Apple has encountered inefficiencies in its Private Cloud Compute infrastructure, designed to process AI queries. Servers specifically designed for this purpose are being used at an average of only 10% of their capacity, and some equipment is gathering dust in warehouses.
According to 9to5Mac, which refer to the publication The InformationApple approached Google with a request to host its new Siri models on its servers, as the fragmentation of its internal systems, with different Apple teams working with isolated technology stacks, prevents the flexible redistribution of resources. At the same time, the finance department is dissatisfied with the growing costs of duplicate infrastructure and is currently unwilling to invest billions in rebuilding it.
Technically, the platform also lags behind: the modified M2 Ultra processors don't provide the performance needed to run large language models (LLM), and software updates require significant effort and time. Against this backdrop, low user interest in Apple Intelligence features has heightened doubts about the continued deployment of in-house data centers.
Apple has approached Google with a request to place its new models Crab on its data centers, since its own Private Cloud Compute infrastructure cannot handle heavy tasks, while Google, on the contrary, already has extensive experience in the mass deployment of servers for LLM thanks to the project Gemini.
Source:
Source: 3dnews.ru
