Apple is all in on AI — its own and everyone else’s

Being “all in on AI” means Apple is treating AI as a major growth opportunity, even when other companies provide some of the underlying models. Apple wants its devices and software to become the main interface for AI use. Its more than 2.5 billion active devices give it a huge audience.

For example, Siri AI uses models developed with Google’s Gemini technology. Apple also plans to let users choose among outside providers for writing tools and image generation. That lets Apple offer useful AI features without developing every model alone.

Apple is increasing R&D spending and discussing AI more often, with 45 mentions on its latest earnings call. It can earn from device upgrades, subscriptions to AI apps, and expanded access to its own server-based features. This approach also limits the need for enormous infrastructure spending, although cloud costs could rise as usage grows.

Apple Intelligence refers to Apple’s AI technology and the features it is adding across its products. The company is using AI to make Siri more capable and to offer tools that help users write, create images, and edit photos. Apple presents these features as reasons to upgrade devices.

The revamped Siri is called Siri AI and launched in beta for US English users in September. Apple expanded it to French, Japanese, Korean, Portuguese, and Spanish in October. Apple also plans writing tools and image generation, including options involving outside AI providers. Intelligent photo editing is another feature that can rely on server-side models.

The current rollout is still limited. Apple says server-dependent features will have usage limits, while some processing can happen on devices. Customers needing more capacity will be able to purchase expanded access in the future, potentially turning Apple Intelligence usage into subscription revenue.

Apple’s infrastructure spending is dramatically smaller than the reported capital expenditures of Amazon and Alphabet. Apple spent $2.2 billion on capital expenditures last quarter. Amazon spent $77.3 billion, while Alphabet spent $51.4 billion. These figures show the different scale of investment surrounding their businesses.

The totals are companywide capital expenditures, not AI-only spending. They also exclude some AI expenses, so they are not a perfect comparison. Still, Apple’s spending was only a fraction of either rival’s total. Apple combines on-device processing, its own servers, and outside AI technology instead of relying only on massive new infrastructure.

Apple’s operating expenses rose 21% to $19.2 billion as it increased research and development, including AI work. Server costs for Siri AI will appear in services cost of sales. Those costs could increase as more requests require servers rather than local device processing.

Apple can make money from AI at several points in the customer journey. It can sell iPhones and Macs that run AI features, collect fees from subscriptions to AI apps, and eventually charge customers who want more access to its own tools. AI can therefore support hardware, services, and subscription revenue.

The mechanism starts with Apple’s large installed base. Customers may upgrade or switch devices to get the latest AI capabilities. Developers and AI companies can reach users through Apple’s platform, while Apple can collect fees on subscriptions to AI apps. Apple can also sell expanded access to server-based features such as Siri AI and intelligent photo editing.

This strategy does not require Apple to build every model itself. It can use outside technology, including Google’s Lumen for Siri AI, while controlling the customer experience. The opportunity is still developing, because Siri AI has only recently begun reaching customers and paid expanded access is planned for the future.

Running an AI model on a device means the computer performs the task locally instead of sending every request to remote servers. That can reduce the number of cloud requests a company must pay for. It can also support private, local processing, which Apple highlights as an advantage of its Mac computers.

Gap is running AI entirely on Mac Mini computers for loss prevention across hundreds of stores. Because those tasks run locally, Gap avoids cloud token costs for each request. Christie’s is using Mac Studio computers and local models for workflows including pricing estimates. These examples show how Apple hardware can become an AI machine.

This approach does not eliminate cloud computing for every task. Some larger or server-side models still require Apple or another provider’s infrastructure. However, local processing can reduce usage-based cloud expenses and make AI more affordable for customers with suitable workloads. Apple says Mac Mini and Mac Studio are pushing local, private AI forward.

Server-based AI requires computing resources each time a request is processed. If more Siri and Apple Intelligence requests move from devices to Apple’s servers, Apple will need to handle more cloud activity. That creates higher operating costs, even if the extra usage also brings more customers and potential revenue.

Apple said Siri AI’s cloud costs will appear in services cost of sales. CFO Kevan Parekh also said it is too early to predict how many requests will require servers rather than on-device processing. Server-side models support features including Siri AI and intelligent photo editing, and Apple plans to limit usage for those features.

The financial effect depends on demand and pricing. More usage could increase subscription revenue if customers pay for expanded access. But if server costs grow faster than that revenue, Apple’s profit margins could weaken. The company already warned that rising memory costs would squeeze margins in the holiday quarter, although that warning was separate from Siri’s server costs.

Capital expenditures, or CapEx, are money companies spend on long-lasting assets rather than ordinary daily operations. In AI, these assets can include data centers, specialized chips, servers, storage, and networking equipment. They provide the physical capacity needed to develop and run AI services.

Large-scale AI systems process huge numbers of requests. Companies must therefore build or expand facilities, buy powerful computing equipment, and connect that equipment with high-capacity networks. The article does not detail each infrastructure category, but it compares the large companywide investments made by Amazon and Alphabet with Apple’s smaller total.

Apple spent $2.2 billion on capital expenditures last quarter, compared with $77.3 billion for Amazon and $51.4 billion for Alphabet. Those figures include more than AI and exclude some AI expenses. Apple reduces its infrastructure needs by processing some AI on devices, using its own servers selectively, and relying on outside technology.