The global conflict between supply and demand for artificial intelligence infrastructure is intensifying among the world's leading technology firms. According to informed sources, Alphabet (GOOG) informed Meta Platforms, Inc. (META) around March of this year that it could not meet the social media giant's full demand for computing power for its Gemini AI model, imposing usage limits. This highlights that even the largest AI service providers are struggling to cope with the surging demand for computational resources.
Reports indicate these restrictions remain in place, having disrupted and delayed several of Meta Platforms, Inc.'s internal AI initiatives. In response, Meta Platforms, Inc. has instructed its employees to improve efficiency in their AI compute usage, promoting careful management of AI tokens internally. Both Alphabet and Meta Platforms, Inc. have declined to comment on the matter.
Escalating Demand Pressures Major Providers
This situation is forcing Alphabet to accelerate its expansion efforts. Earlier this month, the company entered into a compute leasing agreement with Elon Musk's SpaceX worth $920 million per month. Alphabet CEO Sundar Pichai acknowledged during the first-quarter earnings call that the company has recently faced constraints in compute capacity, stating that cloud revenue would have been higher if demand could have been fully met.
Meta Platforms, Inc. is not an isolated case. Multiple sources note that other enterprise clients of Alphabet have also faced varying degrees of restriction, with Meta Platforms, Inc. being the most affected due to the exceptional scale of its requirements. This episode underscores how the explosive growth in AI inference workloads has become one of the industry's most significant challenges.
Persistent Compute Bottlenecks Impact Key Clients
Despite tech giants investing hundreds of billions of dollars in chips, data centers, and power supply, the supply of AI computing power continues to lag behind the pace of demand growth.
Alphabet's cloud business revenue surpassed $20 billion for the first time in Q1, and its backlog of signed but unfulfilled cloud contracts nearly doubled sequentially, exceeding $460 billion. Pichai made it clear that compute constraints are expected to persist in the near term.
Within this context, the impact on Meta Platforms, Inc. is particularly pronounced. Sources indicate that the intense demand from major enterprise clients like Meta Platforms, Inc. is a direct driver for Alphabet's accelerated search for external compute sources. As companies deploy chatbots, coding assistants, and AI agents at scale, inference workloads—the compute power consumed by models performing tasks after training—are becoming the industry's central bottleneck.
Internal Projects Hindered, Accelerating Shift to Proprietary Models
Meta Platforms, Inc. uses Gemini extensively internally for tasks such as platform safety reviews (including identifying scam content and removing harmful material), customer service and advertising support chatbots, as well as certain internal workflows and code development, while also using other models like Anthropic's Claude.
Sources reveal that Meta Platforms, Inc. initially chose Gemini because its performance surpassed that of the company's own open-source Llama model. However, as compute restrictions tightened, Meta Platforms, Inc. is accelerating its migration towards in-house models. Multiple sources state that Meta Platforms, Inc. has recently begun prioritizing the rollout of its newly launched Muse Spark model, which is considered to rival Gemini in performance, helping to reduce reliance on external models.
Meta Platforms, Inc. CEO Mark Zuckerberg has previously committed to increasing investment in AI talent and infrastructure, aiming to build what he terms "personal superintelligence." Unlike Alphabet, Meta Platforms, Inc. does not have a cloud business and is accelerating the construction of its own data center infrastructure, pledging to invest a cumulative $600 billion in the United States by 2028.
Industry Seeks Solutions Through Expansion and Partnerships
Facing compute pressure, Alphabet's agreement with SpaceX this month, along with a similar deal struck by AI lab Anthropic last month, represents efforts to bridge infrastructure gaps.
The restrictions placed on Meta Platforms, Inc. by Alphabet provide a rare window into the real pressures faced by top global AI service providers in allocating compute resources. Currently, the infrastructure bottleneck across the AI industry is extending from the training side to the inference side. Resolving the supply-demand imbalance will depend on the realization of a new wave of large-scale capital investment.
Comments