The Bloomberg reports that Apple is exploring an acquisition of Perplexity AI is unsurprising in one sense and deeply revealing in another. On the surface, Apple’s acquisition of a buzzy AI startup fits the recent pattern: Microsoft acquires (sort of) OpenAI, Google leans into DeepMind and Gemini, and Amazon throws money at Anthropic… and Apple? Apple talks up “Apple Intelligence” while running most of it in the cloud and quietly looking for ways to not miss the AI boat entirely. So yes, they’re shopping.
What is Apple buying?
The answer, at least when it comes to Perplexity, is hype and a UI layer sitting precariously atop someone else’s infrastructure. Apple dismissed LLMs as chatbots (!). But if Apple now wants to spend billions on a chatbot with a nice skin on OpenAI or Claude, then the real problem isn’t missing the AI moment, it’s misunderstanding it.
The moats Apple wants but Perplexity doesn’t have
From Ben Thomson’s interview with the CEO of Cursor, the only company showing actual defensibility at the application layer, we outlined a roadmap that looks like this:
-
Step 1: autocomplete with LLM sauce (think of Copilot);
-
Step 2: smarter completions using foundation models, customized to your local context;
-
Step 3: low-latency, context-aware local models (Cursor lives here);
-
Step 4: project-level context baked into prompts and responses;
-
Step 5: full automation with minimal human interaction.
Perplexity isn’t on this path. It’s building a nice product with zero distribution moat (Google), zero model moat (OpenAI), and zero infra moat (Amazon or Microsoft). Apple, by contrast, has three moats already: devices, distribution, and privacy. Buying Perplexity won’t add a fourth, but it might just be a short-term bandage for a long-term clarity problem.
Apple doesn’t need to win the AI race. But it does need to understand the course. Right now, it’s flirting with startups that look good on the track but are running on someone else’s legs.
