Apple earnings for Q4 2026: what Siri AI and memory costs mean for business devices

On Monday 2 November Apple reported its fiscal fourth-quarter results, and the headline numbers were strong. Apple earnings for the quarter to 26 September showed revenue of $114.5 billion, up 12% on a year earlier, a September-quarter record. The forecast that followed was less comfortable. Apple's chief financial officer, Kevan Parekh, said the company expects gross margin "to be between 45 percent and 46 percent" in the December quarter, and John Ternus, on his first call as chief executive, said: "What we're seeing with memory pricing right now is unprecedented", according to the call transcript.

"apple stock" was the biggest business search on Google Trends the next day. This piece is not about the share price, and nothing here is investment advice: it is about building systems, not buying shares. It sets out what Apple reported and said about AI, where to read the claims carefully, and what it means for a business whose staff work on Macs and iPhones.

Apple Q4 2026 results: the figures

The call was a first in two ways. Ternus became chief executive on 1 September, succeeding Tim Cook, and opened by saying "It's a privilege to be with you for the first time as Apple CEO", joined by Parekh. Both names and titles also appear in Apple's own release.

Sources: Apple's consolidated statements filed with the SEC and the call transcript.

Gross margin for the quarter was 47.9%, "including a favorable impact of approximately 1 percentage point from tariff refunds", and full-year revenue was a record $479 billion, Apple said. Parekh said Mac revenue rose "despite supply constraints". Apple's chief operating officer, Sabih Khan, said the September-quarter shortages on iPhone, Mac and iPad were mostly about advanced chip production, and that in the December quarter "we expect constraints on Mac and iPad, primarily driven by memory".

For the December quarter, Parekh guided to revenue growth of 7% to 10%, Services growth of 8% to 9%, gross margin of 45% to 46%, and operating expenses of $20.4 billion to $20.8 billion. Revenue guidance was above the 6.1% analysts expected, but the margin range sat below the 46.3% average estimate, and the shares fell about 3% in extended trading.

Memory costs, prices and the gross margin squeeze

The margin forecast is the part of these results that reaches a business's budget. Parekh said the guidance "includes primarily the impact of higher memory costs", along with "some costs related to the ramp of Siri AI", on the call. Ternus said Apple expects "even higher memory costs, partly offset by a few factors": price changes, inventory bought earlier, and cheaper non-memory components. He said Apple expects memory prices to keep rising beyond the December quarter.

Apple has already passed some of this on. According to the transcript, it changed prices on Mac and iPad at the end of June and on iPhone in September, and Ternus said "these price increases provided some offset". Parekh said the company had "reluctantly" made those decisions. No further rises were announced on the call, and MacRumors reported that Apple dodged a question on more.

This is the same DRAM shortage that has pushed up the cost of PCs and servers across the market, covered in our piece on Sam Altman's RAM comments and the memory shortage. One more figure is worth noting: Ternus said Apple's "manufacturing purchase obligations have increased by $33 billion from June to the September quarter", as it works to secure supply. Apple is paying more for components and committing to them earlier.

Apple Intelligence and Siri AI: what was actually said

Apple's AI story on the call centred on Siri AI. Ternus said Apple "launched Siri AI in beta for US English users in September", added French, Japanese, Korean, Portuguese and Spanish in October, and described it as something that "understands what's on their screen, and holds a natural conversation".

On where the work happens, Parekh said Siri AI requests are handled both on the device and in Private Cloud Compute, Apple's server-side system. "Obviously on device, it's very, very efficient, no real costs associated with that," he said, while server-side requests carry inference costs. The mix between the two, in his words, is still "very early days". Ternus added that "there are going to be some usage limits on features that rely on server-side models", and that "anyone who needs greater capacity will be able to purchase expanded access in the future". No limits or prices were given.

On spending, Parekh said Apple will "continue to significantly increase our overall investments in AI". The filing shows what that looks like. Research and development rose to $45.8 billion for the year from $34.6 billion, while payments for property, plant and equipment fell to $9.0 billion from $12.7 billion, according to Apple's statements. That is a different shape from the Big Tech AI spending reported last week: Apple is spending on people and models, not on building data centres at the scale of Alphabet or Amazon.

The largest AI partnership was not discussed. In January, Apple and Google said in a joint statement that Google's technology "provides the most capable foundation for Apple Foundation Models", and that "Apple Intelligence will continue to run on Apple devices and Private Cloud Compute", The Register reported. Terms were not disclosed. On Monday's call, an analyst asked about usage on device "versus private cloud compute versus GCP", Google's cloud, and the answer did not address Google directly.

Where to be sceptical about Apple Intelligence

Apple's AI timelines have slipped before. In March 2025 Apple told Daring Fireball that the more personal Siri was delayed: "It's going to take us longer than we thought to deliver on these features". It has now arrived, but as a beta, in a handful of languages, and with usage limits still to be defined.

The privacy claim is strong and specific. Ternus said Siri AI "is built with essential privacy protections so no one can access users' data, not even Apple". That is Apple's design goal for Private Cloud Compute, and it is a claim worth taking seriously. It is also a vendor describing its own product, and the call gave no detail on how requests built on Google's models fit into it.

Customer examples are chosen by the seller. Parekh said "more companies are choosing Mac across a growing range of use cases, including for on-device AI", and cited a retailer running AI on Mac minis across its stores. Apple picks the examples it tells investors. They show what is possible, not what is typical.

Agentic features are a stated ambition, not a product. Ternus called agents "promising tools for users" and said Apple wants users "aware of what they're giving access to". That is a sensible principle. It is not a release date.

What Apple's earnings mean for businesses using Apple devices

Budget for higher device prices, not a dip. Apple has already raised prices twice this year, expects memory costs to keep rising, and warned of Mac and iPad constraints this quarter. If machines need replacing, plan for current prices and longer lead times. If they work, consider whether another year of service is cheaper than buying at the top of the memory cycle. Specify memory properly when you do buy; under-specified machines cost more over their life. The question of how much memory local AI actually needs is covered in our M6 MacBook Pro piece.

Know where each AI feature runs. Apple Intelligence now spans three places: the device, Apple's Private Cloud Compute, and external providers such as ChatGPT when a user allows it. Each has a different privacy position. Work that runs on the device stays there. Private Cloud Compute runs on Apple's side, under Apple's privacy design. An external integration sends data to another company under its terms. Your data-handling policy should say which of these is acceptable for which kind of information.

Use the controls you already have. Apple's device management tools can switch off individual features on company Macs: writing tools, Image Playground, Mail summaries, Notes transcription, and external integrations, which Apple says "currently includes ChatGPT and Google Lens (visual intelligence)". Most of these need supervised devices. Apple notes that some older Apple Intelligence and Siri restrictions "have been moved to modern declarative configurations", so check what your device management service supports before assuming a switch exists for Siri AI. These features are available unless someone turns them off, so the default is a decision too.

Expect limits on server-side AI. Ternus said usage limits are coming and that more capacity will be sold separately. If staff come to rely on Siri AI for real work, that becomes a per-person cost you have not yet been quoted. Do not build it into a process until the limits and prices are published.

Do not plan around features Apple has not shipped. Agentic Siri, wider languages and deeper app integration may all arrive. Apple's own history shows they can arrive late. If a product or workflow depends on a capability, design it around what is generally available today, and treat anything announced but not shipped as a bonus.

Keep important AI work independent of the phone. Siri AI is a personal assistant on a personal device. A business process that needs logging, access control, testing and a choice of model is usually better built as its own system, with the device as one way in rather than the place the logic lives.

The short version

Apple's fiscal fourth quarter was a record, led by iPhone and Mac, and its December forecast was squeezed by memory costs and the cost of running Siri AI. For a business, the lessons are practical: Apple devices are getting more expensive and may be harder to get this quarter, Siri AI is real but early and will come with limits, and the privacy position of each feature depends on where it runs. Set your device management controls deliberately, budget for current prices, and build on what has shipped rather than on what has been promised.