On February 8, 2026, Sherwood News reported that Apple's Q4 capital expenditure was $2.37 billion — down 19% year over year. It was the only Big Tech company whose infrastructure spending declined. Three days earlier, Alphabet had guided $175-185 billion in 2026 capex, Amazon had guided $200 billion, and Meta had guided $115-135 billion. Combined: $490-520 billion. The same day, The Information reported that the spending would "all but wipe out free cash flow" for all three companies. Apple — generating more revenue than any of them — was moving in the opposite direction.
The Ratios
Apple's Q1 revenue was $143.76 billion. Its quarterly capex was $2.37 billion. That's a capex-to-revenue ratio of 1.65%.
Compare:
- Alphabet: $175-185 billion capex on $400 billion revenue — 44-46%
- Meta: $115-135 billion capex on roughly $165 billion revenue — 70-82%
- Amazon: $200 billion capex on roughly $200 billion revenue — approximately 100%
- Apple: roughly $9.5 billion annualized capex on roughly $575 billion revenue — 1.65%
Meta plans to spend 70-82% of its revenue on infrastructure. Apple spends less than 2%. The difference is not incremental. It reflects two fundamentally different theories about what the AI era requires.
The Buy Strategy
Apple isn't abstaining from AI. It's buying it instead of building it. On January 13, Apple signed a multiyear deal with Google to power Siri with Gemini models. The Financial Times reported the deal was a cloud contract worth several billion dollars — and that OpenAI had declined to be Apple's custom model provider. On January 31, Bloomberg's Mark Gurman said Apple's internal development "runs on Anthropic at this point" and that Apple had wanted to rebuild Siri around Claude, but Anthropic "wanted a ton of money."
Apple shopped for intelligence the way it shops for components: solicit bids, negotiate hard, pick the best deal. OpenAI said no. Anthropic priced itself out. Google said yes. The result is that Siri — the product that was supposed to define Apple's AI era — will run on someone else's models.
On February 7, Bloomberg reported that Apple plans to allow third-party AI chatbots in CarPlay. Apple is not just buying intelligence for its own products. It's opening the platform so others can bring theirs.
The Ingredients Question
Apple's executives are not oblivious to what the buy strategy implies. On February 2, Gurman reported that Apple's leadership was questioning whether the company has "the right ingredients to win the AI era" — and that a blockbuster quarter "shouldn't give cover to avoid an AI reckoning."
The evidence suggested they were right to worry. On January 31, Bloomberg reported Apple had lost at least four more AI researchers in recent weeks — two to Meta, one to DeepMind — plus a top Siri executive. On February 6, Apple wound down plans for an AI-based virtual health coach. On February 2, the Wall Street Journal reported that AI companies were outbidding Apple in the electronics supply chain — for chips, memory, and glass fiber — meaning "parts for iPhones to cost more."
The capex race Apple isn't participating in is still raising Apple's costs. The $490-520 billion flowing into data centers competes for the same physical materials that go into iPhones. Apple's restraint on AI infrastructure doesn't insulate it from the infrastructure arms race.
The Other Side of the Trade
Apple's buy strategy only works if there are sellers. On February 8, the Financial Times reported that Anthropic's guidance to investors claims annualized revenue will exceed $30 billion by the end of 2026. In the Feb 1 a16z survey, 75% of Anthropic's enterprise customers were running the latest models. Microsoft was spending nearly $500 million per year on Anthropic's AI. Apple was running its internal development on it.
Anthropic's revenue acceleration and Apple's capex decline are two sides of the same trade. One company sells intelligence. The other buys it. The seller's revenue grows. The buyer's infrastructure spending shrinks. If AI intelligence is becoming a service — like cloud computing before it — then Apple's strategy is not a concession. It's the natural position for a company that has always controlled the integration layer rather than the infrastructure layer.
What the Fraction Reveals
The February 5 capex announcements asked a question: if building AI requires spending like an infrastructure company, will the returns look like a software company's? Apple's answer on February 8 was to reject the premise. Don't spend like an infrastructure company. License the intelligence. Design the silicon. Keep the margins.
On the same day Apple's capex decline was reported, the Wall Street Journal documented who was capturing the half trillion: Corning's fiber-optic business was booming on a $6 billion Meta deal, data center contractors like Sterling Infrastructure and Comfort Systems were surging, and a single Japanese company called Nittobo dominated the supply of T-glass — ultrathin glass sheets critical to advanced chips — with no plans to add capacity for months. The builders were making the supply chain rich.
Apple is betting that intelligence commoditizes — that models converge, switching costs stay low, and the value accrues to the company that integrates AI into two billion devices, not the one that builds the biggest data center. The dissolution of exclusive partnerships supports this bet. If Microsoft can put Claude next to Codex, if Amazon can court OpenAI while backing Anthropic, then the models are interchangeable. And if the models are interchangeable, the company with the distribution doesn't need to build the factory.
The risk is the opposite scenario: that one lab achieves something that can't be licensed, that intelligence doesn't commoditize, that the $490-520 billion buys a moat no fraction can cross. Apple's own executives worry about this. But on February 8, the market's verdict was clear. The builders faced a free cash flow wipeout. The supply chain hit all-time highs. And Apple — spending 1.65% of revenue on infrastructure while its peers spent 44% to 100% — was selling a cosmic orange iPhone in China.