Apple logo is seen on a computerin Krakow, Poland on February 18th, 2026. (Photo by Beata Zawrzel/NurPhoto via Getty Images)NurPhoto via Getty ImagesThe fiscal year 2026 signifies a significant divide in the AI aspirations of Big Tech. Amazon is investing $200 billion in capex – mainly for AI infrastructure – while Alphabet (Google) aims for $175–185 billion, nearly double last year’s figure, and Meta Platforms has plans for $115 to $135 billion. Along with Microsoft (projected to reach around $145 billion), these hyperscalers are pledging nearly $700 billion in a frenzied effort to establish extensive, gigawatt-scale training clusters and data centers.Then there’s Apple (NASDAQ:AAPL): forecasting slightly over $14 billion in capex – essentially unchanged year-over-year, and a small fraction of what its competitors are committing. In fact, during Q1 FY’26, Apple’s capital expenditures actually decreased year-over-year. As the world enthusiastically embraces AI, what accounts for Apple’s markedly different strategy?Apple AI BetApple has chosen a distinct path in AI. Rather than investing tens of billions into the development and training of extensive foundational models from the ground up, it prioritizes secure, tightly integrated inference that depends partly on processing on-device. For the most compute-heavy model training, Apple has sought external solutions. Recently, it entered into an agreement to license Google’s Gemini, reportedly costing only $1 billion annually (approximately 1% of annual free cash flows). This arrangement grants Apple access to a cutting-edge model at a fraction of the expenditure that its competitors commit to developing and maintaining their own systems.This strategy is partly driven by Apple’s business model. Instead of pursuing direct AI revenues through cloud subscriptions, consumption, or ad-enhanced AI features, Apple incorporates Apple Intelligence as complimentary, privacy-focused upgrades to boost sales of high-end hardware (new iPhones and Macs) while speeding up its digital services expansion. Certainly, Apple has not completely abandoned the development of internal models. It is actively working on its own in-house foundational and task-specific models, but on a smaller scale and with a product-centric focus rather than aiming to create the largest and most intricate models.The potential risk is clear. If the most potent models remain scarce and tightly controlled, owning them might become more advantageous than merely renting access. In such a scenario, Apple’s cautious approach could appear less like prudence and more like insufficient investment in the core technology of the upcoming era.MORE FOR YOUSome Big AdvantagesCapital Efficiency and Margin Preservation: The main advantage lies in preserving free cash flow. By steering clear of acquiring expensive GPUs like Nvidia H100/Blackwell clusters, Apple sidesteps the substantial depreciation expenses currently impacting the balance sheets of companies such as Amazon and Google. At this point, we would concur that licensing a leading model for $1 billion annually is financially more advantageous than spreading the costs of a $100 billion infrastructure build-out over time, especially considering the still uncertain financial returns of AI features. This strategy enables Apple to maintain gross margins comfortably above 40% while its competitors struggle with increasing depreciation and energy expenses.Distributed Cost Structure: Apple’s vertical integration of silicon permits it to transfer some of the computational costs to the end-user. By handling certain inference tasks locally on the iPhone’s Neural Engine (NPU), Apple ensures that the energy consumption and hardware wear linked to daily AI interactions are the responsibility of the consumer's device rather than Apple’s cloud infrastructure. This creates a distributed computing network of 2.2 billion active devices, providing an inference capacity scale that centralized clouds cannot compete with in terms of cost efficiency.Commoditization of the Model Layer: Strategic outsourcing positions the foundational model as a commodity utility rather than a factor for differentiation. By viewing the model provider (Google) as a backend vendor comparably to a cloud storage service, Apple maintains the capacity to change providers or adopt a multi-model strategy. If a more advanced model comes from OpenAI or Anthropic, Apple’s architecture allows for easy integration without the sunk costs of prior internal training initiatives.Some Big Risks, TooIt’s important to acknowledge that this tactic entails genuine risks. Licensing Gemini establishes a significant dependency on Alphabet for the core intelligence component of the operating system. Unlike a standard component supplier, Google governs the reasoning engine that facilitates the interaction. This gives Alphabet significant leverage over pricing, feature access, and timeframes. As Gemini becomes further entrenched in iOS, transitioning to another provider could become increasingly challenging for Apple.There is also a potential brand risk. Private Cloud Compute is engineered to anonymize data, yet complex queries still necessitate a transfer to third-party models. Even if the arrangements are technically secure, the assumption that sensitive inquiries leave Apple’s infrastructure might undermine one of its strongest competitive edges, which is privacy.Ultimately, Apple is implicitly wagering that foundational models will become commoditized. If cutting-edge models continue to be rare and strategically governed, then possessing large-scale training infrastructure might turn out to be more defensible than licensing access. Should models fail to become widely utilized, Apple’s asset-light methodology could be perceived more as underinvestment instead of strategic prudence.Are you uncertain if Apple is the right stock for your portfolio? Consider a multi-asset portfolio strategy to safeguard and expand your wealth.How To Build A Resilient Wealth StrategyIn a climate of fear and greed, individual stock selections expose you to unnecessary risks. A holistic wealth strategy positions you to effectively manage risk while taking advantage of global trends.What if you leveraged the ongoing commodity super cycle? Could a portfolio allocating 10% to commodities, 10% to gold, and 2% to crypto, in addition to equities, deliver higher returns over the next 1-3 years? We’ve analyzed the figures. Our wealth management partner skillfully manages these types of intricate multi-asset strategies, combining real assets with high-performing equity components like the Trefis High Quality Portfolio, which has surpassed 105% in returns since inception.