Google curbed Meta's Gemini access as 2 tech giants compete for AI compute

Google's parent company Alphabet Inc. restricted Meta Platforms Inc.'s access to its Gemini artificial intelligence models earlier this year, a sign that demand for AI computing power is beginning to outstrip supply even among hyperscalers building their own infrastructure.

Google informed Meta around March that it could not fulfill the social media company's full request for Gemini computing capacity, according to a Financial Times report. The restrictions disrupted several of Meta's internal AI projects and delayed their timelines, the report said, with Meta affected more than other Google Cloud customers because of its unusually high demand for Gemini models.

"Meta has been affected more than other Google Cloud customers because of its unusually high demand for Gemini models," the Financial Times reported, citing people familiar with the matter. Neither Google nor Meta has publicly commented on the restrictions.

Meta had been using Gemini across a range of applications, including content moderation, scam detection, customer service, advertising tools, and software development. The company turned to Gemini because it outperformed some of its own AI models on certain tasks, the report said. In response to the capacity constraints, Meta has encouraged employees to use AI tokens — a measure of how much computing power AI applications consume — more efficiently.

The restrictions highlight a broader challenge gripping the AI industry. Companies across technology are investing billions of dollars in chips, servers, and data centers, yet demand for generative AI computing continues to outpace available supply. Google has previously acknowledged that limited computing capacity has constrained growth in its cloud business despite strong customer demand.

Meta's Shift to In-House Models

Meta has begun moving some of its workloads from Gemini to its own Muse Spark model, reducing its dependence on external AI providers. The company continues to invest billions of dollars in building its own AI infrastructure, a strategy that could insulate it from future capacity constraints at cloud partners.

The episode underscores a structural shift: access to AI infrastructure is becoming as strategically important as the models themselves. Compute power — measured in tokens, GPU hours, and data center capacity — has emerged as one of the industry's most valuable and scarce resources.

Investment Implications

For investors, the supply-demand imbalance in AI compute supports a bullish outlook for hyperscaler cloud providers such as Alphabet, Microsoft Corp., and Amazon.com Inc., as well as AI chip makers including Nvidia Corp. and Advanced Micro Devices Inc. Capital expenditure narratives will intensify as companies race to expand capacity. However, the constraints also signal rising costs and potential margin pressure for companies unable to secure sufficient compute, making vertical integration — building proprietary chips and models — an increasingly important competitive advantage.

This article is for informational purposes only and does not constitute investment advice.