Amid the explosive growth of global computing power demand, even tech giants face infrastructure bottlenecks. Recently, it was confirmed that due to a shortage of cloud computing capacity worldwide, Google has officially restricted Meta's access to its top AI model, Gemini.

Over the past period, Gemini has been the core of Meta's automated security workflow. The model has shown high efficiency in handling large-scale review tasks such as fraud detection and harmful content filtering, with performance even surpassing Meta's own open-source Llama system at times. However, with the sharp increase in AI inference workloads, despite Google achieving $20 billion in cloud business revenue in the first quarter, the expansion of its physical infrastructure still cannot keep up with the rapidly growing demand for computing power.

This computing power allocation by Google to Meta has directly caused delays in several key AI projects within Meta. Faced with this sudden infrastructure constraint, Meta's management has urgently issued instructions for teams to fully improve the efficiency of AI Token usage.

This situation, constrained by a competitor, has also become a catalyst for Meta to accelerate its independent progress. Under the push of the newly established "Super Intelligence Lab," Meta is migrating a large amount of core security and review workloads to its fully self-developed cutting-edge model, "Muse Spark."

Industry analysts believe that this move is not only an emergency measure for Meta to cope with the short-term computing power crisis, but also an important turning point for it to seek fundamental technical self-control in the field of artificial intelligence. In the current context where computing power resources have become scarce production factors, reducing reliance on third-party model platforms is gradually becoming a consensus among major tech giants.