1. OS 26

It’s widely expected Apple will announce a major design update to unify OS experiences across products. The design changes are rumored to resemble Vision Pro’s OS, which could include rounded app icons and more transparent backgrounds in apps.

I agree with Bloomberg’s Mark Gurman, who reported it will be the biggest revamp since iOS 7 in 2013. The overlooked detail here is that iOS 7 was the first OS designed by Jony Ive. iOS 7 was the beginning of the iPhone design we all know today, as icons transitioned from an outdated realistic feel to a more simplified and cleaner look. The design switch is subtle, yet more noticeable the more you look at it side by side.

The company is also expected to align the naming of the OSs. Currently, we have iOS 18, iPadOS 18, macOS 15, tvOS 18, watchOS 11, and visionOS 2. This will be updated to reflect the coming year, so the next generation of OS will be: iOS 26, iPadOS 26, macOS 26, tvOS 26, watchOS 26, and visionOS 26. My take: Thank you, Apple.

2. Open More Models to Developers

Apple is expected to open its in-house AI models to developers through a new software development kit (SDK). Previously, developers could only integrate Apple Intelligence features, such as AI writing tools and the Image Playground, into their apps. With this new initiative, developers will gain access to Apple’s proprietary models, enabling them to create more sophisticated and customized AI functionalities within their applications. Initially, access will be limited to the smaller models that run locally on devices, with cloud-based model access planned for a later phase.

What jumps out to me is that Apple has a more advanced cloud-based LLM that is rumored to be 150 billion parameters. The more parameters, the bigger the model’s “brain.” This would be comparable to ChatGPT-4o, which is impressive and would mark about a 50x increase in model size compared to the small model Apple currently has running on device.

3. AI-Powered Battery Management

Gurman reported that Apple is developing an AI-powered battery management feature for iOS 26 aimed at enhancing iPhone battery life. The system will analyze individual user behavior and device usage patterns to optimize power consumption. My take: it’s a snoozer. This is another Apple Intelligence feature that isn’t the “AI” users are expecting.

Long Shot Announcements

This year will lack the wow factor we saw the past two years with Vision Pro in 2023 and Apple Intelligence in 2024. While we’re not planning for those kinds of announcements this year, it’s worth noting the potential long shot announcements:

• AirPods with cameras
  Probability: 15%. AirPods with cameras, or a Watch featuring a camera, is on the product roadmap. The question is whether we’ll see it this year or next. I put the odds for a September announcement at 25%, and 75% by fall of 2026.

• The next-in-line Vision Pro
  Probability: 10%. A slimmed-down Vision Pro, essentially in the same form factor as the first iteration. This would be a step in the right direction but still would not solve the product’s core challenge: consumers don’t want to wear a heavy headset.

• Apple unveils its first XR smart glasses to compete with Meta and Google
  Probability: 5%. This would be huge news, given it would show that Apple knows they need to shift the focus away from Vision Pro to a more consumer-friendly wearable.

Reading between the lines of management commentary, the focus has shifted away from Vision Pro.

Apple’s focus on Vision Pro has noticeably faded. Since Vision Pro’s release, we examined the past eight quarters of earnings calls. In the first six quarters of Vision Pro, Tim Cook kicked off the earnings call talking about Vision Pro within the first three paragraphs on average, including the welcoming remarks. In the last two quarters, Vision Pro has fallen down to the 13th paragraph, the last product highlighted in Apple’s lineup before turning to services. In other words, Vision Pro is no longer being highlighted above other products.

The bottom line is I was wrong on Vision Pro. When it was launched, I predicted there would be a $1,500 version in two years. Today, the product starts at $3,500, essentially unchanged since it went on sale. I now believe consumers are more attracted to less immersive AI smart glasses, which have been a focus for Meta and Google.

Recap: Apple’s Missteps in AI

After years of leaning on on-device machine learning, Apple arrived late to an AI race already dominated by OpenAI, Microsoft, Google, and Meta. Two years ago, Craig Federighi’s team flagged the need for tens of thousands of GPUs, but he delayed spending amid doubts about ROI. When Apple finally moved, then-CFO Luca Maestri approved a slimmed-down budget while Apple’s competitors were ramping up investments. Additionally, Apple’s AI chief, John Giannandrea, questioned consumer demand for chatbots and took a cautious approach to scaling efforts. The company’s AI path forward would be focused on partnering.

This left Apple years behind.

Last year, the company looked to catch up with the help of OpenAI powering Apple Intelligence and a rebuilt Siri. The timing of that product roadmap proved to be too aggressive, with an almost 18-month delay of the full Apple Intelligence feature set, now expected early next year.

The bottom line is Apple seemed to underestimate the AI shift, then over-promised features, and is now racing to catch up. My belief is they have a couple of years to close the gap, and will do just that.

When it comes to spending on AI models and infrastructure over the past two years, Amazon, Google, Meta, and Microsoft have taken the high investment road, and Apple the low investment road.

Looking back to 2021, Amazon, Google, Meta, and Microsoft have all increased capex spending by about 30% each year in the race to build out AI. Meanwhile, Apple’s capex actually decreased by 1%. In terms of absolute dollars, the hyperscalers’ capex spending has on average increased from $30B in CY21 to an estimated $77B in CY25. Apple’s capex went essentially unchanged at around $11B during the same period.

The reason Apple’s investment has been so low is it has rented compute to train its models rather than owning GPU clusters, and has deployed them in narrow, controlled ways, not at hyperscaler scale. This keeps costs down but limits internal progress and control, which explains the reliance on a third party like OpenAI until Apple can develop a model that rivals the best. That approach has saved Apple around $80B over the past two years compared to the other hyperscalers, which is real money even for a $3T market cap company.

The reason why I believe investing more is the right approach centers on the long-term control of features. To have that control, which impacts features and privacy, the company needs to own a best-in-class model. If Apple settles for a good enough third-party AI, it risks slowing integration of AI features and paying a perpetual toll to a third party. Over the past decade, Apple has wanted to take more control, including building its own chips and modems, a prospect that 15 years ago seemed preposterous.

Some will say it’s a fool’s errand for Apple to build its own advanced models and infrastructure. They’re not only too far behind, but their track record around services is mixed. The skeptics would point to Siri struggles over the past ten years along with Maps failing to challenge Google Maps. These stumbles underscore why any Apple-built foundation model must launch at or above the standard set by OpenAI, xAI, Google, and Anthropic. In reality, the best solution for Apple is to acquire a top lab like Anthropic, whose last round was at $60B.

Meanwhile, OpenAI’s Jony Ive–designed “third device” is pitched as a companion to “iPhones and Macs,” not a replacement, which suggests Sam Altman’s framing is an olive branch to Apple. But if that device becomes the preferred AI interface, Apple’s urgency to own a frontier-grade model to power the device grows, given OpenAI will likely put a governor on how Apple can license their tech to power a competitive device. There’s also the fallback position of working more with Google, but that would amount to another tech reset that would push the timing of features back.

My take: Apple needs to step up its capex investment to $20–$25B a year, well above the current $11B run rate and still behind the other mega caps approaching $80B annually.

Apple has more time than many realize, around two years, to navigate to an AI-first world given two factors that lock in their user base:

First, the biggest reason is more than half of Apple’s customers have more than one Apple device. I estimate the average Apple customer has 1.7 devices. The reason why Apple customers have, on average, more devices is the products work together seamlessly, making it easy for consumers to manage multiple devices. Additionally, I believe each customer uses an average of 1.5 services (Music, TV, Cloud Storage, News). The combination of those two factors makes Apple customers extremely loyal.

Second, the AI device competition is also behind the curve at launching “must-have” AI features. There are smartphones and devices with AI capabilities outside of Apple products, but none of these devices have features unique enough yet to get Apple users to switch. Additionally, an AI-first device doesn’t exist yet. Use cases are still in development, and the simple fact is the technology is not there yet. That leads all eyes to the Jony Ive and Sam Altman combination, which should yield an AI-first device that will be unveiled next year and likely ship in 2026. That device will likely set the stage for Apple’s next big device.

Disclaimer