Google has limited usage of its artificial intelligence model Gemini, and some Meta AI projects have been halted or delayed, reports say. Analysts say securing computing resources is emerging as a key variable in big tech competitiveness beyond model performance.

On June 29, the online media outlet Gigazine reported that Meta has been running some internal AI projects using Google Gemini, but development schedules are being disrupted because Google has not been able to provide sufficient computing resources.

The issue is reported to have begun around March and has continued to affect some projects as of late June. Meta and Google signed a contract for Gemini usage, but the computing resources available for supply did not keep pace with Meta’s demand. As a result, some internal Meta AI projects designed on the premise of Gemini’s large-scale computing capabilities have been delayed, reports say.

The impact has also affected Meta’s internal operations. Meta has been pushing to reduce AI operating costs and has issued internal guidance to use AI tokens more efficiently, reports say. The computing resource shortage has spread beyond simple service limits to development schedules and overall cost management.

The limits are not only Meta’s problem, another explanation said. A source familiar with the matter said other Google customers are also facing certain constraints, but are not being affected as much as Meta. Meta is among the customers most actively using Google AI models, the source said, meaning it is more heavily hit by the shortage of computing resources.

The case also intersects with Google’s rapidly growing cloud business. In its April earnings release, Google said cloud segment revenue topped $20 billion for the first time. The size of cloud contracts not yet provided was tallied at more than $460 billion, up almost double from the previous quarter.

But as AI demand from large customers is rising faster than expected, analysts say the pace of securing infrastructure such as chips, data centres and power is not keeping up.

Google is also moving aggressively to secure computing resources. In June, it was reported to have signed a contract with SpaceX to secure additional computing resources. The move shows cases of mobilising external infrastructure to expand AI services.

Meta is also accelerating efforts to strengthen its own AI competitiveness. The company is expanding cooperation with semiconductor firms including AMD and Nvidia as it pursues its Personal Superintelligence strategy. It is seeking to enhance its own AI capabilities while also actively investing to secure external computing resources.

Markets are interpreting the case as a sign that the AI industry has entered a new phase. If competition in AI has so far centred on model performance and functions, it has now become an era in which competitiveness is determined by how stably companies can secure computing resources, data centres and power, the view is.

The Financial Times said Google’s case of placing caps on large customers’ AI model usage is a rare example showing infrastructure bottlenecks across the AI industry. It said that despite huge funds being poured into chips, data centres and power facilities, even big tech firms are struggling to secure enough computing resources to handle surging AI demand.

As a result, competition in the AI industry is rapidly shifting beyond model development to a race to secure infrastructure. The industry is expected to focus on whether Meta’s project delays end as a temporary supply shortfall or develop into a structural problem over the allocation of AI computing resources.