Apple Is Winning the AI Race by Not Running It

During the June sell-off that erased $2.3 trillion from the Magnificent Seven, one set of numbers contained the market's entire verdict on who profits from artificial intelligence. Apple fell 8%. Microsoft fell 20%. Nvidia fell 13%. ^[1] On July 17, Apple passed Nvidia to become the world's most valuable company again — not by surging, but by falling less. ^[2] The divergence is not a fluke of a bad week. It is the market pricing two incompatible models of AI, and it has chosen one. On one side sit the hyperscalers. Microsoft, Alphabet, Amazon, and Meta have collectively committed up to $725 billion in capital expenditure on AI infrastructure in 2026. ^[3] Their stock prices are flat or falling. Nvidia's price-to-earnings ratio has compressed to 33, Microsoft's to 23 — both multi-year lows — even as Nvidia reported 85% revenue growth and Microsoft 18%. ^[4] The market is discounting AI infrastructure revenue as potentially unsustainable. Alphabet raised up to $84.75 billion in equity to fund its build-out, potentially the largest such deal in history, with capex possibly reaching $300 billion in 2027. ^[5] Apollo chief economist Torsten Sløk put the question the market is now asking aloud.

But what if the payoff takes longer than consensus assumes? — Torsten Sløk

The hyperscalers are borrowing against a payoff the market no longer assumes will arrive on schedule. Gross leverage ratios across the group have doubled from 0.9x to 1.8x in two quarters, and $800 billion in additional U.S. credit is funding the build-out. ^[6] Tom Essaye of Sevens Report Research drew the parallel to an earlier bubble.

To be fair, this fear has been around for several months, and it isn't appearing yet. However, it's not without precedent because this is exactly how the dotcom bubble burst. — Tom Essaye

Apple is absent from the capex list. It is not building AI data centers. It is not issuing equity to fund them. And the analysts who cover the company have drawn the conclusion explicitly.

Apple is less exposed to capex intensity and better positioned to monetise AI via services, ecosystem lock-in and hardware upgrades. — Toni Meadows

That re-rating is the mechanism beneath the valuation crossover. Apple monetizes AI through three channels, none of which require owning the infrastructure. The first is hardware pricing. In late June, Apple raised prices across its lineup by $100 to $300, with some products exceeding 50%, citing a global memory crisis caused by AI data centers diverting RAM from consumer electronics. ^[7] Tim Cook was blunt.

Price increases are unavoidable. — Tim Cook

On July 18, Apple raised iPhone prices in Japan by up to 11%, extending the pattern. ^[8] Where the hyperscalers absorb AI infrastructure costs on their own balance sheets, Apple passes them through to consumers as a premium feature — and the consumer pays. The second is the App Store. Apple's platform facilitated $1.4 trillion in developer billings in 2025, reaching 850 million weekly users across 175 countries. ^[9] AI-focused app billings grew fourfold, and roughly 40 of the top 100 apps now feature consumer-facing AI built with tools from Anthropic, Stability AI, Midjourney, and Apple's own technology. Apple takes its platform cut from every transaction. It does not pay for the compute that powers those apps; the developers do. The AI boom flows through Apple's pipes, and Apple collects the toll. The third is the purest expression of the model. Apple asked Google to install dedicated servers inside Apple's own data centers to power next-generation Siri, because Apple's server capacity is, by its own assessment, underpowered for AI workloads. ^[10] Apple pays Google roughly a billion dollars a year for the arrangement and is now considering a paid subscription tier for the advanced features those Google servers power, bundling resource-intensive tools like image generation and complex conversation into Apple One or iCloud+. ^[11] The arithmetic is simple: Apple buys AI compute at wholesale from a rival, wraps it in Apple hardware and software, and sells it at retail to consumers who already live inside its ecosystem. It is integration arbitrage — capturing the margin between the cost of AI and the price of AI as a consumer luxury. The financial engine beneath all three channels is not AI infrastructure returns. It is services and hardware pricing. Apple posted a record $111.2 billion in quarterly revenue in the second quarter, up 17% year over year, including $31 billion in services, and authorized a $100 billion share repurchase. ^[12] The hyperscaler model requires capex to be earned back through AI revenue that may take years to materialize. Apple's model earns now. The model has a coherence the hyperscalers do not have, but it is not invulnerable. The vulnerabilities are specific and acknowledged. Apple has withheld Siri AI from iPhones and iPads in the European Union, citing the Digital Markets Act's interoperability requirements, and in China, citing regulatory requirements, with no timeline for release in either market. ^[13] EU regulators have described the decision differently.

EU law is non-negotiable — Thomas Regnier

Apple's integration advantage is geographically constrained — roughly a third of iPhone users are in markets where the company has chosen not to deploy its AI. The Department of Justice is in active settlement discussions with Apple over a 2024 antitrust lawsuit alleging smartphone monopoly and suppression of competition across digital wallets, cloud streaming, messaging, and smartwatches. ^[14] The ecosystem lock-in that underpins Apple's entire AI monetization strategy — the reason consumers pay the premium, the reason developers build for the platform, the reason the App Store toll works — is under legal attack. Apple has already made concessions, opening NFC and supporting RCS, but it still restricts Apple Watch interoperability. And then there is the question the market has not yet answered. As Mark Gurman has observed, Apple has not yet proven its AI is worth using, let alone worth paying for. The market is pricing integration. But integration is only as durable as the ecosystem it sits on, and only as valuable as the AI it integrates.