Apple Beats NVIDIA with Low‑Cost AI Strategy and $5T Valuation

On July 17th, Apple surpassed NVIDIA to become the most valuable company globally, a feat achieved despite its seemingly minimal investment in Artificial Intelligence (AI). This rise to the top, and becoming the second company to exceed a $5 trillion market capitalization, is particularly noteworthy given that Apple was often criticized for being "behind" in AI, spending only about 2% of its sales on AI compared to the 39% spent by hyperscalers.

Apple's AI Journey: From Skepticism to Dominance

Just a few years ago, Apple's position at the top seemed unlikely. From early 2022 to early 2026, Apple experienced mixed performance, including a nearly 19% drop between April and May of last year. This period was marked by lawsuits, tariffs on Chinese components, and, significantly, criticism regarding its AI products.

Siri, initially envisioned as a revolutionary voice assistant, frequently frustrated users with its limited understanding and inability to perform complex tasks. The promised "AI-powered" Siri, intended for integration with Mail, Calendar, and Messages, faced delays, numerous bugs, and an underwhelming reception, leading to its indefinite postponement.

The introduction of Apple Intelligence at WWDC24 also encountered problems, including further delays, postponed features, and general user dissatisfaction. Apple was criticized for overpromising AI capabilities and promoting features that Apple Intelligence could not yet deliver. This led to the discontinuation of a major TV ad for the new AI Siri. The features were subsequently promoted for the iPhone 17 instead of the iPhone 16.

Adding to these challenges, AI chief John Giannandrea announced his retirement in December 2025 due to concerns about product direction. Siri was then moved under Mike Rockwell, who had launched the Vision Pro, a product that also faced its own set of issues.

Apple's initial AI partnership with ChatGPT eventually dissolved, leading to a lawsuit against OpenAI for trade secret theft. Apple then partnered with Google Gemini, a move seen by some as a "self-own" given Google's existing payment of an estimated $20 billion annually to Apple for being the default Safari search engine. Critics pointed out the vast difference in model parameters, with Google's AI boasting 1.2 trillion parameters compared to Apple Intelligence's 150 million for its cloud-based version.

On February 12, 2026, Apple's stock dropped 5%, wiping out its gains for the year, primarily due to further Siri delays just weeks before a major AI update was expected. At this point, some even suggested changing the "Magnificent 7" to "Mag 6" to exclude Apple, with commentators noting Apple's late entry, reliance on partners, sluggish rollout, and limited features.

The Turnaround: Apple's Unique AI Strategy

Despite these setbacks, Apple has remarkably climbed back to the top, even briefly surpassing NVIDIA. While NVIDIA's dip in July contributed to this, Apple's overall performance in 2026, with a 24% increase against NVIDIA's 4%, and a nearly 60% increase over the trailing 12 months, indicates a more fundamental resurgence.

Apple's success can be attributed to its distinct approach to AI, particularly its capital expenditure (capex) strategy.

Capex and AI Spending

Capital expenditure refers to funds companies spend on acquiring, upgrading, or maintaining long-term physical assets. Apple's capex for fiscal year 2025 was $12.7 billion, with an estimated $11-12.7 billion allocated to AI-related expenses. Projections for fiscal year 2026 show a modest increase to $13-14 billion in AI-related capex.

This is significantly lower than other "Magnificent 7" companies:

• Meta: $72 billion in 2025, with guidance of $130-145 billion for 2026.
• Alphabet: $80 billion in 2025, with guidance of $175-205 billion for 2026.
• Microsoft: $64 billion in 2025, estimated $190 billion for 2026.
• Amazon: $107 billion in 2025, estimated $200 billion for 2026 (mostly for AWS, AI data centers, and NVIDIA GPUs).

Apple's capex is roughly one-tenth of Amazon's and less than one-fourteenth of Amazon's 2026 capex. While other Mag 7 companies collectively spend around $700 billion and allocate an average of 39% of sales to AI, Apple stands apart.

The "Rent, Don't Build" Philosophy

Apple's low capex is largely due to its decision to license Google Gemini rather than building its own large language model (LLM). This strategy significantly reduces costs, as licensing a model for a billion dollars a year is far more economical than investing hundreds of billions in developing one from scratch.

Apple's core AI focus is on privacy and local processing. Its bet is that the most useful AI should run locally on devices like iPhones, iPads, and Macs, rather than in remote data centers. Local AI offers speed, cost-effectiveness, and, crucially, privacy.

For heavier tasks that cannot run locally, Apple Intelligence utilizes "Private Cloud Compute." Apple has owned data centers for over a decade, primarily for iCloud services, Apple Maps, and the App Store. They have refitted some of these for Private Cloud Compute, charging customers for cloud-based AI usage.

However, for the most demanding workloads, Apple does not rely on its own data centers. Instead, it leases compute and cloud capacity from Google Cloud. This means Apple is not ignoring AI but is operating with a different business model: on-device first, cloud second, and minimal ownership of a data center empire. By renting infrastructure, Apple avoids the massive capital expenditures of its competitors.

The Power of the Ecosystem: Monetization and Free Cash Flow

The market's renewed confidence in Apple stems from its control over a crucial layer of the AI ecosystem: the users. Apple boasts 2.5 billion active devices, a massive installed base that provides a unique advantage.

• Monetization through Local AI: With AI running locally on devices, Apple's marginal cost per AI prompt is essentially zero. The silicon, electricity, and depreciation costs are borne by the customer who already purchased the device. This also reduces the need for Apple to own vast GPU clusters, which can rapidly depreciate.
• Tiered AI Access: While local AI has usage limits, iCloud+ subscribers receive enhanced AI access, creating another monetization channel.
• Cost Pass-Through: For the most costly AI uses, particularly those requiring Private Cloud Compute, Apple charges users. When these tasks require Google's licensed services, Apple passes on the cost to the user while retaining a small profit.

This strategy allows Apple to win in almost every scenario. Unlike other AI companies that often lose money on free users and only profit from premium subscribers, Apple's model ensures profitability. Investors are recognizing Apple's industry-leading free cash flow, which is a direct result of its low AI investment and high monetization potential through its vast device ecosystem.

Apple's success highlights a key insight: the AI moat is not solely about owning GPUs or data centers, but about controlling the devices and the user base. Apple's approach, which prioritizes owning the layer with high margins and user behavior, has always been consistent. It predicted the commoditization of AI models, understanding that it doesn't need to build the best model itself, but rather leverage the competition among providers like OpenAI, Google, Anthropic, and Meta to ensure a supply of high-quality, increasingly affordable AI.

While companies like OpenAI commit to massive AI spending, they control a layer of the AI ecosystem that is becoming less valuable as AI models become more accessible and commoditized. Apple, by contrast, controls the end-user experience, a far more valuable position in the evolving AI landscape.