The Financial Times (FT) reported on the 28th (local time) that Google imposed usage limits on its artificial intelligence (AI) model Gemini for Meta, Facebook's parent company. FT said Meta demanded computing capacity beyond what Google could provide, showing that even big tech corporations are not free from AI infrastructure bottlenecks.
Citing multiple sources familiar with the matter, FT said Google notified Meta in March this year of the usage restrictions on Gemini. The move reportedly delayed some of Meta's AI projects. Google has imposed usage caps on other clients as well, but Meta was said to be hit the hardest because demand for the Gemini model was exceptionally high.
Sources said Meta is responding to these limits and cutting AI expense by encouraging employees to use AI tokens more efficiently. A token is the basic unit an AI model uses to process information and generate answers.
FT said this is an example of the infrastructure bottlenecks that are deepening across the AI industry. Big tech corporations are making astronomical investments in AI chips, data centers, and power, but they are struggling to secure enough computing resources to handle surging AI demand.
As demand from major clients surges, Google is also speeding up efforts to secure additional computing capacity. Early this month, Google signed a computing capacity lease deal worth $920 million per month (about 1.41 trillion won) with Elon Musk's SpaceX.
Sundar Pichai, Google's chief executive officer (CEO), said in the first-quarter earnings release this year, "Google Cloud revenue topped $20 billion for the first time, and backlog, which we have not yet been able to serve, rose to more than $46 billion, nearly doubling from the previous quarter."
He said, "We are short of computing resources in the near term," adding, "Cloud revenue would have been higher if we had been able to meet all customer demand."