Yesterday at WWDC 2026, Apple finally pulled the curtain back on Siri AI. The demos looked smooth enough: a conversational assistant that reads your emails, recognizes what’s on your screen, and actually finishes multi-step tasks without forgetting what you asked three seconds ago. But watch the stock ticker from yesterday afternoon. Apple shed roughly 5% of its value, wiping out something like $230 billion in market cap within hours of the reveal. When a company announces its most important software overhaul in years and investors treat it like a shrug, you have to ask what the room full of developers and journalists missed.
The truth is that Siri AI isn’t a moonshot. It’s a carefully hedged bet.
What Apple Actually Shipped
On paper, Siri AI is the assistant Apple promised back in 2024 before the delays started piling up. It taps into your messages, photos, and mail for what Apple calls “deep personal context.” It can see what’s on your display and act across apps. There’s even a dedicated chat interface you can search, which finally makes Siri feel less like a voice-only afterthought and more like the conversational layer iOS has needed for a decade. Apple’s official line calls it “an entirely new version of Siri,” unlocked by next-generation Apple Intelligence.
But here’s the detail that changes the narrative. Apple isn’t powering this alone. The core intelligence layer runs on custom Google Gemini models, routed through Apple’s Private Cloud Compute and on-device processing. That partnership is a quiet admission that Apple’s internal AI efforts couldn’t carry the load. After years of marketing privacy-first, independent machine learning, the company is effectively renting brainpower from the same competitor that makes the Pixel. We just covered Google’s latest model drops last week, and now Apple’s most visible consumer feature is built on that same family of models.
And then there’s the beta label. Apple is calling this a preview, with developer seeds going out immediately for iOS 27, iPadOS 27, macOS 27, and visionOS 27, but general availability won’t arrive until later this year, and only in English at first. If you’re in the EU and carrying an iPhone, you’re out entirely. The DMA has locked Siri AI out of iOS and iPadOS in Europe, and China isn’t getting it either. MacRumors confirmed the regional block yesterday, and there’s no timeline for either market.
The Parts Nobody’s Talking About
While the keynote focused on dinner-party planning and email rewrites, the developer forums tell a different story. Apple is deprecating SiriKit and making App Intents mandatory for any app that wants to plug into Siri AI. That’s a forced migration, and it means thousands of developers have homework to do before this assistant can actually control the apps people use every day. The transition might reshape the iOS ecosystem more than any consumer demo, but it got a single slide in a two-hour presentation.
I spent yesterday evening digging through post-keynote threads and early beta installs. The same fatigue keeps showing up. Dig through the Reddit megathreads and you’ll see experienced users pointing out that the underlying architecture looks suspiciously like repackaged Semantic Index work from last year’s aborted Apple Intelligence rollout. Others running the first developer build immediately noticed that Gemini on a Pixel still feels snappier and more accurate for open-ended questions. One prominent tech voice put it bluntly:
There’s also a privacy tension brewing under the surface. Siri AI needs enormous amounts of personal context to work, and while Apple is leaning on on-device processing and its Private Cloud Compute architecture, the beta period raises questions about how user content is handled while the models are still being tuned. Apple didn’t offer clear answers on telemetry during the beta phase, and I’ve already seen developers asking whether their test queries are being logged for model improvement.
Oddly enough, the most underreported piece of the keynote might be the new child-safety and parental-control tools that shipped alongside the AI overhaul. In the rush to cover Siri, most outlets skipped past features that could actually move hardware for families this holiday season. With John Ternus pushing faster product cycles, Apple needs software differentiators that sell today, not just promises for next spring.
And that stock drop? It makes sense. Investors didn’t see a leapfrog moment. They saw Apple catching up, wrapping that catch-up in a beta flag to avoid another public delay, and outsourcing the hardest AI work to Mountain View. It’s pragmatic, but pragmatism doesn’t send a stock to the moon.
I’ll be watching the developer beta over the next month. If Siri AI can actually execute complex, cross-app workflows without hallucinating contact names or misreading calendar invites, it might earn the “entirely new” label Apple gave it. But right now, it feels less like a revolution and more like a very expensive, very cautious patch on a reputation that was starting to fray. The real test isn’t whether Siri can plan a dinner party on stage. It’s whether it can set a timer without apologizing.
