure enough GPUs, cloud capacity and data centre infrastructure to
keep those models running at scale.
Why
Nigeria Should Pay Attention
For
Nigeria and the rest of Africa, the development underscores one of the biggest
barriers to building competitive AI products: access to compute.
Many
African startups, researchers and universities already struggle to access
affordable AI infrastructure, making it difficult to train and deploy
large-scale models locally. If global technology companies worth trillions of
dollars are facing capacity shortages, the challenge is even greater for
emerging AI ecosystems.
The
growing demand for computing infrastructure also strengthens the case for
national and regional investments in AI-ready data centres and shared compute
resources, an area that countries such as Nigeria have already begun
prioritising through initiatives like the Nigeria AI Scaling Hub.