Will Apple abandon private cloud AI?

It may mean something or nothing, but I’m a little concerned on reading a recent The Information report (via 9to5Mac) which seems to describe Apple moving away from use of Private Cloud Compute.
While the article is hidden behind a paywall, I’ve managed to assemble a short reprise of some of the points it makes – I cannot guarantee any of these points are genuinely correct.
Apple considering Google for Siri hosting
- Apple may host new Siri models on Google servers instead of Apple’s Private Cloud Compute.
- Google would run Siri servers in its data centers while adhering to Apple’s privacy standards.
Private Cloud Compute issues
- Underpowered and unable to run the latest AI models (e.g., Gemini-based Siri).
- Only about 10% of its server capacity is being used; many servers are idle.
- Updating software is complex and slow.
Fragmented cloud infrastructure
- Different Apple teams run independent technologies rather than a unified, centralized system.
- Causes inefficiencies: some resources idle while others need capacity.
- Finance team frustrated by duplicate infrastructure costs but reluctant to spend billions on overhaul.
- Attempts to unify the stack have repeatedly stalled.
Usage and demand concerns
- Initial Apple Intelligence features underused, making the Private Cloud Compute buildout seem less worthwhile.
- Expected higher demand for new Siri features, but current infrastructure may be insufficient.
What happens next?
The reports seem to be saying that AI developments may push Apple to reconsider its in-house cloud investments.
What that might mean is that while Apple ends up running the best edge devices for AI, it will no longer have total control of those Apple Intelligence interactions as they would be routed through a third-party, in this case, Google.
The problem I have there – particularly in view of the recent furore concerning Anthropic in the US – is that any weakening of Apple’s control of the interaction could end up weakening the privacy Private Cloud Compute promised. My concern is that by outsourcing the protection of that privacy Apple may end up making that privacy easier to break.
After all, while Apple may resist an attempt to break privacy, will others resist as much? Is it possible Apple is attempting to distance itself from privacy on a more general basis as legislatures everywhere lean on the company in top secret courts? I don’t know.
But the report I spent time considering leaves me concerned at what may, or, indeed, may not, be planned.
Which is why it probably makes sense for people to create their own on-prem AI services in future.
You can follow me on social media! Join me on BlueSky, LinkedIn, and Mastodon.
