The Financial Times reports that around March of this year, Google informed Meta that it would no longer be able to satisfy the substantial compute power requirements for the Gemini model. Consequently, Meta was compelled to reduce its usage and delay several internal AI initiatives. At present, these limitations persist, prompting Meta to instruct its employees to be more judicious in their utilization of AI tokens. This represents a significant shift from the company’s earlier, enthusiastic promotion of AI. Google has set usage limits for the Gemini application based on compute quotas, akin to mobile data plans. Under this system, users are expected to operate within a rolling 5-hour weekly window, with a capped weekly usage limit. This approach is designed to guarantee equitable API resource access for all clients amid this period of swift expansion. Google’s move underscores the ongoing challenge of compute power supply as a fundamental constraint in AI industry development. Despite Google’s continuous investments in AI infrastructure, it still grapples with meeting the soaring market demand.