Why did "AI-lagging" Apple make money first in the AI era?

Ask AI · Why is Apple’s lightweight asset strategy safer amid the AI bubble?

Text | Xiao Jing

Editor | Xu Qingyang

Apple’s AI lags behind many major companies, but it is becoming the first to profit from this track.

By 2025, Amazon, Alphabet, Meta, and Microsoft will have invested nearly $700 billion in AI infrastructure. Morgan Stanley predicts Amazon’s free cash flow will turn negative as a result. In the same year, Apple spent almost nothing on large models but earned nearly $900 million in commissions from generative AI applications. According to analytics firm AppMagic, this number is expected to surpass $1 billion this year.

Competitors are burning money training models, while Apple is collecting tolls.

Additionally, when OpenClaw (commonly known as “Lobster”) made local AI agents a real demand, consumers responded by buying Mac Minis. Quiet, low-power, high-memory, this $599 device is becoming the default platform for personal AI.

Apple hasn’t trained a large model but has simultaneously hit two bottlenecks in AI: distribution channels on mobile devices and hardware for desktop operation.

Siri still can’t produce decent text, but AI applications on 2.4 billion iOS devices continuously pay Apple subscription fees of 15% to 30%; developers’ Mac Minis run local intelligent agents day and night, but no money flows to the cloud.

While competitors burn money training models, Apple charges on both ends. The “lagging AI technology” giants may be the most financially secure in this race.

01 Channel is King: The AI Premium of iPhone

By current AI standards, Siri is clearly outdated. It can handle basic tasks like setting alarms but cannot maintain coherent conversations, nor does it have content generation or deep research capabilities. In 2025, Apple AI head John Giannandrea resigned, widely believed to be due to Siri’s lack of breakthroughs.

In January 2026, Apple announced a partnership with Google, with the new Siri integrating Gemini as its underlying technology, seen as a stopgap after self-developed solutions stalled.

But technical shortcomings haven’t weakened Apple’s bargaining power. No matter how advanced OpenAI, Google, Anthropic, or xAI’s chatbots are, the iPhone remains one of the main channels to reach consumers. This means they still need to pay the “Apple tax”: users subscribing to these companies’ AI services via the App Store pay a 30% cut in the first year, then 15% annually, with slight regional variations.

AppMagic’s monitoring data shows this trend’s evolution: in January 2025, Apple’s monthly revenue from generative AI apps was about $35 million; by August, it peaked at $101 million. Although later, due to a decline in ChatGPT downloads, revenue decreased, the total for the year still approached $900 million.

In 2026, Apple is expected to earn nearly $1 billion in commissions from generative AI apps.

Apple’s AI revenue shows a clear head effect. Just OpenAI accounts for nearly three-quarters of Apple’s commissions from generative AI apps. Elon Musk’s xAI’s Grok ranks second, accounting for about 5%.

$1 billion is a small proportion of Apple’s total revenue, but generative AI apps are becoming a new growth point for services. This segment has grown faster than hardware in recent years, with higher profit margins, and has always been a focus for investors.

02 Light Assets vs Heavy Investment: The Diverging Paths of AI

Comparing Apple’s AI revenue within the entire industry reveals its uniqueness.

In 2026, Amazon, Alphabet, Meta, and Microsoft are expected to invest nearly $700 billion in AI infrastructure.

Heavy investments are already squeezing financial space: Morgan Stanley predicts Amazon’s free cash flow will turn negative this year, with a gap estimated at $28 billion by Bank of America. Barclays believes Meta’s free cash flow could drop nearly 90%. In contrast, Apple’s capital expenditure is only a fraction of its competitors’.

Charles Rinehart, Chief Investment Officer at Johnson Asset Management, notes that if Apple can continue to act as a “toll road” provider for AI, it will remain advantageous in the long term without bearing huge capital expenditure pressures.

Wedbush analyst Dan Ives calls this model “invisible AI strategy.” He estimates that just monetizing AI could add up to $1.5 trillion to Apple’s valuation. The core logic is that 2.4 billion active iOS devices form a vast distribution network—an insurmountable channel barrier for any AI company. He predicts that AI monetization could contribute $75 to $100 per share in the coming years.

03 The Unavoidable “Apple Tax”

In theory, apps can guide users to subscribe via official websites to avoid the App Store cut. But Wall Street Journal tests found that although ChatGPT offers a website subscription option, it provides no price incentives, and users have no reason to take the more cumbersome route. Moreover, this option isn’t available in all markets.

Google tries to break through via its own channels. It dominates Android, develops Pixel phones with integrated AI features, but still can’t persuade iPhone users to migrate en masse.

OpenAI’s strategy is more aggressive. It acquired a hardware startup co-founded by former Apple designer Jony Ive, attempting to build its own device entry point. But competing for market share from iPhones and replicating a full device ecosystem is far more difficult than Apple filling its AI gaps.

Elon Musk has also been rumored to consider entering phone manufacturing, but in February 2026, he explicitly denied on X: “We are not developing a phone.”

04 Blocking “Vibe Coding”: New Moves to Protect the Toll Road

While enjoying subscription revenue, Apple is also wary of a new technological trend that could undermine its business model—“vibe coding.”

This emerging development method allows users to generate executable applications directly via natural language commands. Tools like Replit and Vibecode are entering the market with this approach, helping users create web apps that run directly in browsers. These apps bypass the App Store and thus avoid the 30% cut.

Sources say Apple has already taken restrictive measures against such apps. Replit and Vibecode updates have been delayed because their “vibe coding” features are deemed to violate App Store review guideline 2.5.2, which prohibits apps from downloading or executing code that alters their own or other apps’ functions.

Apple’s review stance is that when users generate new apps via Replit and preview them within the app, this involves “executing code that introduces new functions.” Replit argues that the generated code runs in an isolated virtual machine, and the preview is just a web view within the app, similar to clicking a link in X or Facebook to open a webpage. The stalemate has lasted months.

Recent developments suggest the deadlock may be broken. Insiders say Apple plans to approve Replit and Vibecode updates if the developers make concessions. Replit would need to change the preview from an embedded web view to an external browser; Vibecode would be asked to remove the feature that generates dedicated apps for Apple devices.

Not all “vibe coding” apps face resistance. Vercel’s v0 continues to release updates, and apps from Snap and Canva retain similar features, allowing users to generate filters or mini-games with AI. This indicates Apple’s intervention is selective, not a total ban.

Legal experts focused on app ecosystems, like Gene Burrus—who represented Spotify in antitrust disputes against Apple and now advocates for developers—believe Apple remains vigilant against technologies threatening its platform control. The real risk of “vibe coding” apps is that they could foster a web app ecosystem operating outside the App Store.

Additionally, Apple’s own developer tool Xcode is introducing AI coding features, some based on Anthropic’s Claude and OpenAI’s Codex models. Apple prefers developers to continue using Xcode for app development within its ecosystem rather than shifting to cross-platform “vibe coding” tools.

05 Unexpected AI Windfall: OpenClaw Boosts Mac Mini

While the App Store continues to generate revenue, a hardware product is also being propelled by AI technology—closely related to the rise of OpenClaw (commonly known as “Lobster”).

An X user noted that OpenClaw’s popularity reflects changing demands for dedicated hardware. The reason is that truly intelligent software not only consumes computing power but also hardware.

When users follow OpenClaw’s design—staying online, multi-agent collaboration, deeply integrated into workflows—single laptops are no longer sufficient. Quiet, low-power, stackable dedicated devices have become the new choice, with Mac Mini and Mac Studio fitting this niche.

This aligns well with OpenClaw’s core features: local operation, no data upload to the cloud, but high memory consumption. This directly boosts demand for high-memory Mac Minis.

Alex Finn, founder and CEO of Creator Buddy, believes Apple will win the AI race. He argues that when powerful personal AI appears, consumer reactions have already shown the trend: people go out and buy Mac Minis.

“Ordinary users don’t want their personal assistant running in the cloud,” Finn said. “They want to run locally, at home, on their own device, storing their own data. They don’t want tech executives reading their chats with AI girlfriends. They want private, easy-to-use, stylish devices.”

A netizen summarized Apple’s AI strategy succinctly: Apple doesn’t talk about AGI, only silent hardware.

While others debate which company’s chatbot is smarter on social platforms, Apple quietly sells the $599 Mac Mini, enabling every developer to run AI agents locally. Eventually, people wake up to find the world running agents on Mac Minis, and no one remembers who said, “We don’t do AI.”

06 Is the AI Bubble Not Apple’s Problem?

A notable bet on Polymarket asks: Will the AI bubble burst by the end of 2026?

Currently, 17% of bets are on “Yes,” with a trading volume of $2.2 million. Conditions include Nvidia’s stock dropping 50% from its all-time high or OpenAI declaring bankruptcy.

If such scenarios occur, Apple might be the least affected player.

Special contributor: Jin Lu