Apple told analysts on November 2, 2026, that the server costs of running Siri AI will be charged against its Services business, that the portion of each device sale it defers will now cover the assistant, and that heavier users will be able to buy more capacity. On John Ternus's first earnings call as chief executive, Apple guided operating expenses for the December quarter to between $20.4 billion and $20.8 billion, up 11% to 13% from a year earlier, a rise chief financial officer Kevan Parekh attributed to research and development, "in particular AI." Siri already handles more than 2.5 billion requests a day, according to Ternus. Neither executive said how many people use Siri AI, what its servers cost, or how much of the work runs on Google's infrastructure.

In Short

Apple's new Siri AI answers some questions on the phone itself and sends others to Apple's computers in the cloud, and Apple said on its earnings call that every cloud answer costs it money. That matters because those bills come out of Apple's most profitable business, the services side that includes iCloud and the App Store, and Apple plans to let people who use up their daily Siri AI allowance pay for more. Apple expects its operating costs in the last three months of 2026 to be up to 13% higher than a year earlier, driven by AI research, and it still has not said how many people use Siri AI.

What Apple said about Siri AI on November 2

Apple held the call after publishing results for its fiscal fourth quarter, which ended on September 26, 2026. Revenue was $114.5 billion, up 12% from a year earlier, a September-quarter record. Ternus, who succeeded Tim Cook as chief executive on September 1, opened with the product line-up and then turned to AI. Quotes in this article come from the transcript of the call published by Investing.com, which says the article carrying the transcript was generated with the support of AI and reviewed by an editor. Apple posts its own replay at apple.com/investor.

The rollout dates were the first concrete facts. "We launched Siri AI in beta for U.S. English users in September and expanded to French, Japanese, Korean, Portuguese, and Spanish in October," Ternus said. The September step was the beta that shipped with iOS 27 on September 14, excluding users under 13. His description of a US English launch is narrower than the material Apple published for the Apple TV 4K on October 13, which listed English-language Siri AI in eight markets, among them the UK, India and Australia.

Then came a number. "Siri has long been popular with our users, with more than 2.5 billion requests every day," Ternus said in reply to Erik Woodring of Morgan Stanley. The figure covers Siri as a whole, not the new assistant. Apple gave no count of Siri AI users, no share of devices running the beta and no measure of how often the server-backed features are used. "We are very pleased with the reviews and feedback we've received, and we're so excited that more and more users are getting to experience it," was as far as Ternus went.

He was also clear that the current product is a starting point. "It is primarily request-based today, but as we talked about at WWDC, it is built on a modern, orchestrator-driven architecture that gives us real room to expand its capabilities over time," he said. At WWDC on June 8, Apple described a system orchestrator that coordinates capabilities across the platform, drawing on the Spotlight semantic index and an app toolbox that runs on device, with models custom-built in collaboration with Google's Gemini family and adapted to run on devices and on Apple's servers. Request-based means Siri AI acts when asked. The orchestrator is the part Apple says can grow into something that acts across apps.

Privacy carried the rest of the pitch. "Like everything we make, Siri AI is built with essential privacy protections, so no one can access users' data, not even Apple," Ternus said in his prepared remarks, and he returned to the point in three separate answers.

Who pays for the servers

The most detailed AI disclosure came from Parekh, in response to Amit Daryanani of Evercore, who asked how much of the coming fall in gross margin came from memory, product mix and the Siri AI ramp. Parekh set out two accounting effects.

"As we continue to roll out Siri AI to more customers, the associated cloud costs will be recognized through services cost of sales," he said. The second concerns revenue rather than cost: "Consistent with how we have always accounted for Maps, Siri, and iCloud for the devices we sell that include these bundled services, we defer a portion of the revenue and recognize it over the life of the device. Our hardware deferrals as of the December quarter will reflect the addition of Siri AI. Both of these items will have an impact to the gross margin."

Two pressures on one margin

The first effect places a variable, usage-driven cost in Apple's highest-margin line. Services gross margin was 76% in the September quarter, up 40 basis points from June, against 37.2% for products, according to Parekh's prepared remarks. Every Siri AI request that goes to a server now adds to the cost of a business whose profitability is the reason Services matters so much to Apple's earnings. Users do not pay per request. Apple does.

The second effect is quieter. Part of the price of each iPhone, iPad and Mac is already set aside and recognized over the device's estimated life, because the device comes with services such as Maps and iCloud. From the December quarter, that slice will reflect Siri AI as well, which in PPC Land's reading means a slightly smaller share of each sale is booked as revenue at the moment of purchase. Apple did not say how large the extra deferral is or over how many years it will be recognized.

Both effects sit inside the gross margin guidance. Apple expects a gross margin of between 45% and 46% in the December quarter, against 47.9% in September, a figure that included about one percentage point from tariff refunds. Parekh said in his prepared remarks that the range "includes primarily the impact of higher memory costs, as well as some costs related to the ramp of Siri AI." Daryanani put the underlying decline, once the refunds are stripped out, at about 140 basis points at the midpoint, and Parekh accepted the magnitude. "We expect most of that decline to be driven by higher memory costs," he said. The Siri AI share of the remainder was not quantified.

On the device or in the cloud

Wamsi Mohan of Bank of America pushed on the cost question directly, framing it in tokens, the units in which AI models read and write text. He asked Apple to "give us some sense of how the token consumption associated with Siri AI is being done on-device versus Private Cloud Compute versus GCP," whether usage was accelerating, and whether faster adoption would change Apple's compute spending plans.

Parekh confirmed the split but not its proportions, and identified the server-side expense as inference costs. "The mix of that is still very early days, so it is hard to predict exactly what the mix of those costs are going to look like. Obviously, on-device is very, very efficient, no real cost associated with that. But on the server side, we would have some inference costs," he said, adding that Apple was "in very, very early stages of understanding the kind of mix of those costs, as well as any further requirements we might have on the compute side."

That answer explains the economics. A query handled by the Neural Engine in a customer's phone costs Apple nothing at the margin, because the customer has already paid for the chip. A query sent to Private Cloud Compute carries a cost that grows with use. How much of Siri AI's work can stay on the device? Parekh would not estimate it, and he did not say whether usage was accelerating.

The Google question

He also did not address the third option in Mohan's question. GCP is Google Cloud Platform, and neither Parekh nor Ternus mentioned Google or Gemini at any point on the call.

The connection is documented elsewhere. Apple and Google announced a multi-year agreement in January under which Google's Gemini technology would underpin Apple's next-generation models, and neither company disclosed financial terms. At the time, the companies said Apple Intelligence would continue to run on Apple devices and Private Cloud Compute. Then, on June 10, Help Net Security reported that Apple was extending Private Cloud Compute to Google Cloud infrastructure and Nvidia GPUs, citing Apple's security research blog. According to that report, Apple keeps control of the software running on those machines, maintains a verifiable ledger of the Google Cloud hardware in its fleet, and said the full set of protections would be phased in during a summer preview.

If that arrangement is in production, "Private Cloud Compute versus GCP" is not a clean distinction: a server-side Siri AI request could run inside Apple's privacy system on Google's hardware. Parekh's phrase, "the Private Cloud Compute environment that we have outlined," does not say where that environment physically sits or who is paid for it.

Charging for more Siri AI

The clearest statement on how Apple intends to recover some of those costs came in the first question of the call. Michael Ng of Goldman Sachs asked about iCloud+, saying "it's becoming a more strategic product, just given the tie-in with home security as well as the potential monetization opportunity for Apple Intelligence."

Ternus answered by describing expanded access. "There's going to be some usage limits on features that rely on server-side models, which include Siri AI and intelligent photo editing and some other things like that. Anyone who needs greater capacity will be able to purchase expanded access in the future. So, kind of to your point, in the beginning, iCloud tiers were largely about storage, but we've been adding more capabilities and features to it."

The limits themselves are not new. Apple's product pages have carried footnotes on daily usage limits since the September launch, and the iPad mini footnotes in late October said expanded access would be available for a fee in the future, without a price or a date. What the call added was the framing. Ternus placed paid capacity inside an answer about iCloud tiers, beside the news that iCloud+ plans now include Apple TV and Apple Arcade in more than 100 countries and regions. He did not say that expanded Siri AI access will be sold as an iCloud+ tier, how much it will cost, or when it will start.

Apple already sells AI behind a subscription. In October, it put a new Keynote and Pages assistant, built on Apple models and OpenAI, behind its $12.99-a-month Creator Studio plan. Cloud services, the category that includes iCloud, set an all-time revenue record in the September quarter, according to both executives.

Parekh widened the revenue case beyond subscriptions. "We also see a wide range of ways that AI can contribute to revenue growth over time," he said. "On the product side, we feel AI offers a compelling reason for customers to upgrade and switch in order to get the latest capabilities, and we also feel there are opportunities across third-party apps and partnerships as well as first-party opportunities, as John highlighted earlier, with the expanded access to server-side features of Siri AI and Apple Intelligence features as well." He did not name a partnership or say what share of iPhone upgrades Apple attributes to AI.

Read together, the two disclosures form one mechanism, in PPC Land's reading. Daily limits cap how much free server-side use each customer can generate, which bounds the cost that flows into services cost of sales. Anything above the cap becomes something Apple can charge for.

Agents, apps and who owns the customer

Three analysts asked, in different words, whether AI agents threaten Apple's position between users and the services they buy. Woodring framed it first: "As the world of AI evolves from chatbots to consumer agents, it seems like there are growing opportunities for Apple. It also seems like there are risks." He asked how Apple views "any potential risks of some of these agents trying to encroach on your services business."

Ternus treated third-party agents as a permissions question. "As far as agentic capabilities, we think these are promising tools for users. We just want to make sure that users are aware of what they are giving access to. We want to make sure this is done in a way that protects security and privacy," he said, adding that "Apple devices, we think, are the best places to use all forms of AI."

Daryanani went further. "Do you see the potential for the primary customer relationship to shift from the device to the agent over time?" he asked, before ending with: "How far along do you think Siri is in bringing those pieces together right now?" Ternus answered with the intelligent personal hub, his phrase for the iPhone. "So much of your personal data is on your iPhone, whether it's your contacts, your communications, health data, financial data, all of that. Siri can tap into that and help you access it in a way that's very secure and private and that nobody else has access to. Even Apple doesn't have access to it." He did not say how far along Siri is.

Krish Sankar of TD Cowen asked how "the agentic capability of Siri built into the OS platform versus AI agents that are confined within an app architecture" would evolve. Ternus repeated that Siri is "primarily request-based" and said: "We think fundamentally, what is most important to us is that we are giving people these capabilities that are private and secure, and we are really focused on protecting people's data."

The competition the analysts described already runs on Apple's hardware. Google restricted its Gemini Spark agent for macOS to $99.99-a-month Ultra subscribers when it launched the beta at the end of June. On July 9, OpenAI folded its Atlas browser into ChatGPT Work, an agent inside a desktop app, and on October 15 its ChatGPT extension for Chrome gained the ability to operate open tabs, filling forms and sending messages. Alphabet said on October 28 that its Gemini app had passed one billion monthly active users. That is a different measure from Siri's daily requests, so the two cannot be compared directly.

The shift toward agentic AI puts a specific asset at stake. The personal data Ternus listed - contacts, messages, health and financial records - is the context an agent needs to act for a user, and it is the context Apple says only Siri can reach in a private way. Whether third-party agents will get comparable access through Apple's frameworks, and on what terms, was not discussed.

The Mac as a local AI machine

Apple's other AI argument was about hardware that removes the cloud bill altogether. "More than 40 years after the first Mac, the category is having its biggest moment yet, with every Mac in our latest lineup a powerful platform for AI," Ternus said. "Mac mini and Mac Studio in particular are pushing the boundaries of local, private AI, giving users the ability to do things once considered impossible."

Parekh gave two enterprise examples. "Gap Inc. is running AI entirely on Mac mini for loss prevention across hundreds of stores, eliminating cloud token costs. Christie's, a world-leading auction house, is running local models on Mac Studio to improve core productivity workflows such as pricing estimates." Cloud model providers bill by the token, so a model running on a machine the customer owns turns a recurring charge into a one-time hardware purchase. Apple did not say how many Macs either company bought or which models they run.

The chips are built for it. The A20 Pro in the iPhone 18 Pro "features a new dual 16-core Neural Engine, accelerating on-device AI and computational photography," according to Ternus, and the M6 in the new 14-inch MacBook Pro and iMac carries the same dual 16-core Neural Engine "to drive private on-device AI tasks." In October, Meta released its 816-billion-parameter Muse Spark Open model running on two linked Mac Studios, with Ternus congratulating the team in a post on X.

Mac revenue rose 24% to $10.8 billion in the quarter, and Apple has "never sold more Macs in a fiscal year," Ternus said. Supply is the constraint. Chief operating officer Sabih Khan said: "For the December quarter, we expect constraints on Mac and iPad, primarily driven by memory, as demand has been higher than what we expected." Local AI depends on large amounts of memory. That is the component Apple says it is shortest of.

Memory, compute and capital

Mohan placed Apple's memory problem in a wider market. "Clearly, you are now competing for memory more so versus historically from pretty well-capitalized companies who are signing long-term agreements," he said, asking whether Apple was changing how it buys. He did not name the companies. Khan answered that Apple values its supplier relationships and that "our manufacturing purchase obligations have increased by $33 billion from the June to the September quarter," a figure Apple said would appear in its annual report on Form 10-K. When Sankar asked whether Apple was signing long-term agreements with memory suppliers, Khan gave the same answer.

Ternus was blunt about prices. "What we're seeing with memory pricing right now is unprecedented," he said, and "looking beyond the December quarter, we see the pricing for memory purchases continuing to increase." Apple has passed part of that on. "We did reluctantly make the decision to change pricing on a number of our products, starting with the iPad and Mac in the June timeframe, and then iPhone more recently in September," Parekh said. The June round took the entry 14-inch MacBook Pro to $1,999, with the MacBook Air, iPad Pro and iPad Air also rising.

Apple's own AI build-out looks different from its rivals'. Its capital expenditure for fiscal 2026 fell 28.9% to $9.04 billion, according to its condensed cash flow statement. Alphabet spent $51.4 billion on property and equipment in its September quarter alone. Apple's AI spending shows up instead in research and development and, from now on, in the operating cost of Siri AI.

Research spending has risen steeply all year. R&D grew 34% to $11.419 billion in the March quarter and 32% to $11.729 billion in the June quarter, when Apple's Form 10-Q cited higher infrastructure costs including investments in artificial intelligence. In the September quarter it was $11.73 billion, up 32%, according to Apple's condensed statements, and total operating expenses reached $19.2 billion, up 21%.

David Vogt of UBS noted that Apple was "coming off of a year of very aggressive investments in R&D" and that the December guidance implied moderation. Parekh agreed the growth rate was coming down from fiscal 2026 levels, but said "we continue to significantly increase our overall investments in AI while continuing to invest in the product roadmap." Ternus added that "we feel confident that we are spending what we need to spend to achieve our roadmap."

Europe and China: where Siri AI is missing

The call said nothing about the European Union, where Siri AI remains unavailable on iPhone, iPad and Apple Watch. When the beta shipped on September 14, Apple said Mac and Apple Vision Pro users in the EU could use Siri AI in a supported language, while iOS, iPadOS and watchOS were excluded at first. Apple has linked the exclusion to the Digital Markets Act and says it is working to find a path forward. Europe still set a September-quarter revenue record, and Services set all-time records in the Americas and Europe, according to Ternus.

China is the sharper case. Siri AI and the other new Apple Intelligence features are not available there while Apple works through regulatory requirements. Yet Greater China was Apple's fastest-growing region: revenue rose 21% in the September quarter and 28% for the full year, according to Ternus. Asked by Richard Kramer of Arete Research to explain the recovery, he credited iPhone, Mac and the new line-up, and concluded: "We think the success comes from having really great products." Parekh's argument that AI gives customers a reason to upgrade cannot be tested in the market where Apple grew fastest, because the assistant is not on sale there.

What the call left open

Several numbers that would let outsiders judge Apple's AI economics were absent:

  • No Siri AI user count, activation rate or request volume separate from Siri's 2.5 billion daily requests.
  • No price, launch date or product name for expanded access.
  • No estimate of server-side inference costs, either per request or in total, and no split between on-device and cloud processing.
  • No answer on how much Siri AI work runs on Google Cloud.
  • No size for the additional hardware deferral or the period over which it will be recognized.
  • No timeline for Siri AI on iPhone and iPad in the EU, or in China.

Investors noticed the cost side. Apple shares fell 2.8% in after-hours trading to $328.85, according to Investing.com, which linked the reaction to near-term margin pressure from memory and to the costs tied to Apple's AI rollout.

Why this matters for marketers

Siri AI is designed to sit between a user and the apps on a device. The orchestrator Apple described at WWDC reads on-screen content and personal context and acts through app actions, which means an assistant, not a home screen or a search box, increasingly decides which app handles a request. That matters for every developer who buys Apple Ads to win installs. The App Store had "a record number of nearly a billion average weekly visitors" in the September quarter, according to Ternus, and advertising set an all-time record, "including our first-party Apple Ads," according to Parekh. If tasks move from tapping into apps to asking Siri, the value of a top slot in App Store search could change. Apple gave no sign of how Siri AI chooses between apps or whether that choice will ever carry advertising.

Search sits behind the same question. Woodring asked about the durability of Services "as certain things like search evolves and regulatory changes impact App Store monetization." Google's payments for default search placement in Safari, roughly $18 billion to $20 billion in 2023, depend on users typing queries into a browser. Parekh named third-party licensing as one of the main headwinds for December-quarter Services growth, guided at 8% to 9%. Ternus's response on new revenue was brief: "We think intelligence is going to be another opportunity as well." Apple did not say whether AI answers inside Siri will reduce Safari searches, or how that would affect licensing income.

The privacy framing also shapes the ad market. Apple's claim that "no one can access users' data, not even Apple" positions Siri AI as the one assistant that can use personal context without exposing it. The same company builds the operating system, runs the assistant, sells ads in the App Store and Maps, and decides how much access third-party agents and ad tech receive. For marketers, the call confirmed that Apple's AI will carry real running costs and that Apple intends to charge for heavier use. What it did not confirm is how much of the relationship between users and apps Siri AI will end up controlling.

Timeline

Summary

Who: Apple, its chief executive John Ternus, chief financial officer Kevan Parekh and chief operating officer Sabih Khan, answering analysts from Goldman Sachs, Morgan Stanley, Evercore, Bank of America, TD Cowen, UBS and Arete Research.

What: On its fiscal fourth-quarter call, Apple said Siri AI's cloud costs will be recognized in services cost of sales, that hardware deferrals will reflect Siri AI from the December quarter, that users needing more server-side capacity will be able to buy expanded access, and that December-quarter operating expenses will rise 11% to 13% to between $20.4 billion and $20.8 billion, driven by research and development, in particular AI. Ternus said Siri handles more than 2.5 billion requests a day. Apple gave no Siri AI usage figures and did not address how much of the workload runs on Google Cloud.

When: The call took place on November 2, 2026, covering the quarter ended September 26, 2026. The guidance covers the December 2026 quarter.

Where: Apple reported from Cupertino, California. Siri AI is in beta in English, French, Japanese, Korean, Portuguese and Spanish, but not on iPhone, iPad or Apple Watch in the EU, and not in China.

Why: Siri AI turns AI from a fixed research cost into a running cost that grows with every server-side request and lands in Apple's most profitable business. How Apple pays for it, through paid capacity, upgrades or partnerships, and how much of the customer relationship it keeps as agents from Google and OpenAI spread across its devices, will shape the App Store, Apple Ads and the search licensing income that marketers and publishers depend on.