Apple will spend about $14 billion on capital projects this year. Its four largest rivals will spend close to $700 billion. On Monday, investors decided Apple picked the better number.
Apple closed Monday, July 27, at roughly $4.94 trillion, passing Nvidia at about $4.77 trillion after Nvidia shares fell close to 5%. Apple stock has climbed about 24% in 2026 while the company spends roughly 3% of its revenue on capital expenditure, against 46% at Alphabet and 54% at Meta. The most valuable company in the world is the one that refused to buy the AI stack.
For four years, Wall Street treated Apple as the megacap that missed the moment. Siri stayed broken. Apple Intelligence slipped. Every rival announced another data center, another gigawatt, another hundred billion in committed spend, and Apple announced buybacks. In 2024 the company lost the top spot in the stock market for the first time in more than a decade. Analysts asked Tim Cook when Apple would get serious about artificial intelligence.
Apple never got serious in the way those analysts meant. It bought the intelligence instead of building it, kept its balance sheet clean, and let Google absorb the capital risk. That decision now carries a $4.9 trillion price tag.
What Happened
Nvidia shares dropped almost 5% on Monday as investors reassessed the cost of the AI buildout, cutting the chipmaker’s market value to roughly $4.77 trillion. Apple rose about 1% to close near $4.94 trillion, reclaiming the title of most valuable public company for the first time since April 2025.
The daily arithmetic favors Nvidia’s decline over Apple’s advance, and the two stocks have swapped the crown more than once this month. Look at the year instead of the day and the picture holds up. Apple has gained about 24% in 2026. Nvidia has gained roughly 6%. Tesla, Microsoft, and Meta are all down. The iShares Semiconductor ETF lost 14% over the past month as money rotated out of AI hardware.
Daniel Newman of Futurum Group described Apple to the New York Times as a flight to safety, an AI-adjacent holding that trades more like an index than a bet. That reading explains Monday. It does not explain a 24% run in a year when the rest of the Magnificent Seven struggled.
The Backstory
Apple’s absence from the AI arms race started as a failure rather than a strategy. The company tried to rebuild Siri and thread generative features through its devices starting in 2024. Those efforts ran into delays and quality problems, and Apple postponed the launch. Internal evaluations reportedly showed Siri mishandling a third of complex requests. Twelve to fifteen senior AI researchers left for Meta, OpenAI, and Anthropic.
Apple solved the problem by writing a check to its biggest rival. In January 2026 it signed a multi-year agreement with Google, reported at roughly $1 billion a year, to run a rebuilt Siri on a custom Gemini model. Apple showed the result at WWDC in June and said it would ship this year. Craig Federighi used the keynote to take a shot at the rest of Silicon Valley, describing competitors as racing forward in pursuit of AI for its own sake without regard for the people it serves.
Set that $1 billion against the $20 billion Google pays Apple each year to remain the default iPhone search engine. Apple is buying frontier AI for 5% of what one distribution deal already pays it.
The Plan
Apple’s fiscal 2025 capital expenditure came to $12.72 billion on $416.2 billion of revenue. Analysts expect $13 billion to $14 billion for fiscal 2026. Over the same stretch, Amazon guided to about $200 billion, Alphabet raised its 2026 range to $195 billion to $205 billion, Microsoft set roughly $190 billion, and Meta lifted its range to $125 billion to $145 billion.
John Ternus takes over as chief executive from Tim Cook, and he told analysts in April that he plans to hold the financial discipline that defined Cook’s tenure. Apple returned $91 billion to shareholders through buybacks in fiscal 2025 while its peers issued bonds to fund servers. Meta priced $30 billion of investment-grade debt in October 2025. Alphabet reported negative free cash flow of $5.9 billion in its most recent quarter after a $44.9 billion capex quarter, a shift we covered in our breakdown of Google’s $514 billion cloud backlog.
The Business Model Angle
The AI buildout converted the most profitable software businesses on earth into capital-intensive infrastructure businesses. Meta, Microsoft, and Alphabet now run capital intensity in the range utilities and telecom operators live in. Oracle sits at 86% of revenue.
Apple sits at 3%. That single ratio explains the repricing better than any narrative about safety.
Owning the compute layer looks like vertical integration, and Apple has spent forty years teaching the market that vertical integration wins. The Apple business model rests on controlling silicon, the operating system, and the customer relationship. Apple still does not own a fab. It designs the M-series and A-series chips and hands manufacturing to TSMC, capturing the design margin while someone else carries the $20 billion plants and the cyclical risk.
Apple is running the same play on intelligence. Own the layer with pricing power, which is the device, the operating system, and the two billion active users. Rent the layer without it. Frontier models keep getting cheaper and keep converging on similar capability, which turns them into a purchased input rather than a defensible asset. Apple can swap Google for OpenAI or for its own model without touching its distribution. The reverse is not true: Google cannot swap out the iPhone.
That is the founder-level lesson buried in a market cap headline. Vertical integration pays when the layer you absorb has durable pricing power. When a layer commoditizes, owning it converts your income statement into a depreciation schedule. We made the same argument about the whole stack in our piece on margin migration in the AI economy, and Apple is the cleanest live test of it.
The Risk
Renting has a bill attached, and it arrives later.
Apple does not control the roadmap, the pricing, or the cost curve of the input that will define its most important interface. Google can raise the price at renewal. Google can also ship the better assistant on Android first and keep the frontier model a quarter ahead of the one it licenses out. Sundar Pichai has already called Google Apple’s preferred cloud provider, which is a comfortable phrase for Google and an awkward one for a company that sells privacy.
The deeper exposure is interface risk. Apple’s moat is distribution: it sits between two billion users and everything they do. If the assistant becomes the interface, and the assistant thinks with someone else’s model, Apple’s device turns into glass and the value accrues to whoever supplies the reasoning. The Apple SWOT analysis treats a capable Gemini-powered Siri as an opportunity. It is also the mechanism by which Apple could hand its rival a seat inside its own ecosystem.
Regulators are watching the same seam. The $20 billion search deal already sits at the center of the DOJ case against Google. Adding an AI licensing agreement on top gives the government a second exhibit for structural separation, and Apple books that search payment as near-pure profit.
There is also the possibility that the market is wrong right now and correct four years ago. If the hyperscaler buildout produces durable advantage, Apple bought a rental at the exact moment the owners started collecting rent. Nvidia’s business model has not broken. Its stock rotated.
Quick Questions
Why did Apple pass Nvidia on July 27, 2026? Nvidia fell close to 5% on concerns about AI infrastructure costs, cutting its value to about $4.77 trillion, while Apple rose about 1% to roughly $4.94 trillion.
How much is Apple spending on AI compared with rivals? Apple guides to roughly $13 billion to $14 billion of capital expenditure in 2026. Amazon, Alphabet, Microsoft, and Meta together plan $650 billion to $700 billion.
Is Apple building its own AI models? Apple runs smaller models on device and through Private Cloud Compute, and it licenses Google’s Gemini for the rebuilt Siri under a deal reported at about $1 billion a year. It continues developing a larger in-house model.
What does Apple pay Google, and what does Google pay Apple? Apple pays roughly $1 billion a year for Gemini. Google pays Apple roughly $20 billion a year to stay the default iPhone search engine.
Will Apple keep the crown? Apple and Nvidia have traded the top spot several times in July 2026, and both sit within a few percentage points of each other. The gap closes on any single trading day.
The Business Model Analyst Take
The market spent four years pricing Apple’s AI absence as incompetence and now prices the same absence as discipline. Nothing about Apple’s strategy changed. Investors changed what they believe about the return on a trillion dollars of GPUs.
Apple made the least fashionable choice available to a technology company in 2026. It treated intelligence as a component to be sourced rather than a capability to be owned, the same way it treats displays, modems, and fabrication. Whether that looks like genius or complacency in 2030 depends on one question: does frontier AI commoditize like memory chips, or consolidate like cloud infrastructure? Apple has bet roughly $14 billion a year on the first answer. Four rivals have bet close to $700 billion on the second.
One of those bets is cheap to unwind.
