How big could an Apple AI services business be?

Apple investors breathed a sigh of relief on June 10th when the company’s CEO, Tim Cook, and his deputies held a two-hour presentation about their intention to bring artificial intelligence to the iPhone, iPad, and Mac. Apple had been seriously lagging the giants of AI, Microsoft, and Alphabet’s Google, but they managed to turn things around sharply with a vision for AI assistants that actually offered some nice functionality.

The biggest payoff for investors, however, could be yet to come, and it’s not in any Wall Street spreadsheets at the moment. Apple has emphasized the security aspects of its iCloud facilities for the purpose of protecting user data. That opens the door to a meaningful boost in Apple’s services revenue, its most profitable business and its fastest-growing.

Services, including sales of music and video and apps, was a fifth of Apple’s total revenue in the fiscal year that ended in September, at $85 billion, a staggering sum considering it was just $14 billion a decade ago. While Apple’s total company sales dropped 3 percent, held back by a sluggish iPhone 14 sales cycle, services revenue rose 9 percent.

In the most recent quarter, reported in May, services sales growth sped up, reaching 14 percent, year over year, and it crossed over to more than a quarter of Apple’s revenue, even as total company sales once against slipped.

Services’s profit, moreover, is almost twice what it is for iPhones and iPads and Macs, at 75 percent of sales last quarter versus a profit of 37 percent for product sales.

Anything that contributes to services sales should be music to investors’ ears. AI could be a shot in the arm if Apple decides to broadly sell an AI predictions service in the cloud.

The June 10th unveiling of “Apple Intelligence” was mainly about things one can do on the device itself. However, some of the processing of AI tasks that requires more heavy lifting will be sent from the device to Apple’s iCloud, where it will be run on custom-built Apple servers.

That wasn’t a total surprise, as rumors leading up to the event had discussed how Apple might run some computing in the cloud to handle big AI workloads. What was surprising was the company’s emphasis on security in the cloud. That is an approach that was clearly meant to distinguish the company: the perceived lack of protection in services such as chatbots that can leak data, especially, personally identifiable information.

Apple’s head of engineering, Craig Federighi, made a big point of noting that Apple will provide a secure version of its chips and its software for those cloud computers. For example, Federighi claims all user data will be wiped from the cloud servers once an AI operation is performed. He promised that Apple’s cloud software will be open to third-party audits to verify that claim, a statement that is echoed by Apple engineers in a blog post that accompanied the keynote.

In essence, the cloud becomes a mirror of the same sales pitch Apple makes with its iPhone, which is that data is never compromised.

The question for investors is, Is all the work to secure the cloud meant simply as a way to sell more iPhones, or is it a plan for something bigger?

It’s conceivable Apple could offer a retail cloud computing AI service, not unlike the services offered by OpenAI, search engine Perplexity.ai, Google, and privately held Anthropic, which makes a family of generative AI programs that compete with OpenAI called “Claude.”

All those services have coalesced around a price of $20 per month, which is the price for ChatGPT Plus, Perplexity.ai’s Perplexity Pro, Google’s Google One subscription, which includes its Gemini language model, and Anthropic’s Pro account. (Versions meant for corporate teams and for enterprise come with higher price tags.)

It’s the same price, in fact, at which Apple sells its Apple One account, a bundle that includes iCloud storage, an Apple TV+ subscription, Apple Music, and the Apple Arcade service.

The opportunity for Apple is to add another offering to the bundle with an AI component.

What’s interesting is that Apple has scaled way beyond what the others have when it comes to subscriptions. The company reported over a billion paying subscribers last month. The only outfit that comes close is Google, which reported in April it has a hundred million subscribers to Google One, and the same amount of subscribers to its YouTube Music offering, and its YouTube Premium subscription, and a mere eight million subscribers for what it calls YouTube TV.

The disparity in subscriber scale points to what most already know about Apple, which is that it’s a brand that can get customers to pay up as long as they perceive value. Using typical Wall Street math, if Apple got even 10 percent of its billion-plus subscribers to sign up, it would immediately become one of the biggest paid AI offerings around.

So, what could Apple offer? Probably, Apple doesn’t want to try to compete with OpenAI or Google or Anthropic in developing the biggest generative AI programs. For one thing, it’s not clear that Apple has the intellectual firepower among its engineering teams in AI to equal those firms.

For another thing, building the largest AI programs is very capital-intensive. Google spent $32 billion on capital investments last year, three times the amount Apple spent, even though Apple has to build physical retail stores and buy capital equipment for some device assembly.

Not all of Google’s big spending is AI spending, but AI is taking up an increasing chunk of the spend. Apple’s CFO Luca Maestri indicated last month that “we are obviously very excited about the opportunity with Gen AI,” while indicating there will be no massive increase in capital spending. “We plan to continue along the same lines going forward,” he said of capital outlays.

The thing is, Apple doesn’t have to be at the forefront of generative AI development. A lot of what goes on with “large language models” and chatbots is rapidly becoming a commodity. Not only are there leading open-source AI programs, such as Meta’s Llama-3, there are also smaller companies that can offer a lot of functionality without the huge budget of a Google or a Microsoft. They include startup Databricks, which has demonstrated very competent offerings that are focused on practical use cases without trying to create the be-all and end-all of generative AI.

Apple’s best path would be to put together a curated group of AI offerings that are focused on users’ own concerns, and their data. For example, a health offering could tie into Apple’s devices to give consumers more insights into the vast health data gathered on the device. In one version, that service could be a free basic “tier” that simply measures how health and fitness have changed for an individual over time.

In a “Pro” type of account, the AI model could compare how such health and fitness metrics for the individual subscriber compare against all other paying Pro subscribers. The comparison would not be made with specific user metrics, which would violate privacy, but with a statistical average of the cohort.

Other examples include small-business services using AI to analyze business trends. That is a feature that every software company in the land is rolling out in one form or another. Most of those software vendors, however, do not have Apple’s relationship with small business customers; they’re trying to sell to enterprises. And they don’t have the access to data in a secure fashion that Apple can offer.

That last point is the key. If Apple has truly built a secure “enclave” in the cloud for users’ data, as it claims, then Apple doesn’t have to lead with science projects the way OpenAI does. It can offer tailored, focused uses of AI, built on commodity generative AI offerings, that speak to users’ actual priorities in a way that is less likely to make those customers feel robbed of their personal information and their intellectual property.

Just as Apple did with music and movies 20 years ago with iTunes on the device and in the cloud, there’s a chance to take a whole market that is currently up in the air, filled with features and also with fear, confusion, and risk, and make it palatable to a customer base that has proven willing to pay for what it values.