Mark Gurman on Apple's secret Gemini deal, Siri's brand damage, and the race for smart glasses

Summary

Apple's AI strategy is entering a critical phase, anchored by a secret deal to embed Google's Gemini model as the under-the-hood engine powering a revamped Siri, expected to roll out in March or April 2026. The arrangement is structured as a straightforward supplier relationship — Apple pays Google for the model, with no revenue share — and is deliberately white-labeled. Without Mark Gurman's reporting, the Google involvement would likely remain unknown, and Apple has no plans to advertise it. Gurman expects Apple to eventually replace the Gemini layer with a proprietary model, making this a transitional rather than permanent architecture.

The competitive process to win the Siri contract was intense. OpenAI was eliminated early. Anthropic was the frontrunner until roughly two months before the deal closed, at which point financial terms pushed Apple toward Google. OpenAI had demanded an investment stake in addition to a starting fee of at least $1 billion in year one, doubling annually for three years — terms Apple rejected. Anthropic and Google's models were judged roughly equivalent in quality, making cost and commercial structure the deciding factor.

The new Siri will offer three announced capabilities — on-screen awareness, personal context search, and related features — plus an unannounced AI web search product positioned as a direct competitor to ChatGPT and Perplexity. That search layer will be Apple-built and will eventually extend into Safari and across iOS. The ChatGPT integration already in Siri is not being removed. Longer term, Gurman is bullish on Siri becoming a commerce layer — enabling Uber bookings, food delivery, and similar transactions — once reliability reaches the threshold where users can actually trust it to execute. At that point, revenue-share agreements with app partners become viable. Currently, Siri's decade-old app extension framework for those functions is, in his words, "unbelievably terrible and unreliable."

The Siri brand itself may be the deeper problem. Fifteen years of underdelivering have done meaningful damage, and Gurman argues Apple should consider a name change when the upgraded assistant launches — though he does not expect them to act on it. The gap between Apple's marketing promises on AI and actual product performance has cost the company credibility in a way that a single good release may not fully repair.

iPhone and Hardware

The iPhone 17 line is set up for a strong upgrade cycle, timed to coincide with the five-year anniversary of the COVID-era device surge. Gurman expects Apple to report a $140 billion quarter when it reports results in early 2026. The key decisions — prioritizing battery life, thermal performance, camera quality, and the return to aluminum from titanium — reflect deliberate market research rather than engineering ambition for its own sake. The aluminum reversal is a functional necessity: as chip performance improves, heat dissipation requirements increase, and titanium simply cannot keep pace.

The iPhone Air is, by Gurman's account, a technology exercise rather than a volume product. Its commercial purpose is to develop the thinner form factors, battery technology, and advanced materials needed for an eventual foldable iPhone — a market where Samsung has a seven-year head start and Google has three years. Apple has not cited the iPhone Air as a key revenue driver, and it does not appear as a meaningful line item in company filings.

Smart Glasses and the Vision Pro

Apple is moving forward with smart glasses, with a no-display first-generation product targeted for 2027, potentially announced before the end of 2026. A display-equipped version follows after that. Gurman also notes Apple is developing cameras embedded in AirPods, creating an internal competition between form factors — though he expects both to ship. The glasses will integrate tightly with iPhone, a deliberate advantage over Meta Ray-Bans, which Apple has actively limited from syncing cleanly with iOS.

Meta deserves significant credit for creating the smart glasses category. Gurman rates the Meta Ray-Ban display glasses as two to three years ahead of anything Apple currently has, and calls the $800 price point well-calibrated for the feature set. He expects Google to enter the space with Samsung and through partnerships with Warby Parker and other eyewear brands within months. The 2026-to-2028 window is where competition in the category intensifies.

The Apple Vision Pro, launched in February 2024 at $3,499, remains the best video-watching experience on the market by Gurman's assessment, but its form factor and weight limit it to a narrow user base. He argues Apple has done a poor job marketing features like Personas and shared viewing experiences, relying on executive interviews rather than TV advertising and hands-on demos.

M&A and the Perplexity Non-Deal

Gurman confirms he broke the story of Apple exploring an acquisition of Perplexity, and says the interest was real — not a leak from Perplexity itself. The strategic logic was defensive: Apple feared the DOJ antitrust case could invalidate its Google search default deal, worth billions annually, and Perplexity represented a plug-and-play replacement that Apple could monetize through its ad network. The deal became less urgent once it was clear the court would not immediately unwind the Google arrangement, giving Apple time to build its own AI search product in-house.

Apple also looked at Mistral, the European LLM developer, though regulatory friction with the EU makes any such acquisition a long shot. The company's largest acquisition to date remains the $3 billion Beats deal in 2014. That acquisition discipline — buying small, buying for talent or capability, avoiding large-multiple bets — has not changed.

OpenAI Hardware Ambitions

Sam Altman's recent comments tying OpenAI's massive compute spending commitments to an upcoming consumer device registered as a signal, but Gurman is skeptical. Every AI-first hardware product launched in the current cycle has underperformed, and the Meta Ray-Bans — the category's most successful entrant — consume a fraction of the compute Altman is referencing. Apple's posture toward OpenAI hardware is that of a fast follower: confident it can replicate any product a competitor brings to market if the concept proves out.