Industry News

Apple Is Paying Google $1 Billion a Year to Rebuild Siri — Here's What's Actually Changing

Q: Why is Apple paying Google to power Siri?

Apple determined that building and running a 1.2 trillion parameter frontier AI model in-house would be prohibitively expensive and time-consuming. Google's Gemini model, combined with Apple's Private Cloud Compute infrastructure, allows Apple to deliver frontier AI capabilities while maintaining its privacy standards. The deal is worth approximately $1 billion per year — cheaper and faster than building a competing model from scratch.

Q: Does Apple's Gemini deal mean Google sees your Siri queries?

No. Apple routes all Siri queries through its Private Cloud Compute (PCC) infrastructure — a system of Apple Silicon servers that processes requests with stateless, ephemeral computation. Google receives only the processed request, not raw user data. Apple states that Google never directly sees what users ask Siri. However, privacy researchers note that the architecture creates a new potential vector that did not exist before the Google partnership.

Q: When does the new Gemini-powered Siri arrive?

The initial rollout is happening in iOS 26.4, beginning in March–April 2026 to all compatible iOS devices. A full 'Siri 2.0' with proactive agentic capabilities is expected with iOS 27 in late 2026. The Gemini model powering the new Siri is internally designated as Apple Foundation Models v10.

Q: What happened to Apple's deal with OpenAI?

The Apple-OpenAI partnership for ChatGPT integration in iOS has been effectively superseded by the Google Gemini deal. Apple used OpenAI as a stopgap while negotiating the longer-term, deeper Gemini integration. ChatGPT will remain available as an opt-in option in Siri settings, but Gemini is now the primary AI engine. Apple's annual $1 billion payment to Google dwarfs any arrangement with OpenAI.

March 30, 2026 · 7 min read · Happycapy Guide

TL;DR

Apple finalized a deal in January 2026 to pay Google approximately $1 billion per year for a custom 1.2 trillion-parameter Gemini model that will power a completely redesigned Siri. The rollout is happening now in iOS 26.4, reaching over 2 billion iOS devices worldwide. The Apple-OpenAI partnership is effectively over. The new Siri hits 92% on complex multi-step tasks (up from 58%) and responds in under 0.5 seconds, with Screen Awareness across all apps.

1.2T

Parameters in Google Gemini model powering Siri

$1B/yr

Apple paying Google annually for the model

2B+

iOS devices receiving the Siri overhaul

92%

Task success rate on complex queries (vs 58% legacy)

What Apple Just Admitted About AI

Apple spent years insisting it could build world-class AI in-house. Its own Apple Intelligence features — launched in late 2024 — were widely criticized as underwhelming compared to ChatGPT and Claude. The company then patched the gap with an OpenAI partnership, making ChatGPT available as an opt-in inside Siri. That was a stopgap.

The real move came in November 2025, when Bloomberg reported that Apple was negotiating with Google for access to a custom 1.2 trillion parameter Gemini model. The deal closed in January 2026 at approximately $1 billion per year. This is Apple's largest annual AI spend ever — and it is going to a competitor.

“The computational cost of frontier AI models requires a partner with Google's data-center footprint. Apple's pragmatic admission is that building a 1.2 trillion parameter model in-house, today, is not feasible on the timeline users demand.” — Nerd Level Tech analysis, March 2026

What the New Siri Can Actually Do

The Gemini-powered Siri launching in iOS 26.4 is not an incremental update. Here are the headline capability changes:

New Siri Capabilities (iOS 26.4)

Screen Awareness: Siri can see and understand content across every app in real-time — no more copying and pasting text to ask about it.
128K context window (expandable to 1M tokens): Remembers the full conversation across weeks, not just the last few exchanges.
Sub-0.5 second response time: Down from the legacy 2–4 second latency that made Siri feel slow compared to ChatGPT.
92% complex task success rate:Up from 58% on multi-step queries like “find my last three emails from John, summarize them, and draft a reply.”
Cross-app workflows: Can pull data from Calendar, Mail, Photos, Notes, and third-party apps in a single query.

“Siri 2.0,” expected with iOS 27 in late 2026, will go further: proactive agentic intelligence that executes complex workflows without being asked — booking appointments, ordering groceries, and managing tasks autonomously.

The Privacy Architecture (And Its Limits)

Apple's partnership with Google raises an obvious question: does Google now see what you ask Siri? Apple's answer is no — and the architecture is technically sophisticated.

All Siri queries are processed through Apple's Private Cloud Compute (PCC)infrastructure: dedicated Apple Silicon servers that handle requests with stateless, ephemeral computation. Apple's PCC acts as a cryptographic intermediary — Google receives the processed query but not raw user data or device identifiers. Apple has released the technical specifications for PCC for independent security research.

Privacy researchers note a nuance: PCC is a new attack surface that did not exist before the Google deal. The stateless design reduces but does not eliminate risk. For most users, the privacy tradeoff of a dramatically more capable Siri is acceptable. For users with sensitive data needs, the calculus is different.

Siri Gets Smarter. Your Work Gets More Complex.

Even a Gemini-powered Siri is a general assistant built for quick tasks. Happycapy Pro gives you dedicated AI agents — research, writing, analysis, code, automation — with 50+ models for $17/month. For work that goes beyond “set a reminder.”

Try Happycapy Pro — $17/mo

AI Assistant Comparison: New Siri vs. ChatGPT vs. Claude vs. Happycapy

Feature	New Siri (iOS 26.4)	ChatGPT (Plus)	Claude (Pro)	Happycapy Pro
Model	Google Gemini 1.2T	GPT-5.4	Claude 4.6 Opus	50+ models incl. all three
Device integration	Deep (iOS-native)	Limited	Limited	Mac Bridge + integrations
Context window	128K → 1M tokens	1M tokens	200K tokens	Model-dependent
Agentic tasks	Coming (iOS 27)	Available now	Available now	Available now
Privacy	PCC buffer (Google receives query)	OpenAI servers	Anthropic servers	Multi-provider
Price	Included in iPhone	$20/mo	$20/mo	$17/mo
Non-Apple platforms	iOS/macOS only	All platforms	All platforms	All platforms

What This Means for Google, Apple, and the AI Race

This deal reshapes the AI landscape in three ways. First, it signals that frontier AI model training is now a natural monopoly problem — only a handful of organizations can afford to build and run 1 trillion+ parameter models. Apple, with $160 billion in cash reserves, decided to buy rather than build.

Second, it validates Google's Gemini as the preferred model for device integration. Android already runs Gemini. Now iOS does too. Samsung is deploying Gemini on 800 million devices. Google's distribution advantage — reaching users through Apple and Samsung simultaneously — creates a massive data and revenue flywheel that OpenAI, without device partnerships, cannot replicate.

Third, it is a direct revenue blow to OpenAI. The ChatGPT-Siri integration had put OpenAI on 2 billion iPhones. That integration is now subordinate to Gemini. ChatGPT remains available as an opt-in — but it is no longer the default, and defaults are everything in consumer software.

Every Device Getting Smarter. Use That to Your Advantage.

The new Siri handles quick tasks. Happycapy Pro handles your actual work — long-form research, complex writing, multi-step agent workflows, and integrations across your tools. 50+ models. $17/month.

Start Free — Try Happycapy

Frequently Asked Questions

Why is Apple paying Google to power Siri?

Apple determined that building a 1.2 trillion parameter frontier model in-house is too expensive and slow. Google's Gemini combined with Apple's Private Cloud Compute lets Apple deliver frontier capabilities immediately. The deal is worth ~$1 billion per year — cheaper and faster than building a competing model from scratch.

Does Apple's Gemini deal mean Google sees your Siri queries?

Apple says no. All queries are processed through Private Cloud Compute (PCC) — Apple Silicon servers with stateless, ephemeral computation. Google receives the processed query, not raw user data. Privacy researchers note PCC is a new attack surface, but for most users the tradeoff of a dramatically more capable Siri is acceptable.

When does the new Gemini-powered Siri arrive?

The initial rollout is in iOS 26.4, beginning March–April 2026 to all compatible iOS devices. The full 'Siri 2.0' with proactive agentic capabilities comes with iOS 27 in late 2026.

What happened to Apple's deal with OpenAI?

Effectively superseded. Apple used the OpenAI ChatGPT integration as a stopgap while negotiating the deeper Gemini deal. ChatGPT remains available as an opt-in in Siri settings, but Gemini is now the primary AI engine. Apple's $1 billion annual payment to Google dwarfs any arrangement with OpenAI.

Sources

Bloomberg: “Apple Plans to Use 1.2 Trillion Parameter Google Gemini Model to Power New Siri” (November 5, 2025)
9to5Mac: “Apple to unveil results of Google Gemini partnership as soon as next month” (January 25, 2026)
Gadget Hacks: “Apple Siri Gets $1B Google Gemini AI Upgrade in 2026” (January 28, 2026)
Nerd Level Tech: “Apple's Siri AI Overhaul: The Gemini Deal in 2026” (March 25, 2026)
udit.co: “Apple Redesigns Siri with Google's Trillion-Parameter Gemini” (March 2026)
FinancialContent: “Apple Inks $1 Billion Deal with Google to Power Gemini-Fueled Siri Revamp” (February 6, 2026)