From Singapore: Apple's artificial intelligence strategy took a candid turn this week when reports emerged that the company has approached Google about hosting a new, Gemini-powered version of Siri directly on Google's own servers. The development, first reported by The Information, signals that Apple's internal cloud ambitions may be running well behind the demands of a genuinely competitive AI assistant.
The timing is awkward. As recently as January, Apple and Google jointly announced a multi-year collaboration under which the next generation of Apple Foundation Models would be based on Google's Gemini models and cloud technology, with those models set to help power future Apple Intelligence features, including a more personalised Siri. Critically, that announcement stated Apple Intelligence would continue to run on Apple devices and Private Cloud Compute, while maintaining Apple's industry-leading privacy standards. The new reporting suggests that commitment may be harder to keep than Apple let on.
The numbers behind Apple's cloud problem are striking. Private Cloud Compute is reportedly underutilised in its current state, with Apple using only about 10% of its capacity on average, leading to some already-manufactured servers sitting dormant on warehouse shelves. Analysts at Morgan Stanley have estimated that Apple committed approximately $4.5 billion to its Private Cloud Compute buildout through fiscal year 2025, with additional spending planned for 2026. That is a considerable sum for infrastructure that is largely going unused.
The structural reasons are telling. Apple's cloud infrastructure is reportedly very fragmented, with technologies from different teams run independently rather than drawing from one centralised pool, leading to inefficiencies where parts of the stack sit idling while other parts lack access to available server capacity. This is precisely the kind of bureaucratic inefficiency that costs shareholders money and, in this case, may have pushed Apple toward a competitor's data centres.
For Australian iPhone users, the trade implications are direct. Apple's Private Cloud Compute system was designed around a core promise: that user data sent to the cloud would not be stored, accessible to Apple employees, or exposed to third parties. If the upgraded Siri runs on Google's servers instead, that architecture changes fundamentally. Apple is reportedly in advanced talks with Google to run the new Siri inside Google's data centres, a company which already has extensive experience with large-scale language model server buildouts; Apple already relies on Google's cloud for some iCloud features, including cloud storage. But extending that relationship to Siri, which has access to users' messages, calendars, and personal context, is a materially different proposition.
Defenders of Apple's approach make a reasonable point: the company has consistently said it will maintain privacy standards regardless of the infrastructure partner. Apple has focused on privacy with its AI rollout, with much of processing happening on-device or through tightly controlled infrastructure, and says it will maintain those privacy standards throughout its partnership with Google. Apple CEO Tim Cook reinforced this on the company's Q1 2026 earnings call, stating that Google's AI technology would provide the most capable foundation for Apple Foundation Models and that processing would continue to run on-device and in Private Cloud Compute while maintaining industry-leading privacy standards. Whether the same guarantees can extend to Google-operated servers is the question privacy researchers and regulators will want answered.
There is also a competitive logic to Apple's position that deserves acknowledgement. Apple's fiscal 2025 capital expenditures were modest compared with hyperscalers, at roughly $12.7 billion for the year, with its 2026 capex budget likely to remain in the mid-teens of billions rather than the $100-plus billion range committed by rivals. Choosing to partner rather than build at scale is a legitimate corporate strategy, and one that insulates Apple from the risk of massive AI infrastructure overcapacity if consumer demand disappoints. Apple is also investing in its own silicon: the company is developing dedicated AI server chips, with analyst Ming-Chi Kuo reporting that mass production is slated to begin in the second half of 2026, before official deployment in 2027.
The honest assessment, though, is that Apple is catching up, not leading. Pressure on the iPhone maker to deliver an impressive Siri AI voice upgrade has intensified since the company delayed its promised overhaul last year despite running advertisements for it. Multiple industry analysts have noted that Apple Intelligence adoption has been tepid, with a January 2026 survey by creative agency CIRP finding that fewer than 30% of eligible iPhone users had actively engaged with Apple Intelligence features beyond the initial setup period.
For Australian businesses and consumers exposed to the Apple ecosystem, the picture is complex rather than simply alarming. Apple's partnership with Google may well deliver a genuinely better Siri faster than Apple could have managed alone. The privacy questions are real but not yet settled; much depends on the contractual architecture Apple imposes on any Google hosting arrangement. What is clear is that the tech giant's famous self-sufficiency has limits, and in the AI race, those limits are showing. Pragmatic Apple users will reserve judgement until the company publishes specifics about how data will be protected on third-party infrastructure. Given Apple's brand equity rests substantially on that privacy promise, getting those details right matters enormously, for the company and for the Australian consumers who trust it with their most personal data.