Apple deepens its reliance on Google, handing Siri over to Google Cloud

Apple is seeking to host the next generation of its voice assistant Siri on Google Cloud, reflecting the deep-seated challenges the tech giant faces due to its long-term underinvestment in cloud computing, and marking a further deepening of its reliance on external infrastructure.

According to a report by The Information on Tuesday, at Apple's request, Google has begun a feasibility study on deploying dedicated servers within its data centers to run the new version of Siri powered by Gemini. Sources revealed that the related plan must meet Apple's privacy standards. Apple has previously relied on Google Cloud for online storage and internal AI model training services, and if the hosting arrangement is realized, the cloud computing collaboration between the two parties will enter a new phase.

The backdrop of this move is the severely low utilization rate of Apple's own cloud infrastructure. According to former employees, Apple's private cloud computing system designed specifically for AI has an average utilization rate of only 10%, with some servers still stored in warehouses and not yet installed.

Meanwhile, Apple's progress in the AI field continues to lag behind expectations, with the more conversational new version of Siri yet to be launched, and the market has reacted lukewarmly to its released AI features.

Financial Orientation Suppresses Cloud Computing Strategic Investment

Apple's difficulties in cloud computing can be traced back decades, rooted in the company's finance team long viewing cloud computing as a cost burden rather than a strategic asset.

According to about a dozen former Apple executives and engineers, the finance department has consistently suppressed large-scale investments in its own cloud infrastructure over the years, preferring to rent computing power from external providers like Google and Amazon Web Services (AWS) to maintain flexible control over infrastructure costs. This orientation has led to a continuous loss of talent in the cloud computing field.

Mike Abbott, who joined Apple in 2019 and led the integration of cloud infrastructure, left in 2023, and many of his team members subsequently followed him to General Motors, further weakening Apple's internal cloud capabilities.

In terms of capital expenditure, the gap between Apple and its peers is particularly significant. Tech companies like Meta, Microsoft, Google, and Amazon have made substantial investments in data centers in recent years to meet the explosive growth in AI computing demand, while Apple has largely remained on the sidelines.

Private Cloud System Issues Constrain AI Implementation

Apple's private cloud computing system, released in June 2024 alongside Apple Intelligence, was actually incomplete at the time of release, delayed by about six months from the original plan, and only officially launched by the end of 2024.

Former employees pointed out that the system's performance within Apple's own data centers has not been ideal. Firstly, the software updates for private cloud computing servers take much longer than for other types of servers; secondly, the private cloud computing servers equipped with Apple's self-developed chips are not designed for AI workloads, showing significant limitations when running large models like Google's Gemini.

Igor Naverniouk, a former Apple engineer who participated in the development of the next generation of Siri and left in December last year, pointed out the fundamental differences in engineering culture between Apple and Google: "At Google, most resources are centrally managed, and everyone shares the same supercomputer. At Apple, technology selection is done independently." "He previously worked in Google's infrastructure team for over a decade.

Google's Cooperation: From Forbidden Zone to Core Dependency

The collaboration between Apple and Google in the field of cloud computing has not been smooth sailing. For many years, Apple explicitly prohibited AI engineers from using Google Cloud, citing privacy protection as the reason—Craig Federighi, the software chief responsible for privacy affairs, has repeatedly vetoed proposals to use Google Cloud for AI computing power.

The turning point came in 2023. Google upgraded its security systems to meet Apple's privacy requirements, prompting Apple to begin incorporating Google Cloud into its AI infrastructure and to adopt Google's self-developed Tensor Processing Unit (TPU) chips—Apple estimates that its operating costs are significantly lower than those of comparable chips from NVIDIA.

In January of this year, Apple announced an agreement with Google to introduce the Gemini model into Apple products, further confirming the deep binding between the two parties at the AI level. If Google ultimately takes on the cloud hosting of the new Siri, Apple's dependency on this competitor will extend from the model layer to the infrastructure layer, forming a more comprehensive strategic dependency.

The New Siri May Become a Turning Point

Despite the low utilization of private cloud computing systems, Apple has stated that it will launch a new version of Siri this year. Once this product gains widespread adoption among users, the demand for AI computing power may rise rapidly, and the existing infrastructure will face a severe test of whether it can support this demand.

The Information cited informed sources stating that Apple and Google are discussing hosting the new Siri, partly to prepare in advance for the potential surge in computing power demand following the launch of the new Siri.

For investors, this series of dynamics reveals Apple's structural shortcomings in the AI race: as competitors continue to ramp up their own computing power infrastructure, Apple's cloud computing capability development remains constrained by financial culture and historical issues, and the gap between the execution pace of its AI strategy and market expectations is unlikely to be bridged in the short term