Even tech giants with their own LLMs are struggling to find enough computing power.
Google was forced to limit its use of Meta’s Gemini AI model as Mark Zuckerberg’s company outgrew its computing power, people said. financial times newspaper. The incident revealed that even tech giants with their own LLMs struggle to secure enough computing power for themselves, let alone their customers.
Meta does not run its own cloud business and is rapidly expanding its data center buildout, pledging $600 billion in cloud computing investments over the next two years. Google reportedly warned social media companies about capacity limits in March, which forced Meta to require employees to use their tokens more efficiently.
Gemini AI is used by Meta for customer service, advertiser chatbots, and coding, alongside processes such as harmful content removal and fraud detection. Sources said Meta initially chose Gemini for these purposes because it performed better than its Llama open source model. The company also employs other models such as Anthropic’s Claude for similar purposes.
Despite spending billions of dollars on data centers, large companies struggle to find enough capacity to meet their usage needs. Google itself recently agreed to pay SpaceX $920 million a month to use xAI’s data centers, citing the additional computing power needed for Gemini Enterprise.
Analysts say AI power users are benefiting from the boom, but providers like OpenAI are not yet profitable. This is because the revenue generated from AI has so far only covered a fraction of the cost. Recently, the price of tokens has soared, forcing some companies to back off from using AI, including the AI companies themselves.
