Google limited Gemini for Meta due to a lack of capacity and exposed the shortage of chips, energy, and data centers
Nuevo
Agregar La Derecha Diario en
Compartir:
Gemini found itself at the center of an unexpected situation after Google reduced the available capacity for Meta. The decision forced the company of Mark Zuckerbergto ration the use of tokens and exposed a problem that affects the entire artificial intelligence industry.
According to Financial Times, the limitation is not due to a drop in demand, but rather a lack of sufficient infrastructure to support the sector's growth. The shortage of chips, energy, and data centers is already impacting even the largest tech companies in the world.
The decision forced Mark Zuckerberg's company to ration the use of tokens
Why Google limited the use of Gemini for Meta
According to information published by Financial Times, in March Google informed Meta that it could no longer supply all the capacity of Gemini that it needed.
As a result, Meta asked its engineers to optimize the use of artificial intelligence tokens and accelerated the return to its own Llama models for certain tasks.
Meta was using Gemini for a key function: content moderation, aimed at detecting scams, removing harmful posts, and maintaining platform safety.
Why did Google limit the use of Gemini for Meta
The infrastructure shortage limits AI growth
The case revealed that the main challenge for artificial intelligence is no longer just the development of new models, but the physical capacity to run them.
According to the publication, neither Google nor Meta can currently acquire all the processing capacity they need, even with virtually unlimited financial resources.
The lack of infrastructure limits the growth of AI
The main bottlenecks are concentrated in:
Data centers.
Semiconductors.
AI GPUs.
Energy needed to operate that infrastructure.
Why the case contradicts the idea of an artificial intelligence bubble
The report argues that this situation contradicts one of the most repeated arguments by the financial market: that artificial intelligence is experiencing a bubble.
While a bubble is typically characterized by an oversupply and insufficient demand, the current scenario shows exactly the opposite. Companies reserve capacity before it is available and continue to face restrictions in accessing infrastructure.
Why the case contradicts the idea of an artificial intelligence bubble
As an example, the order book of Google Cloud exceeded US$ 460 billion in future revenue commitments, while Meta raised its capital budget for 2026 to a range of US$ 125 billion to US$ 145 billion.
The multimillion-dollar contracts that reflect the lack of capacity
The shortage was also reflected in the agreements signed by major tech companies to secure additional processing capacity.
The million-dollar contracts that reflect a lack of capability
According to the report, in June Google agreed to pay SpaceX around US$ 920 million monthly for about 110,000 Nvidia GPUs housed in the data centers of xAI. The company described that contract as a "bridge" to meet the demand for Gemini Enterprise.
For its part, Anthropic also signed an agreement with SpaceX for US$ 1.25 billion monthly, according to the published information.
Which companies benefit from this situation
The publication notes that companies that own critical infrastructure are the ones currently holding the most negotiating power.
Which companies benefit from this situation
Among them, it mentions Alphabet, which seeks to expand its capacity through its own TPUs, and Amazon, which develops Trainium chips and operates hyperscale data centers.
The same scenario also favors companies linked to chip supply, power grids, and data centers, sectors where the available capacity continues to be insufficient in the face of growing demand.
