The acquisition everyone misread
In June 2025, Bloomberg reported that Apple executives — M&A chief Adrian Perica and services chief Eddy Cue among them — had held internal talks about acquiring Perplexity, the AI startup last valued at roughly $20 billion. The framing was immediate and unanimous: Apple needs a search engine. The ~$20 billion-a-year arrangement that makes Google the default on iPhone is under antitrust threat, the reasoning went, so Apple is shopping for a replacement.
That read isn't wrong. It's just small. Buying Perplexity to win web search is like buying Beats to sell headphones — technically true, and completely beside the point of what the asset becomes inside Apple. The interesting question isn't what Perplexity does today. It's the one capability Apple has spent years trying and failing to build for itself.
What Apple just proved it can't build
For most of 2025, Apple was developing Project Mulberry: an AI health coach designed, in the reporting's own language, to “replicate” a human doctor. It would read the signals coming off the Watch and iPhone — heart rate, sleep, activity, blood oxygen — and turn them into the kind of personalized guidance a physician gives. It was slated to ship as a Health+ service in 2026.
In February 2026, Apple wound it down. Eddy Cue — who had taken over the health division after Jeff Williams retired, and who had also been in the room for the Perplexity talks — decided Mulberry wasn't good enough and shelved the standalone launch, keeping only a scaled-back set of features. The lesson is the most important data point in this entire thesis: Apple owns the body and cannot build the brain. It can ship FDA-cleared hypertension detection, ECG, and blood-oxygen sensing across hundreds of millions of wrists. It cannot, on its own timeline, build the reasoning layer that turns those signals into trustworthy guidance.
That is precisely the half a company like Perplexity already is.
The two halves of a doctor
Strip a doctor down and you get two distinct machines. The first is a sensor— it observes what's actually happening inside your body over time. The second is a reasoner — it interprets those observations against everything medicine knows, and tells you what to do about it.
Apple has built the most capable consumer sensor ever deployed. As of September 2025, the Apple Watch detects signs of hypertension with FDA clearance, takes an ECG, and tracks blood oxygen, sleep stages, and cardiovascular fitness — continuously, longitudinally, across years of a single person's life. No one else has that. You cannot scrape it, buy it, or back-fill it.
Perplexity is the other half. It is not a chatbot; it is a retrieval-and-citation engine — an answer machine whose entire reason for existing is to ground a response in sources you can check. In every domain that's a feature. In medicine it's the whole game. The reason an AI doctor is hard isn't generating text — it's producing guidance you can trust enough to act on, with provenance attached. Apple has the signal with no answer. Perplexity is the answer engine with no signal. Fused, they are a thing neither could become alone.
Why this becomes the biggest moat ever built
A moat isn't one advantage. It's a stack of advantages that compound and that no single competitor can assemble at once. Four of them converge here.
- The sensor moat.Health data isn't a snapshot — it's a time series. The value is in the years. Apple has a multi-year, multi-signal, FDA-pathway record of your actual physiology that no rival can reconstruct after the fact. Google can index the web; it cannot index your heart rhythm since 2022.
- The trust moat.The binding constraint on AI in health is liability and trust, not capability. Apple's privacy brand plus on-device processing plus Perplexity's citation-first architecture is the only combination that produces guidance a person — and eventually a regulator — will treat as credible.
- The distribution moat.Apple has more than two billion active devices. The doctor doesn't need to be downloaded, marketed, or acquired user by user. It ships as an OS update to a population larger than any health system on earth.
- The flywheel.More users generate more biometric signal, which sharpens the model, which produces better guidance, which pulls in more users. And because the input is your body rather than your behavior, the flywheel can't be cloned by a competitor with a bigger ad budget.
Any one of these is defensible. All four, vertically integrated from the silicon to the sensor to the answer to the lock screen, is not a moat. It's a category no one else can enter.
The forecast
Here's the call. The “Apple needs a search engine” framing will keep dominating the coverage, because search is the legible story and the body is the counterintuitive one. But the strategic logic points one direction: Apple acquires Perplexity (or builds the equivalent by acqui-hiring its way there), and the answer engine gets pointed inward — not at the open web, but at the wrist.
By 2027, the product that matters isn't Apple-branded search. It's a grounded, private, always-on clinician that lives where your health data already lives — turning a decade of passive sensing into active guidance. The hardware was the hard part, and Apple already shipped it. Mulberry proved the reasoning layer is the part it can't build alone. Perplexity is the cheapest, fastest way to buy it.
Everyone's watching Apple shop for a way to answer the web. The real move is buying the only thing that can finally answer the body.
— EV